ABSTRACT. "We live in interesting times. Information technology has become an integral element of all work; rapid advances in technology is leading to continuous redefinition of the why, what and how of all work and businesses. This presents significant challenges as well as opportunities for everyone, enterprises and individuals alike. In this presentation, we will discuss imperatives for enterprises and individuals to stay relevant and thrive in this rapidly changing world. "

ABSTRACT. A decade of activity in the field of data science at high pace has resulted in significant value creation for some businesses. On the flip side, various statistics show that data science projects are not successful. In this talk, I will discuss various aspects of data science with some case studies from simple industries to highly complex industries. The talk will be centered around CORE values : Challenges, Opportunities, Reach and Evolution for transformation in today's era of Industry 4.0 and beyond.

ABSTRACT. Large time series collections are common in many domains, including natural and social sciences, IoT applications, cloud computing, and supply chains. These datasets are immensely important for improving forecasting and anomaly identification, and ultimately decision making. Modern datasets can contain millions of associated time series, making them extremely high dimensional (one dimension for each individual time-series). To improve prediction, global patterns must be combined with local calibration. But probabilistic forecasting of these dependent time series collections is challenging. Time series methods can't capture complex data patterns, while multivariate techniques struggle to scale. Based on strong structural assumptions, they are data efficient and provide uncertainty estimates. Contrarily, deep neural network models may learn complex patterns and correlations given enough input. This article proposes a hybrid model that incorporates both approaches' benefits. A latent, global, and deep component makes our new method data-driven and scalable. It also controls uncertainty using a conventional local model. Our strategy is validated both conceptually and experimentally by decomposing exchangeable time series into global and local components. Our findings confirm the model's accuracy and computational complexity.

ABSTRACT. The Olympics are one of the leading international sporting events, which are broadly classified into summer and winter sports. Being one of the toughest competitions, where an enormous number of athletes from various parts of the world participate in a diversity of competitions. These games are considered to be the oldest events which also makes them one of the world's foremost sports competitions, where we witnessed active participation of 205 nations over the globe in the 2020 Tokyo Olympics. During the study for a suitable model, we discovered that there are several socio-economic factors/variables that are good predictors and are significantly impacting a nation’s Olympic success. There were initially 10 features in the model, which were further reduced, based on the techniques of feature selection. The prediction of the Medal Tally is a difficult task, as the distribution of classes for all attributes is not separated linearly and is caused by different scales. Regression[15] techniques like Linear, Polynomial, Ridge, Lasso, Bayesian etc, have been compared on various performance criterion for prediction of Medal tally in this work. Our source data for medal prediction was taken from Official Olympic website and Wikipedia

ABSTRACT. Ayurveda is an ancient concept that believes in holistic healing using herbs. Ayurveda is made up of two words “Ayu” which means life and “Veda” which means knowledge. As per the Ayurveda “Tridosha” principle used for determining the “Prakriti” of a person. Prakriti is a composition of Panchmahabhutas (Five Elements) categorized into three doshas. If three doshas are balanced, then a person is healthy otherwise he/she is prone to diseases. Ayurveda recommends a specific diet, exercise, and medicine that can restore the balance in this Prakriti to provide health. Tridosha is the base concept in the Prakriti Nishchitikaran (Prakriti Certification) and has been studied for a long time; the quantitative reliability for Prakriti Nishchitikaran is studied in this research paper using supervised Machine Learning algorithms. This paper helps to identify the main Prakrities - “Vataj” (V), “Pittaj” (P), “Kaphaj” (K) and subtypes - “Vataj Pittaj” (VP), “Pittaj Kaphaj” (PK), “Vataj Kaphaj” (VK) and “Vataj Pittaj Kaphaj” (VPK).

ABSTRACT. This paper presents a machine learning based system to classify Electroencephalogram (EEG) signal for walking and sitting actions. This system is proposed to generate control signal for the artificial limb which is aimed help the limb amputees to carry out movements. EEG data is collected from 8 healthy male subjects with a mean age of 21 years while they performed walking and sitting actions. The montage activity was captures using 7 electrodes considering frontal, central and parietal regions. The variations of the EEG signal contain information of the physical activity being performed. the Statistical features namely Mean, Minimum, Maximum, Kurtosis, Standard Deviation, and Skewness are extracted from EEG signal. An array of five classifiers Linear Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Decision Tree and Random Forest was used for activity recognition. Random Forest provided 100% accuracy of classification.

ABSTRACT. Over the past decade, online shopping has become the norm. Everything that a person needs is available in online platforms such as Amazon, Flipkart, etc. There is no dearth in the quantity and categories of products available in these online platforms. Given this influx of products in these online marketplaces and a rapid shift shown by people towards online shopping, it becomes imperative for an efficient product recommendation system to be in place in order to make product recommendation for a user as effective and seamless as possible. In this paper, a method for a recommendation system that is based on the aspects present in the user inputs and reviews is proposed, implemented and evaluated. As opposed to just depending on user ratings and the sentiment scores and orientation (positive versus negative) of user reviews to recommend products, the proposed method performs aspect extraction of the given user reviews, calculates the sentiment associated with the aspects along with its positive and negative connotation and orientation and based on this, recommends relevant products to the user based on the requirements that the user enters in our system. The classifiers used for this purpose are k-NN, SVM, SentiWordNet and NLTK VADER. Thus, based on the product requirements entered by the end user in our recommendation system, relevant products are recommended to the user.

ABSTRACT. Document clustering is the task of organizing textual content into groups so that they are more similar to one another than to those in other groups. Several text clustering algorithms have been proposed recently by various researchers. However, the majority of them limited their research to English-language documents. Odia is the language spoken by the people of odisha and its appearance on the digital platform is on the rise recently. This paper proposes an optimized feature representation using PCA of Odia documents for efficient document clustering. The proposed work first extracts four different features from Odia sentences: word-level TF-IDF, character-level TF-IDF, word-embedding, and sentence embedding vectors. With a Silhouette Coefficient of 0.964, Rand Index of 0.352, Normalized Mutual Information score of 0.001, and Davies-Bouldin Index of 0.022, it was found that the use of PCA-based optimized word-level TF-IDF features performed better than other feature representations.

ABSTRACT. Abstract Peptides have garnered interest as therapeutic agents owing to several beneficial advantages. However, a major downside of peptides as therapeutic agents is that they have a higher probability of inducing inflammatory response (proinflammatory). Proinflammatory peptides also serve as important targets for developing therapeutic strategies. Consequently, identifying peptides that are proinflammatory in nature is important. Here, we develop a machine learning method that uses peptide sequence information and employs distributed representation of protein sequences in reduced alphabet form to predict proinflammatory peptides. The reduced alphabet schemes used include Kyte-Doolittle hydrophobicity scale, conformational similarity index and categorization based on chemical properties that was successfully used in predicting ligand binding sites. In the distributed representation method, word embeddings are extracted by treating the protein sequence as a text in a corpus. These word embeddings are then used as input features to a classification method. In our work, we used the reduced alphabet representation of the protein sequences to extract the word embeddings. These word embeddings are then given as input features to the Support Vector Machine classifier. We also identified motifs that are found in the proinflammatory sequences and used them as input features in addition to the word embeddings. Our results show that the combination of motif information and distributed reduced alphabet representation achieves a significantly higher MCC value and accuracy than the existing methodologies. Thus, our work underscores the importance of incorporating functional information (motifs) and the choice of reduced alphabets in creating more accurate predictive models for proinflammatory sequences.

ABSTRACT. As declared in Sustainable Development Goal (SDG) 8, the economy, especially economic growth is very important to generate wealth, overcome poverty and pave the way for more sustainability. For this paper, 118 different economies from underdeveloped to developed countries were analyzed and the results were visualized using principal component analysis and k-means clustering. The principal component analysis allows ranking the different countries not only by one economic indicator but a combination of all indicators. It also allows to visualize the results by reducing the variables. Two countries, Germany and Sierra Leone, from different clusters, were picked out to perform a forecast using an ARIMA model. The annual GDP per capita growth rate was predicted for those two countries and the global economy to provide further insight into the future development of those economies.

ABSTRACT. Wildfires pose a major risk to humans and other species, but thanks to advances in remote sensing techniques, they are now being continuously observed and regulated. The existence of wildfires in the environment is indicated by the deposition of smoke in the atmosphere. Observation of fire is critical in fire alarm systems for reducing losses and other fire hazards with social consequences. To avoid massive fires, effective detectors from visual scenarios are crucial. A convolution neural network (CNN)-based system has been used to improve fire detection accuracy. Separating data into training and testing subsets is a vital aspect of the Inception-v3 architecture [5]. By default, a maximal and minimal amount of data is used for training and testing, and accuracy vs loss graphs for training and testing data are plotted for data visualization.

ABSTRACT. Clustering of an evolving data stream is always very complex as data objects are not static in nature. To identify the real-time clusters is a task. The characteristic of the data objects in an evolving data stream is also evolving. If these evolving pat-terns of the data objects are labelled and defined it may be possible to identify the context change and such repetitive change in the context will explore the trends in the evolving data stream. However, it is necessary to understand the occurrence, re-occurrence and diminishing property of these clusters over time. Improved Affinity Propagation (IMAP) Clustering algorithm identifies these changes in the characteristics of the clusters over time. Continuous observation and registration of new clusters leads us to identify the trends in the evolving data stream. Analyzing these trends may able to understand the patterns or change in the context so that future prediction is possible. The proposed IMAP algorithm is more efficient over pre-estimation of number of clusters, and probing at right time the evolving data stream over time without any loss of information. The Algorithms is also robust in identifying the outliers. Experiments on real data sets are presented to demonstrate the benefits of the trend analysis method.

ABSTRACT. Aircraft Maintenance is an advanced task requiring highly skilled engineers. Facilitating aircraft maintenance by providing proper tools and equipment is essential in ensuring good maintenance work. Aircraft Maintenance Technicians (AMTs) require precise knowledge and customized tools to perform their duties. They are responsible for an airplane’s safety and efficiency, and rely on a few basic pieces of equipment for a wide range of jobs pertaining to airplane maintenance. Specific maintenance tasks require unique tools. And while the AMTs could probably improvise and get the job done anyway, specialized tools exist for a reason — they help get the job done correctly and improvising will lead to unnecessary labor and a compromised aircraft. For example, an incorrectly sized screwdriver or screw causes wear and tear and makes the job harder. Besides, traditional tool management requires employees to manually check in and out each tool, which is time consuming. A Tool Selector app which recognizes and tags tools in real time will help AMTs in determining how it is used in a particular task. Through this app, the AMTs can be guided through animations to perform specific tasks, such as replacement of Oil Filter from an aircraft engine.

ABSTRACT. Text summarization generates a summary that highlights key sentences while condensing all information from the original document into a few sentences. Abstractive and Extractive are the two types of summaries seen in general text summarization process. A hybrid summarization technique which generates an abstractive summary over an extractive summary is proposed in this paper. The initial phase incorporates the use of a semantic model to generate word embeddings for each sentence and these embeddings are used to improve the lack of semantic disintergrity in extractive summaries. The second phase takes in the concept of WordNet, Lesk algorithm and POS tagging for generating an abstractive summary from the extractive summary. The paper uses two different data sets: DUC 2004 and Daily mail/CNN for evaluating the performance over ROUGE and BLEU metric. The results highlight the relevance of developing hybrid approaches to summarization compared to complex abstractive techniques.

ABSTRACT. Depression is a genuine medical condition characterized by lethargy, suicidal thoughts, trouble in concentrating, and a general state of disarray. It is both a "biological brain disorder" and a psychological state of the mind. The World Health Organization (WHO) estimates over 280 million people worldwide suffer from depression, regardless of their culture, caste, religion, or their whereabouts. Depression affects the way a person thinks, speaks, or communicates with the outside world. The key objective of this study was to try to identify and use those differences in linguistics in Reddit posts to determine if a person may be suffering from depressive disorders. This paper proposes novel Natural Language Processing (NLP) techniques and Machine Learning approaches to train and evaluate the models. The proposed textual context-aware depression detection methodology consists of a hybrid transformer network consisting of Bidirectional Encoder Representations from Transformers (BERT) and Bidirectional Long Short-Term Memory (Bi-LSTM) with a Multi Layered Perceptron (MLP) attached in the end to classify depression indicative texts that can achieve incredible results in terms of accuracy - 0.9548, precision - 0.9706, recall - 0.9745 and F1 score - 0.9725.

ABSTRACT. Chemical pesticide spraying over a large area is not only a waste of herbicides and labor, but it also pollutes the environment and compromises food quality. As a result, properly identifying weeds and spraying them are critical tactics for increasing agricultural sustainability. In this work, the networks were trained using a dataset containing 17,509 labelled images of weeds native to Australia. Despite the advances of Deep learning models, it still faces challenges like high computational power, overfitting, need of a balanced labeled dataset, therefore transformers used mainly in Natural language processing tasks can be considered as a possible replacement for CNNs. This paper aims to analyze and compare Vision Transformer with CNN models for achieving good results on weed detection and classification from crop images. As the final results of the experiment, the Vision transformer gave the highest accuracy of 96.41% followed by the ResNet-50 model which gave an accuracy of 95.70% compared to 95.04% of the Xception model. Inception V3 model gave an accuracy of 94.7% and Inception-Resnet V2 gave an accuracy of 94.15% on the dataset images.

ABSTRACT. Generating a meaningful summary for the given natural language text is one of the challenging and popular tasks in the present era. Researchers have come up with various techniques for abstractive and extractive summarization. This experimental study is focused on the extractive summarization. In graph-based extractive text summarization techniques, the sentences of the input document are used as the nodes of the graph, and various similarity measurements are used to weight the edges of the graph. Each node's rating is determined using the graph ranking algorithms, and the top-ranked nodes (sentences) are then added to the output extractive summary. In this work, we first translate the publicly available dataset into Hindi text using the Google Translate service. Next, we apply a pre-trained multi-lingual transformer to generate embedding vectors of each sentence of the document. We use these embedding vectors as the nodes of the graph. The rest of the approach remains unchanged. At last, we evaluate the generated extractive summaries on the basis on ROUGE score. Evaluation results indicate that the use of pre-trained multi-lingual transformer can be effective in generating more meaningful extractive summaries.

ABSTRACT. This is Electronic voting system based on blockchain technology. This will help Voters to vote secretly and more confidently. In this research paper we wrote all the problems and errors we faced to develop this project

ABSTRACT. Senior management commitment, risk management, and flexibility improve project responsiveness to volatile and high-impact scenarios, especially in large projects and programs. The aim of this study is to determine how project flexibility interacts with and affects the relationship between senior management commitment, project risk management, and success in IT projects. A cross-sectional survey of 166 managers was used to derive empirical data from the financial services industry and used to test the conceptual framework based on recent project management literature. Ordinal regression analysis demonstrated a significant relationship between senior management commitment, project risk management, and success in projects which is influenced by significantly positive moderations established through flexibility in projects. The study findings can assist project managers and senior leaders to accomplish their short-term and long-term project goals and achieve success in projects by reducing the chances of failures. This paper adds value to existing research in the context of IT projects and the role of project flexibility on their performance.

ABSTRACT. Artificial Intelligence has emerged as one of the very highly researched areas in recent times. The antenna research community has been exploring AI techniques for the design and development of various types of antennas for communication, radar, aerospace, and other applications for several years. This research has gained large significance in the last few decades due to the emergence of many wireless devices used in communications and computing systems, health monitoring systems, remotely operated robotic systems, and strategic systems including those used for Radar and Electronic warfare. The recent growth of interest in 5G, MIMO, IoT, and RF ID devices has also been a stimulant for adapting AI techniques for novel antenna designs. The present study reviews about 6000 documents over the period 1991-2020 to look at the growth profile of Antenna research in relation to various Artificial Intelligence techniques. The paper provides a landscape view of use of the AI techniques in antenna research and identifies research themes and trends, Knowledge Gaps, and Linkages between the different techniques. The study will be useful for new researchers to help plan their research activity, for experienced antenna engineers to trace the growth profile of antenna research related to different AI techniques, and for other researchers to trace AI applications in different technologies.

ABSTRACT. The Android operating system (OS) currently occupies the majority of the global smartphone market. Even IoT specific applications have prevailing OS as Android into their end device or intermediary communication channels. These Android devices may retain confidential information such as SMS, contacts, banking information, Personal Identification Number (PIN), location-specific information, photographs, videos, IoT devices workflow and so on. Furthermore, Android devices are popular among users due to their extensive capabilities and multiple connectivity options, making them a perfect target for attackers. To get their task done, attackers are shifting to methods that neatly disguise existing state-of-the-art equipment and targets. One such strategy is evasion, which is used to deceive security systems or conceal information flow in order to evade detection. Covert channels, on the other hand, disguise the existence of communication itself, making it unidentifiable to both users and cutting-edge technology. These covert channels, by employing evasive methods, become extremely undetectable and bypass security architecture, ensuring the security of the user's data. The research evaluates and analyses existing state-of-the-art technologies, as well as identifies potential defense mechanisms for mitigating and detecting such threats.

ABSTRACT. With the availability of android phones and internet in every hand, the rate of cyber crimes have increased tremendously. This increase in cyber crimes has led to the need for effective cyber training in all fields. Awareness and training are strongly needed at this point of time. Cyber training is a much debated but the least worked on topic. Everybody talks about this, but very few take steps in this regard. The reason behind this is the lack of expertise and knowledge in the field. The issue of cyber training against social engineering attacks remain in the forefront of organisations. But when we enquire into the larger context of its implementation, then it is revealed that no substantial measure is being taken due to low human resource in the field. Even after so many trainings that are provided there is no valuable decrease in the number of cases of hacking. This paper aims to bring about the ongoing developments in this field and also the vulnerabilities in the organizational structure.

ABSTRACT. In this work, a quantum web-based health analytics system has been proposed. The system uses a Neuro-Fuzzy algorithm which uses a selective rule-based strategy to process data sets. When a query is being raised, the datasets required are identified and processed to formulate inference for a knowledge repository. The knowledge repository is decentralized and ensure authentic access for a particular query. The quantum block chain algorithm reduces processing time and memory consumed by 24% and 28%, compared to the conventional block chain approach.

ABSTRACT. The use of automobiles has increased dramatically in today's globe. As a result of the increased use of automobiles, traffic has increased, and road accidents have increased. Because of the lack of rapid safety facilities, this has a negative impact on the property as well as human life. Although it is impossible to completely eliminate accidents, the consequences can be minimized. The system that is proposed is effective. A concerted attempt to serve victims with emergency services in the quickest possible time The drivers in large corporations engage in criminal behavior. As a result, the organization suffers monetary and time losses. Aside from these uses, the technology can also be used to track people. Vehicle theft, travel luggage, fleet management, and vehicle sales, to name a few examples. A single embedded board is used in the system. GPS and GSM modems are coupled to a microcontroller in this setup. The full setup is already in place in the vehicle. It makes use of a MPU6050 a six axis accelero-gyroscope sensor. It detects movement at the site where it is installed. After that, the signal is compared to the standard values. which also confers the car's accident, unneeded shock or vibration caused by machines, and the car's tilt with relation to the road. The level of acceleration can be used to determine the earth's axis. The Global Positioning System (GPS) is used to locate objects. GSM is utilized to send precoded numbers the exact location of the vehicle. Longitude and latitude will be included in the message. The location of the accident can be established using these values. By using a SIM card, a GSM modem allows for two-way communication. A module like this, functions in the same way as a standard phone. The project intends to create an intelligent security system that provides situational awareness. vigilance and nimble safety

ABSTRACT. India is an independent country with Population of 1.3 billion. But there are few people who are still not independent and they have to rely on other people for help. Yes, the Physically Challenged People of India constitute almost 2.1% of entire Indian Populace. To help these people, as budding Engineers, we thought it was our duty to help these People out and thus created the Bluetooth controlled Wheelchair, so that they won’t have to rely on the surrounding people for help, atleast most of the times.

ABSTRACT. Methane is the primary contributor to ground-level ozone, a hazardous air pollutant and greenhouse gas that potentially kills one million people each year. It is 80 times more potent than carbon dioxide at warming over a 20-year period. There is a critical demand for timely methane leak detection in remote places throughout the energy sector. Algorithms such as matched filters are commonly used to detect affected areas. These algorithms use the spectral signature of methane and try to match it with the signature of each pixel in a hyperspectral image. Considering the complexity of the situation and the performance variation of these algorithms, it is essential to compare them in order to find a quick, effective, and reliable method among these to detect methane leaks. In this paper, we compared three algorithms for identifying methane contaminated areas: RX-algorithm, Matched Filter, and Adaptive Cosine/Coherence Estimator (ACE). We first tested these algorithms on synthetic data and then, we used them to detect methane contaminated areas of the Santa Susana Mountains near Porter Ranch, Los Angeles, California. This methane leak disaster is known as Aliso Canyon gas leak (also called Porter Ranch gas leak). Hyperspectral data for this region was acquired by NASA using Airborne Visible/Infrared Imaging Spectrometer (AVIRIS).

ABSTRACT. The security of modern cryptographic systems mostly depends on mathematical hardness to solve a particular problem manually. The most popular and widely used classical cryptosystem, RSA (Rivest, Shamir, Alderman) is based on the mathematical hardness of prime factorization of the product of two large prime by any of the present classical processors. But the advancement in technology particularly in quantum information provides a threat to the present communication protocols. It motivates the researchers to move to new technologies that are fundamentally more secure through the principle of quantum mechanics. Such protocols rely upon quantum bits or qubits for data transmissions securely. The application of qubits instead of bits in cryptography using quantum key distribution protocols makes encryption robust But the quantum communication protocols are prone to noise due to environmental factors, fragile and traverse’ a long fiber optical path. Hence, the states of qubits are easily tampered with by external factors. Noise alters the information, eventually, the message may not be meaningful to receivers. Except for bit flip and phase flip noises, depolarization, amplitude damping, decoherence, thermal relaxation, and phase damping are obstacles to achieving long-distance communication. The objective of this paper is to present a comparative study of quantum communication protocols in a noisy environment with different classes of noise.

ABSTRACT. The talk will survey a series of results in Multi-Agent Swarm Robotics, concerning the design of local interactions that ensure the emergrence of desired global behaviors.

ABSTRACT. The virus has been a debatable topic for decades as it cannot be placed anywhere in the classification, as viruses are neither dead nor alive. These are small parasites that require the host body to replicate. A virus consists of DNA or RNA, either single or double-stranded or circular or linear. They also have a protective covering made of a protein coat called a capsid. There are two different methods by which we can analyze the genomic functions either by direct experiments or by computational methods. The computational techniques are slowly taking over the research. In this paper, we have tried to find out the hot regions present in the Covid-19 genome using Continuous Wavelet Transform (CWT) using different wavelets at different scales. As a result, we are able to find the mutation sites in the Covid-19 genome.

ABSTRACT. The goal of the research work is to create a prediction model for brain stroke. People who live at higher altitudes have a lower risk of dying from strokes. The effect can be found at between 2,000 and 3,500 meters. When the blood supply to part of your brain is disrupted, a stroke can happen. The brain cells begin to die. When the blood flow to even a single part of the heart is blocked leads to heart attack. The muscles of the heart dies if there is not enough oxygenated blood. A stroke happens when a blood vessel in the brain is damaged. Machine learning is being applied to the healthcare system to predict diseases early. Data is the main necessity of Artificial intelligence. The dataset is used to build a machine learning model. The aim is to predict the chances of stroke using Machine Learning techniques. A comparison is made for better accuracy with the use of four different algorithms. Aim is to create an application that is easy to use and navigate.

ABSTRACT. Statistics show that 39 million people living in this world are blind and 246 million people have poor vision. Although a lot of hardware products and mobile devices are developed to provide them with useful features, their effectiveness and practicability don't help visually impaired people to the fullest. They are dependent on people around them for their daily needs. If no one is available next to them, they become helpless to do many simple tasks on their own. Thus, the target audience of the work presented in this paper is the visually impaired people, who with the help of The Third Eye android mobile application can become more self-reliant. They will be able to identify objects around them, find the object they need with the Yolo object detection algorithm; learn whatever they wish to be read with the optical character recognition algorithm; and operate an android application, use mobile internal features completely on their own using text to speech converter and voice listener functionalities. This android mobile application eliminates their need for a person around them and makes their life smart and simple. It is developed using multiple domain ideas and efficient algorithms that are well designed and implemented to be an ideal product for visually impaired people. This is a great development of technology that helps them to overcome the challenges they face in their daily life.

ABSTRACT. Nowadays, Machine Learning (ML) plays the important role in the E-commerce industry and its customer relations to perform different kinds of tasks such as prediction of purchases, segmentation of customers according to their reviews/sentiments, recommendation of products to the active users etc... various ML algorithms are implemented to get trained with data patterns to perform the above-mentioned tasks. In this paper, the customer segmentation and recommendation of women’s clothing based on the reviews are presented. The comparative study is done using five different ML algorithms namely Regression analysis, Naïve Bayes, Decision Trees, Support Vector Machines, and Clustering analysis. The results show that the Naïve Bayes algorithm has better performance when compared to other algorithms by showing better accuracy.

ABSTRACT. Protein structure with the contact maps of atoms is predicted in neural network model. The model or the data sample is collected form CASP13 database. Segregation of data sample into alpha fold separately. Hence, the orientation of each molecule and contact maps the prediction of inter residues atoms present in alpha or beta fold is discovered with the neural networks. Neural network model training the input data set of alpha fold protein structure and inter connecting residues with atomic orientation of molecules is predicted. We validate the entire training and test set with high level accuracy in atomic residues with similar structures. The neural network yield high percentage of efficiency in predicting the atomic structure compare to the other model. The accuracy of model is demonstrated with the previous experimental model is comparatively high in neural network model. The novel approach of deep learning incorporates the knowledge about the protein alignment of amino acid residues prognosis for biological genomes in an efficient technique.

ABSTRACT. This study utilizes cross-country survey data to analyze differences in attitudes towards cryptocurrency as an alternative to traditional money issued by a central bank. Particularly, we investigate women’s general attitude towards cryptocurrency systems. Results suggest that women invest less into cryptocurrency, show less interest in future cryptocurrency investment, and see less economic potential in these systems than men do. Further evidence shows that these attitudes are directly connected with lower literacy in cryptocurrency systems. These findings support theory on gender differences in investment behavior. We contribute to the existing literature by conducting a cross-country survey on cryptocurrency attitudes in Europe and Asia, and hence show that this gender effect is robust across these cultures.

ABSTRACT. When someone arrives at a new place, the first thing that they do is look for a place to dwell, and when the company they are with is not appropriate, it leads to disputes and sometimes shifting to a new place which can be very tedious. Finding the right roommate is very important as it affects the physical and mental health of a being. The present solutions in the market for this problem include websites like roomster.com, olx.in, indianroommates.in, etc., and applications like FlatMatch, Roomster, etc. A detailed analysis of potential competition was done in order to figure out our standing amongst them. The paper analyzes them on the basis of their features, ratings from users, etc. Roommate-finding platforms exist, but they just display a list of users without considering their preferences. This is where Troomate has an advantage. A detailed literature survey was done about how the pairing of people could be done based on their idea of the perfect roommate. This paper includes various algorithms like Gale-Shapley, Elo rating score, and techniques like clustering in order to effectively match on the basis of powerful filters like social traits, diet habits, sleeping schedules, etc. With an interactive, well-designed UI, dependable backend, and reliable algorithms, Troomate aims at solving the problem effectively.

ABSTRACT. The steel industry is one of the most risk and accident-prone industries. The workers meet with a lot of accidents due to being exposed to hazards present in the workplace. Safety strategies are used to prevent accidents from occurring in the steel industry. Such safety strategies require certain costs. Benefits are derived from incurring such costs on the safety strategy of accident prevention. Cost-Benefit Analysis is required to establish the feasibility of such strategies. This paper aims to discuss the relationship between costs and benefits. Two safety strategies were taken and linear regression modeling was utilized as a statistical tool to find out the extent to which costs (direct and indirect) and benefits of accident prevention are related. From the analysis, it can be concluded that the costs have a positive impact on the benefits of accident prevention. It is also observed that the indirect costs are more significant than the direct costs.

ABSTRACT. Monitoring urban floods caused by heavy rains is critical for planning rescue and recovery responses. Due to climate change, water logging issues in coastal cities are getting worse as economic development activities speed up. Moreover, the water-logging risk zoning map offers crucial decision support for the management of urban water-logging, urban development, and urban planning. Urban water-logging will affect economic growth and people’s life safety. Synthetic Aperture Radar (SAR) is not affected by cloud cover, unlike optical imagery, and provides the required data under cloudy conditions, so SAR provides an efficient method to monitor changes in water bodies across wide areas and identify water-logging brought on by heavy rain. In the present work, we discuss the initial exploration results of the Sentinel-1A SAR GRD data for Mumbai on July 2, 2019. Mumbai experienced a severe flood on July 2 because of heavy rains, and many suburbs of Mumbai city were affected. The SAR images show enhanced double bounce because of the floodwater, which correlated well with the flood news reports. Further investigation is required using complete polarimetric data.

ABSTRACT. India is one of the topmost producers of agricultural products; however, this productivity is not at the optimal level. To increase the productivity and quality of the crop, there is a need to identify the factors that can improve the existing agriculture scenario. One of the most influential factors in crop production is the selection of the best suitable crop for geographical-geological conditions. But there is not enough scientific training on agriculture literacy in the farmer community in this regard. Most of the farming practice is based on conventional routine methods. The way out of these issues is to handle these problems computationally. Machine learning algorithms can really play an important role in bridging the knowledge gap between agriculture experts and farmers. The objective of the proposed research work is to propose a crop recommender that can suggest the best suitable crop for given geographical-geological conditions. So it is anticipated to have a standard dataset that acts as a domain expert to recommend the decision. Then, there is a computational process, which utilizes this agriculture dataset and applies machine learning techniques to develop trained models. Test data is provided to these trained models for the prediction of the crop. In this paper, we have proposed a crop recommender using statistical measures Mean, Standard Deviation, Skewness, Kurtosis, Peak2Peak, and RMS features and machine learning algorithms. The statistical features extracted from the training data are used by the Naïve Bayes (NB), Decision tree (DT), K-nearest neighbor, KNN classifiers, and Hybrid ensemble model for crop recommendations. The Performance of crop recommendations with and without statistical features is also compared. For all machine learning algorithms under consideration, the performance of crop recommendations with statistical features is better for the hybrid ensemble model.

ABSTRACT. Early detection of violent material such as a Molotov cocktail, firelight, etc., carried by individuals in a mob, or the presence of any fire in a crowd may aid security personnel in taking further necessary actions in managing the group. Many recent automated fire detection techniques have been proposed using machine learning and deep learning techniques. However, most of these methods suffer from a high rate of false alarms, slow detection, extraction of hand-crafted features, localization of fire region, and less accuracy. In this research work, we propose a YOLO (You only look once) based fire detection technique for fast, real-time classification and positioning of fire objects in crowd images. An extensive set of annotated fire images is required to train data-intensive CNN (Convolutional Neural Network) architectures robustly. Hence, a method for automatically annotating fire images based on color and HSV channel characteristics using morphological image processing operations is proposed, along with data augmentation techniques. A customized fire image dataset, "DIAT-FireDS," is created using the web scraping technique and the proposed annotation technique. Generated customized dataset is used for training and fine-tuning YOLO architectures. Experiments are conducted using a series of YOLO architectures against the standard and presented dataset to achieve real-time detection accuracy of about 0.74 mAP.

ABSTRACT. This paper suggests dual two staged novel fine grain rotated network for aerial solar panel health classification. The neural network architecture can detect different types of uncleared solar panels of any arbitrary orientation installed in various environments. Three different types of solar panels were identified and categorized based on accumulation of dust. A synthetic dataset was generated to assess the precession of the solar panels' detection in various aerial spatial situations. Furthermore, no datasets were available to support this research. An amalgamation of two datasets were used to draw a conclusion. we primarily focus on identifying different types of dirty PV panels, and we specifically address dust accumulation. The model proposed in this work is effective because it is extremely sensitive to the collection of dust which are hard to detect from an aerial field of view.

ABSTRACT. As healthcare has become more data-driven in recent years, the amount of data produced has increased. Digital data can take the form of audio, pictures, videos, transcripts, clinical records, electronic medical records, and free text. More and more clinical records are being created as a result of the development of information technology systems, and these records need to be processed and examined. Examining and interpreting medical data can be a challenging task that takes a significant amount of time, resources, and human effort. It takes a medical expert to complete the laborious work of assessing a large volume of data. Therefore, artificial intelligence technologies are being used to analyze data in healthcare. The main aim is to build a multi-label classification system that predicts the morbidities that may occur in the future taking clinical notes as input. BERT model exploiting transformer architecture is used to deal with constraints of the small datasets and improve the performance of the model.

ABSTRACT. There are several websites today that compare products. However, the majority of them only use textual data. Through the creation of Content-based image retrieval (CBIR), the visual software for product photos presented in this work provides a revolutionary technique for visually locating products. Consumers' comparison-purchasing strategy is influenced by product value, complexity, and durability. Comparing prices is frequently used to describe contrast searches as a whole. Contrast searching is expanding beyond just finding the cheapest greenest goods online, though. Nowadays, a vast variety of things are available online. Customers can access various records about the products they are interested in by using production advice structures. The core of this problem from a computer and technology perspective is to extract information from the items so that it can be utilized to match the related products. In order to fit a collection of goods gathered in a database, this work provides a method that merges records from items and, as a result, the outline of the products (textual format). With the help of the comprehension offered by the usage of CNN technique, the definition of the products is integrated. A simple weighting technique is used to integrate photo and text content information. Cosine similarity measurement is used to execute the matching. Online retailers provided the information. The processing of the product reviews has been done using NLP. Android app to compare product prices while using reviews and pricing as help.

ABSTRACT. Failure mode and effects analysis (FMEA) is an engineering analytical technique that has found application in the identification of failure modes through investigation of a system and quantification of the risk of occurrence of each failure mode and eliminating them. Failure modes are evaluated in terms of three risk criteria, namely, severity (S), occurrence (O), and non-detectability (D), and prioritized based on their risk priority number (RPN), that is equal to the product of the three risk criteria. However, conventional RPN has a number of drawbacks. This paper proposes a framework that involves a clustering-based approach for prioritizing the failure modes developed and applied to a particular system defined in an integrated steel plant. First, the relative significances of the risk criteria in terms of their weights were computed using the Fuzzy-Full Consistency Method (F-FUCOM). Linguistic terms were considered for comparison among criteria and rating every failure mode in terms of S, O, and D, which were subsequently converted to Triangular Fuzzy Numbers (TFN). Then the failure modes were clustered in the fuzzy environment into four groups using K-means, Agglomerative and Fuzzy C-means (FCM) algorithms, based on weighted Euclidean distance. The results obtained by the clustering algorithms were finally compared. While calculating the weights of the risk criteria, S was considered the most important. Four clusters were formed and named Minimal, Moderate, Major, and Extreme, based on the increasing centroid RPN values respectively. The largest number of failure modes were clustered in the Major-risk cluster followed by the Extreme-risk cluster. Finally, K-means clustering was found to be the superior clustering algorithm of the three.

ABSTRACT. Portfolio optimization is defined as the process of assets allocation to achieve optimum expected returns and/or minimizing financial risk associated. It is crucial for a financial risk manager to provide the best returns possible in the market and calculation of risks like value-at-risk. The problem of portfolio optimization is not new to the financial world and approaches like efficient frontier is already known. While the work on optimization of portfolio is voluminous, this paper describes the portfolio optimization approach using reinforcement learning. This approach is particularly useful in the case where the search space is very large, and distribution of asset must be done in real-time. Algorithmic trading could be a good candidate for this optimization approach. In this paper, the description is given for a few famous optimization algorithms also that is used in the financial industry. The main idea of this RL (reinforcement learning) based approach is that agent learns the weights distribution across the portfolio by rewarding and punishing the weights ratio and by continuously doing so it can produce real-time distribution of the weights. The study has been done on a portfolio of four stocks though can be extends to any number of stocks in the portfolio.

ABSTRACT. Fetal brain segmentation has been a field of interest since a long time. However, it is a challenging task as well for reasons, like blurred images due to fetal motion. Recently deep learning has been successful in performing this task with good accuracy. In this paper, we developed 2-way Ensemble U-Net model, a Convolutional Neural Network architecture for performing segmentation on the fetal brain image to divide it into its seven components: Intracranial space and extra-axial Cerebrospinal Fluid spaces, gray matter, White matter, Ventricles, Cerebellum, Deep Gray Matter, and Brainstem and Spinal Cord. The fetal brain image can be obtained by segmenting it from the fetal Magnetic Resonance Images using any of the previous works on fetal brain segmentation, which presents our work as an extension of the already existing segmentation works. The Jaccard Similarity and Dice Score for this task are 83% and 88% respectively. This is higher than that returned by any of the previous models, when trained for the same task, thus showing the potential of our model in segmentation related tasks.

ABSTRACT. Medical coding is an integral part of the healthcare process. Translation of medical procedures and diagnoses described in a patient’s medical record into codes used for billing and claims. This process is essential for reimbursements, provides data that can be used for tracking, and promotes consistency throughout the medical field. In this paper, we determine the medical codes using the derived diagnosis from medical narrations. We will discuss the process to convert ICD10-CM index and tabular data into hierarchical graph and derive the medical codes. We propose a knowledge graph-based solution that facilitates interoperability without sacrificing accuracy.

ABSTRACT. Quality classification of artifacts such as weld-joints depends on various factors which may not be equally important in determining the quality. Previously well received methods for this task are the use of deep neural networks, fuzzy inference systems and a combination of both these methods (Adaptive Neuro Fuzzy Inference System (ANFIS)). Although promising, these methods have presented notable roadblocks which, if addressed, can help make significant progress in this field of automated industrial quality check process. This work endeavors to infuse domain knowledge into the working of a regular ANFIS (DKI-ANFIS) and the modeling of a customized membership function to improve the computational capacity.

As an essential contribution, this approach allows flexible modification of a DKI-ANFIS which can be a crucial means of improving automated quality inspection and also the saving of precious industrial resources as well as time as a consequence.

ABSTRACT. Abstract—Plant diseases are the primary cause of plant destruction and lead to overall yield reduction. Data collection is one of the significant challenges in agriculture due to varying weather and lighting conditions, and it is a complex task in practicality for crops like Coffee. The existing applications mainly depend on Vegetation Indices (VIs), structural, textural, and color features are more time-consuming and involve domain expertise for accurate data labeling for Coffee plant diseases. An adequate data enhancing model can overcome this issue; hence, understanding Deep Learning (DL) algorithms, specifically Generative Adversarial Networks(GAN), can facilitate data analysis and thus enhance research in agriculture. In this work, we present an effective data augmentation approach called CoffeeGAN. This model works based on the segmentation of images using the OpenCV Color mask algorithm. The algorithm results in the segmentation of diseased spots of the leaves by giving more attention to the Region of Interest(ROI). Then the segmented images with ROI are utilized for the data augmentation process. This work employed an effective Wiener filter for data pre-processing to produce quality images for the data augmentation process. Using a transfer learning approach, we compared our model CoffeeGAN with the base CylceGAN model by training augmented data from both models in ResNet-152. The empirical results proved that CoffeeGAN performed better than CycleGAN with an adequate accuracy of 93.4% compared with 92.0% in CycleGAN

ABSTRACT. Optical Mark Recognition (OMR) Sheet, also called as Bubble sheet is a special type of form used to answer graded multiple choice question examinations, where students have to mark or darken the bubbles to answer the questions. OMR sheets are used by various school, college, university and competitive examinations. Larger institutions like universities have specialized machines to optically detect the marked answers and grade the student. However, these scanners are significantly heavy on the pocket and cannot be afforded by small tuitions or individual teachers. With the help of mobile phones, it is easily possible to scan multiple sheets within seconds. The user is expected to upload a single file of all responses and view the results in a tabular format. Using various image processing techniques, the algorithm evaluates the responses of the student and displays the grade and percentage on the input image along with the correctly, wrongly marked and actual answers with green, red and yellow colors respectively. The user also has access to an interactive dashboard which presents various analytics regarding the students’ performance. A unique feature in this implementation, is that users can choose the number of questions, number of choices, marks per question etc. after which he/she would be able to download a customized OMR response sheet. We experimented on 300 sample OMR sheets with 12800 questions and obtained an average accuracy of 99.12%. This paper aims to present a cost-effective solution to accurately scan OMR sheets without the need of scanners.

ABSTRACT. The promise of data science for social good has not yet percolated to public health, where the need is most, but lacks priority. The lack of data use policy or culture in Indian health information systems could be one of the reasons for this. Learning from global experiences on how routine health data has been used might benefit us as a newcomer in the field of digital health. The current study aims to demonstrate the potential of data science in transforming publicly available routine health data from India into evidence for public health decision-making. Four case studies were conducted using the expanded data sources to integrate data and link various sources of information. Implementing these data science projects required developing robust algorithms using reproducible research principles to maximize efficiency. They also led to new and incremental challenges that needed to be addressed in novel ways. The paper successfully demonstrates that data science has immense potential for applications in public health. Additionally, the data science approach to public health can ensure transparency and efficiency while also addressing systemic and social issues such as data quality and health equity.

ABSTRACT. The COVID-19 pandemic has revealed the flaws in our health system and provided new opportunities to build resilience. Data science is one of the six scientific gaps that emerged in the global experience with the COVID-19 pandemic. Massive Open Online Courses (MOOCs) emerged as a popular way to address capacity-building gaps around issues of public health priority, such as pandemic and outbreak management and public health data science. The objective of the study was to perform a scoping review with thematic analysis and identify the key themes using qualitative evidence synthesis. A total of 458 unique records were found, of which 69 were relevant to the pandemic and outbreak management context. Thematic analysis identified three cross-cutting themes (i) personal and professional competencies, (ii) institutional and organizational response and (iii) community participation—the growing role of MOOCs in building competencies that are empowering and enabling. The growing need for domain expertise in public health data science could be addressed through such courses. Such competencies are central to strengthening the health system and effective public health response. MOOCs present a promising opportunity to deliver public health data science education, integrating domain expertise, public health ethics and data science methods.

ABSTRACT. Globally the most rampant arboviral disease yearly is Dengue fever with approximately 96 million cases and with absence of global approved vaccine. Scientist are in continuous search to devise an epidemiological system which could timely predict, diagnose and treat Dengue infected person to lessen global yearly mortality rate associated with the disease. In this paper a quick review is done on studies done till December 2020 focusing forecasting Dengue epidemiology covering various aspects like diagnostics, prognosis, and surveillance prediction to emphasize overall scheme of current research progress and obstacles faced in it. This paper highlights various predictive approaches along with open issues that require further investigation like mosquito breeding sensory reporting, human mobility, diagnosis of disease prognosis to risky state to ensure precise prediction model of Dengue infected patient and after care. This paper has a good blend of papers encompassing various research directions that are both traditional and innovative in Dengue epidemiology and are further sub grouped according to their proposed approach and objectives to enhance deeper understanding.

ABSTRACT. We report the effect of comorbidity on the survival rates of COVID-19 using quantum machine learning. Recent understanding of the novel Coronavirus aims to verify the target organ of the virus, which could lead shortly to significant advances in the diagnosis and treatment of infected patients. An overview of the impact of the SARS-CoV-2 virus based on many different parameters such as age, type of comorbidity and gender have been studied. The data calculations are done manually by referring to parent articles and using machine learning and quantum machine learning algorithm. It is helpful to verify the target age group, gender at risk of infection, and survival rates of the person. The data used is classical data and quantum algorithms were run on it. We found out that the accuracy has increased to the classical machine learning state vector machine results. We found that pulmonary diseases are the most harmful type of comorbidity when an individual gets infected with COVID-19.

ABSTRACT. The metaverse has opened up the next door in digital evolution. It has the ability to expand the domain of online services by providing a life-like online experience. The digital experience is powered by blockchain technology, tokenomics, and decentralization. The Non-Fungible Token(NFT) is a unique token on the blockchain that is traded, bought, and sold on various NFT marketplaces. NFT marketplaces can also be curated in the metaverse, providing a more interactive experience. Key concepts and principles regarding metaverse, blockchain, decentralized applications, and user experience have been studied and a recipe to understand and create a metaverse NFT marketplace has been presented in this paper.

ABSTRACT. The rapid growth of IoT Technology offers huge opportunities and also brings many new challenges related to the security of IoT networks. Device Authentica-tion is considered to be one of the main challenges. IoT devices are not able to pro-tect themselves due to their limited processing and storage capabilities due to which they are unable to secure and defend themselves and can be easily hacked or compromised. Using password or predefined keys have drawbacks that limit their uses for different IoT applications. Blockchain technology has the capability to address these issues such as providing secure management, authentication, and access to IoT devices and their data, in a decentralized manner with high trust, in-tegrity, and transparency. A solution discussed in this paper is to make use of fog computing and blockchain technology to authenticate IoT devices using smart to-kens generated by an upgradable smart contract.

ABSTRACT. The quantum circuit simulator on HPC system plays a key role in algorithm implementation, algorithm design, quantum education and associated other research areas. In this work, we have implemented cuQuantum for accelerating quantum computing workflows on 64-bit Floating Point and TensorFloat-32 based accelerated system. The commonly used quantum algorithms like Shor’s, Quantum Fourier Transformation (QFT), and Sycamore circuit are implemented on HPC-AI system. These algorithms are further accelerated using cuQuantum. The observed performance for GPU enabled circuits increases linearly on 2, 4 and 8 A100 GPUs for the given qubit size. GPU enabled performance obtained on PARAM Siddhi AI system is compared with those observed from CPU only run as well as observed from previous generation volta architecture (V100) GPUs. For Shor, QFT and Sycamore circuit, the relative speed up, between CPU only and eight A100 GPU enabled run is observed as ~143x, ~115x, and 104x for 30, 32, and 32 qubits respectively. Similarly, the relative speed up, between CPU only and 4 V100 GPU enabled run is observed as ~43x, ~29x, and 24x for 30, 32, and 32 qubits respectively. In view of the better compute capability and memory, the relative performance between four A100 and four V100 GPUs varies from 1.5x to 2.2x for all the three algorithms.