ABSTRACT. Welcome to ICDMAI 2022. Like its 5th edition, due to the continued threat of the COVID-19 pandemic, the 6th edition of this popular International Conference on Data Management, Analytics and Innovation is again being organized virtually, keeping into mind the safety of the organizers and participants. However, our commitment to provide a global platform for highly enriching deliberations on this futuristic technology area remains as before. I recently got an opportunity to review an interesting book titled “Open Data for Sustainable Development” in which the authors propose the use of DSAI (Data Science & Artificial Intelligence) approach to achieve the SDGs (Sustainable Development Goals) identified by the United Nations to ensure peace and prosperity for all people on this planet. According to them, properly acquired, pre-processed and visualized data in various application domains shall provide useful insights to the decision makers to deploy the available limited resources (funds, manpower, infrastructure, etc.) in smarter ways to benefit larger sections of the society. The authors try to justify the usefulness of this approach through a detail analysis of three application domains – healthcare, agriculture and environment. Such books and research monographs re-confirm the need of the hour to accelerate data science research and to produce good data scientists. This conference aims at contributing to these objectives by bringing together researchers, scientists, academicians and students working in the area of data science and associated disciplines. Right from its first edition in 2016, ICDMAI has attracted eloquent keynote speakers, paper presenters, and participants from various countries, making it a truly international event. With our commitment to focus our research for solving socially relevant problems, I am sure that this year’s ICDMAI will have several interesting deliberations on data science techniques for handling various types of challenges faced by different sections of the society due to the pandemic. I thank the Society for Data Science (S4DS) for taking the initiative to organize this important annual technical event. I wish this year’s version of the conference to be equally enriching and successful as its earlier versions, in-spite-of being conducted virtually.

ABSTRACT. Quantum Technology is going to be a major disruptive innovation for the whole human kind in days to come. This will revolutionise the entire technological chain and affect at component level, thus changing the complete face of technology in years to come. The various components of quantum technology and its effect are discussed in this presentation. The status of development in the entire world and in India are also analyzed.  The inescapable necessity to focus and achieve quantum supremacy is also brought out. Quantum technology consist of communication, computing, sensors, materials etc., thus penetrating every walk of life of the human being. It directly affects the financial markets, security, defence, health and education. Therefore, a concentrated massive effort is to be put up to acquire and develop these technologies at the earliest in the country.

ABSTRACT. Ontologies are the backbone of the knowledge management system and play very important role in semantic web. Ontologies help to reduce the bottleneck of knowledge acquisition. Ontology learning process involves identification of concepts and relationship between these concepts. Many approaches based on machine learning and deep learning are used to learn ontology. Automated Ontology based on machine learning and deep learning efficiently identifies concepts and relationship between them. But automatically learned ontologies may have duplicate and similar relationships. Identification of such duplicate or identical relationships are very important to fine tune ontologies. In this paper we propose a method to Recognizing similar relationship to fine tune Ontologies.

ABSTRACT. A various techniques have been presented in this paper are based upon the structure or geometrical shape of an object and by extracting these features, we can detect an object and recognize the same. In this work, we are firstly detecting or counting a number of objects available in an image, then each object is cropped and resized and boundary values of an object is detected, which helps in extracting the relevant features of an object. A number of features which are extracted in this work are contiguous horizontal and vertical peak extent feature, non-connected and connected contour segment features, vertical and horizontal balanced division point, and chord features etc. All of these features further help in finding the shape of an object and it will further help in detection and recognition of an object. In this work, we use linear-SVM and k-NN classifiers for classifying of an object. In this work, we have taken total 1020 images from mpeg dataset, these images includes both i.e. training and testing. The dataset includes total 51 classes and each class contains 20 images. In this, we achieve the accuracy of 91.0% and 90.0% by using Linear-SVM classifier for object recognition using proposed vertical and horizontal peak extent feature extraction technique.

ABSTRACT. Artificial Intelligence is playing pivotal role in automation of processes that were considered hard problems previously, but trustworthiness of these systems is still under question as many of these systems fail to meet expectations. Trustworthiness of artificial intelligence based systems depend on many factors. This paper analyzes human trust lifecycle and proposes an end to end agile and automated machine learning framework for automation of development, deployment, monitoring, and enhancements of AI/ML processes. Further this paper presents results of initial deployments of proposed framework and compares them with bench-mark results.

ABSTRACT. Digital technologies are now becoming part of all the sectors be it banking, automobile, infrastructure, and more. These technologies are empowered by “Data”. This is raising the need for the digitization of documents to fulfill the need for data for driving the digital transformation throughout sectors. Digitization requires the extraction of a huge amount of data from paper-based documents. Automating data extraction from paper-based documents can help in dealing with large volumes of data at a lower cost with lesser efforts. A solution is proposed which uses open-source components to automate the process of data extraction from scanned documents with minimal user input. The solution is capable of generating the structured output reflecting the document layout with the data in a document. The solution is capable of extracting data from tables and stamps present in documents in a well-structured format. The solution is driven by a configuration file, which can help in fine-tuning different processes to improve extracted data. The solution generates an XML for the scanned document which can be used further for storing and processing the data present in paper-based documents by different digital processes.

ABSTRACT. With the evolution of the social media-based application, users share their views, opinions and emotions liberally in language English and any native language like Bengali. The posts or comments collected from the social networking sites are needed to be analysed for identifying the positive or negative public sentiments regarding the products or services of any organization in order to improve their services. Over the last few years, research on sentiment analysis using Machine Learning (ML) tools is very popular. However, the sentiment analysis of Bengali data corpus using a Machine Learning tool is limited. It has been observed that the ML models for sentiment analysis are suffering from misclassification and researchers are still trying to come up with better solutions. In this paper, a Machine Learning approach is proposed for sentiment analysis of the Bengali data set collected from Facebook. Since the posts or comments that exhibit negative polarity are more actionable than the positive ones, the prime aim of this paper is to identify the negative sentiments without being misclassified and introduces a new technique to achieve almost 100% specificity by tuning the class weight of the ML classification model. We have also tried to improve the accuracy of ML classifiers during the sentiment analysis by combining the Part-of-Speech (POS) tagging with Term Frequency - Inverse Document Frequency (TF-IDF) vectorization. The overall accuracy of each ML classifier for combining the POS tagging with TF-IDF shows an increment of 2.74-4.17% from the TF-IDF vectorization without combining with POS tagging.

ABSTRACT. Modern mobile devices have become important part of day to day life as they offer various tools and services which are very useful. Mobile devices handle lot of sensitive information hence malware writers try to exploit vulnerabilities to gain access. Android is one of the most popular OS for mobile device. The open source nature of Android Operating System has attracted users because of open source nature and large number of users it has become lucrative for malware writers to target android devices. With time there is continuous evolution in variety, volume and sophistication of malwares. Researchers are continuously working to develop better malware detection methods which can deal with sophisticated malwares. This paper presents comprehensive overview of latest malware detection methods. The rapid increase of mobile devices and android operating system has led to security issues with android operating system. Malware writers attracted with the increasing use of smart phones and android OS. Hence there is a need to study existing malware detection system for android OS. This paper gives the detailed review on android malware detection system.

ABSTRACT. Natural Language processing applications such as sentiment analysis, spam detection and stance detection extracts author’s emotion, feelings, and categorization such as favor or denial from the piece of text sentences or corpus. Various researchers keep on working in these areas. However, in this research work, the relation between text target entities (say: climate, population) along with the author’s stance and its country’s economic growth level are used to derive a hypothesis. This hypothesis is being proved, discussed and concluded as per the result obtained. All the missing country information for each account is filled up with a technique that uses meta information of tweets. Subset of the data is annotated with predefined seeding features, rules and then applies the best performed supervised machine learning model trained using those annotated set to predict remaining unlabeled tweets. Tweets are labeled as believer and denier for each countries and a hypothesis is being assumed based on rich and poor countries. The statistical result shows that there exists a positive correlation between GDP growth rate and the number of deniers and believers in a country. These techniques, experimental findings and statistical analysis are presented in this paper.

ABSTRACT. Severe thunderstorm prediction by analysis of cloud imageries is becoming an interesting area in research field. It causes destruction to daily life. Therefore correct prediction of severe thunderstorm has important significance. Here in this study Support Vector Machine (SVM) has been applied on cloud imageries for the prediction purpose. Colour is one of the most important features of image. Here colour has been considered as only feature for the classification purpose. Two sets of cloud imageries have been considered here, one for ‘squall days’ and another for ‘no squall days’. The imageries for squall days have been indicated by ‘1’ and ‘no squall days’ by ‘0’. The linear Support Vector Classifier (SVC) has been applied here for classification. Principal Component Analysis (PCA) has been applied here for feature reduction purpose, which yields better result. This prediction has a lead time of 5-6 hours which is enough to save society from devastation created by severe thunderstorm.

ABSTRACT. Regulators, Patients and Industry share the space and work together, with this they share their ‘real-time’ data. Which helps in faster analysis, investigation and generate greater insights and helps in better prediction. Regulators and industry will be able to identify the risks faster and then act on it. If we follow a data driven approach the relations with regulatory becomes advantageous. With the help of advance analytics, we detect new data patterns and trends which help in effective approach and help further in compliance mapping and change assessment. The ultimate goal is for industry is to lead and ‘self-regulate’ and improve transparency. In this paper, we discuss a new paradigm, where in life-sciences companies and regulators work together in a collaborative way to build stable rules and regulations instead of applying restrictive and punitive approach to regulations.

ABSTRACT. Traditionally, the data entry process at Business Process Services (BPS) is a human driven activity. Associates manually type the fields which need to be captured in target IT Systems. BPS have very strict Service Level Agreements for digitizing data from documents like Invoices, Mortgage related documents and Financial Documents. BPS providers most often use machine automation techniques like OCR/ICR on scanned handwritten documents to keep up with the SLA’s. This has had limited success because the documents contain a very large number of variations in terms of format and structure. Often a human in the loop is needed to locate the required information based on certain context. The proposed “Voice Driven Data Entry” can create a significant impact in terms of SLA’s compared to the traditional structured way of automation involving integration and imaging.

ABSTRACT. Today’s online education is very essential tool for learning. The advancement of information technology has allowed the creation of virtual courses which have created new challenges for teachers in motivating students. The online classes are becoming monotonous and hence not creating more interest among learners since it has no gamification environment. There are no proper application of gamification elements. Most of the online learning platforms are either text or video based or live communication between teachers and learners. The objective is to define gamified element in the current online education platforms to motivate students. In this paper gamified modeling environment is proposed and overall implementation mechanism of gamified learning architecture is proposed. In this paper we have focused how neural network can be used in the gamified learning environment to improve the overall performance of system.

ABSTRACT. The deaf and dumb community uses sign language as a means of communication. Sign language is a language of signs and facial expressions and not of spoken words. It is a visual mode of communication. The position of hands, the movement of fingers and the expressions on the face play a vital role in sign language. Sign languages have a very limited set of words. The grammar is difficult to understand. On the contrary, spoken languages across the globe have a rich vocabulary. It is difficult for signers to understand a spoken language. There is a need to develop a system that establishes a link between spoken and sign languages. Translating from spoken languages to sign languages or vice versa is a challenging task. This paper presents the state-of-the work that has been done in the field of translating English (a spoken language) to Indian Sign Language.

ABSTRACT. Location-based services are an important aspect of living, as these not only provide time benefits but also save a lot of energy. With the increase in complex building structures, people of different ages may find it difficult to navigate within such structures. When we say different age groups, it includes the age groups from little ones who don’t understand directions to the older ones who are in an urge to find places early due to getting drained out easily. Indoor navigation is not only an asset to sighted people but for the impaired ones too. Indoor navigation when integrated with voice assistants allows the impaired people to navigate hassle-free. For this, an Indoor navigation system is required which localizes the user, takes the account of their current location, and then corresponding to the destination point, guides the user through the path. The literature survey takes into account the different localization approaches considered by different researchers in indoor localization and the necessary positioning algorithms to reach the destination for the paths. This paper also reviews the plus points and limitations of the research work that has been done in Localization and Pathfinding algorithms.

ABSTRACT. In this paper we present a comparative study of two approaches for the detection of foreign objects in an industrial assembly line setting and proposes a complete solution from the findings. The methodology is vision based and can be used for processing 3D objects conveyed at a constant velocity. Out of the two methods, the CNN based approach is recommended to the company sponsoring this research. The design of the system is accomplished using a fixed camera, a display unit, a conveyor belt and further a raspberry pi or equivalent hardware to run the solution. The novelty of this solution is the possible full automation of the assembly line with low latency and high performance and with a small training dataset.

ABSTRACT. Increasingly heterogeneous information on social discussion forums, fake news arises to create rumors to change the reliability of information resources due to contextual terms miss classification. The problem is that features definitions and their relational contexts are not properly extracted to analyze the contextual terms. The Original sense of contextual terms is affected by lexical terms, interrogative terms, extortion, sematic features, and sarcastic terms. To concentrate originality, subjective terms are extracted as positive credibility correlation scores between positive and negative content ratios. To propose a Fake content analysis based on Inverse Contexture Abstractive Term Frequency Model using Surf Scale Diffusive Neural Network for public forum social content. By analyzing the positive correlation of the sentence using Reliable Subjective Influence Score (RSIS), this selects the positive terms depends on mutual content dependencies between the originality terms. To analyze the Inverse Contexture Abstractive Term Frequency Model (ICATFM) for feature selection to select the credibility score in the sentence to relate to subjective real or non-real terms. The selected features are trained into a deep neural classifier optimized with Surf Scale Diffusive Neural Network (S2DNN). This implementation proves the best performance by identifying the fake detection to produce higher precision and recall rate to increase the classification accuracy compared to other methods.

ABSTRACT. Every human being has the fundamental right to equal opportunity. The world's deaf society, like every other person, deserves access to all forms of information. Gestures have long served as a means of communication for deaf communities around the world. There should be a way for hearing and deaf people to communicate directly in order for this to happen. A human translator can help in this situation but this will violate the privacy of the conversation. In this 21st century, it has become imperative to automate this communication so that deaf people are not reliant on human translators. This paper deals with the Sign Language generation systems based on different machine translation techniques.

ABSTRACT. Enterprises are rethinking their cognitive transformation journey in post-pandemic new world order. Data driven platforms are increasingly realizing the benefits of AI and rebuilding businesses around experience and reimagining growth. The session will focus on shifting AI adoption from experimenting to strategic envisioning and growing importance of data capabilities to deliver trust, transparency and decisions. The session will also cover how Managing data in a multicloud world needs an architecture driven pattern based data strategy to accelerate transformation. AI at scale built on digital data foundation is strategic imperative today to build a resilient Enterprise that is innovative at scale, responsive to change and intelligent in operation.

ABSTRACT. Breast cancer is one of the most common forms of cancer that is diagnosed in most women and some rare cases even men. In recent years, breast cancer survival rates have increased significantly, due to factors such as earlier detection. Treatment for breast cancer largely depends on identifying the type of mass of tissue formed, which is known as a tumor. If normal cells grow in an uncontrollable manner the tumor is called benign (non-cancerous). But, if the cells’ growth is out of control and their behavior is abnormal, then the tumor is called malignant (cancerous). During the invasive, (i.e. curable) stage of cancer, only 10-15% part of the breast contains cancerous cells. Therefore, it is difficult to diagnose it using mammography. However, the development of machine learning techniques has allowed early detection of breast cancer in clinical trials. This paper presents a comparative study of different machine learning techniques for detecting breast cancer. It also presents an auto-encoder model that performs breast cancer detection in an unsupervised manner. It attempts to identify a compact representation of features that are strongly related to breast cancer. The techniques are tested on the Breast Cancer Wisconsin (Diagnostic) Dataset, which is publicly available on Kaggle. The auto-encoder beats its competitors with a precision and recall of 98.4%.

ABSTRACT. The rapid growth of scientific literature in the fields of computer engineering (CE) and computer science (CS) presents difficulties to researchers who are interested in exploring publication records based on standard scientific categories. This urges the need for automatic classification of text documents into scientific categories using content and contextual information. Document classification is a significant application of supervised learning which requires a labeled data set for training the classifier. However, research publication records available on Google Scholar and dblp services are not labeled. First, manual annotation of a large body of scientific research work based on standard scientific terminology requires domain expertise and is extremely time-consuming. Second, hierarchical labeling of records facilitates a more effective and context-aware retrieval of documents. In this paper, we propose an ontology-driven classification technique based on zero-shot learning in conjunction with agglomerative clustering to automatically label a scientific literature data set related to CE and CS.

We study and compare the effectiveness of multiple text classifiers such as logistic regression, support vector machines (SVM), gradient boosting with Word2vec and bag of words (BOW) embedding, recurrent neural networks (RNN) with GloVe embedding, and feed-forward neural networks with BOW embedding. Our study shows that RNN with GloVe embedding outperforms other models with an above 0.85 F1 score on all granularity levels.

Our proposed technique will help junior and experienced researchers in identifying new emerging technologies and domains for their research purposes.

ABSTRACT. Due to the prevalence of tuberculosis, it has become a source of concern in the world. From day to day, new cases are still discovered. To help governments and health policymakers make appropriate decisions and impose restrictions to reduce the prevalence of tuberculosis, efficient forecasting methods are required to be applied. Machine learning and deep learning techniques are commonly used due to their abilities in producing accurate results. In this paper, seven different models are applied to accomplish this objective, including SARIMAX, LSTM, CNN-LSTM Hybrid, MLP network, SVR, XGboost, and RF Regression models. These models are applied to forecast pulmonary positive, negative, and TB incidence cases. The models with the lowest errors are then chosen and used to forecast the number of pulmonary positive, negative, and TB incidence cases. The results of the experiments showed that the CNN-LSTM Hybrid and MLP networks achieving the lowest forecasting errors compared to the other models and being chosen for forecasting pulmonary negative, positive, and TB incidence cases from 2020 to 2029. The forecasting results revealed that there would be 117.861557 new pulmonary negative incidences, 153.029385 new pulmonary positive incidences, and 414.4704 new tuberculosis incidence cases.

ABSTRACT. Agriculture is the chief support for the nation, measures have to be taken to solve the problems faced by the farmers which mainly includes the increasing jeopardy caused by animals in farm fields. They mainly damage the crops during sowing, seedling and ripening stages and cause economic loss to farmers and reduce the country's GDP. Ultrasonic Repellers are flourishing and its functionality can be increased by installing it on drones. Furthermore, Machine Learning and Artificial Intelligence can be incorporated for detection of animals destroying the farm field and produce the upper limit frequency(ultrasonic frequency) of the detected animal. Raspberry Pi and a rebarbative system are used which is mounted on the drone. The Raspberry Pi has a camera attached, to capture the live video/images of the animals destroying the farm field and the Raspberry Pi having the detection code programmed in it, detects the animals destroying the farm field. The rebarbative system is interfaced with the Raspberry Pi. Upon detection of animals invading farm fields the rebarbative circuit is triggered to produce the ultrasonic frequency of the detected animal. As the gadget is not static, that is the Raspberry Pi along with the rebarbative circuit for producing ultrasonic frequency is placed on the drone, the repelling capacity is not bounded to a specific area but the entire field. Invading animals do not get adjusted to the ultrasonic frequency and thereby the repelling capacity is efficient. Generated ultrasonic frequency will not get attenuated and is directed to targeted animals as directional speakers are used. Since wild animals and birds species are deleterious this device is more convenient to repel them and outshines the existing system as efficient and advancing technologies like machine learning and electronics technologies are inculcated.

ABSTRACT. Abstract

Extrusion film casting (EFC) is a very important industrial process employed for producing large amounts of thin films for multiple uses. It is important to identify input process parameters for maintaining stable and continuous trouble-free operation. A priori modelling and solution of governing phenomenological equations to achieve this involves large computational time and costs. As an alternative in this work we have employed Machine learning to build data driven EFC classification models to predict stability of any given set of operating conditions. We have used accurate input data from available literature for certain combinations of parameters along with their respective output classes to build robust classification models. We used Support Vector Machines (SVM) and Random Forest classifiers for this purpose. We created four different data sets with different process parameter combinations and number of relaxation modes. We also used the Synthetic Minority Oversampling Technique (SMOTE) to handle data imbalance. Our simulation results indicate that prediction of stability/instability classes for different process parameters can be achieved with high degree of confidence with robust machine learning models

ABSTRACT. In the previous decade, the measure of CO2 outflows has seen a critical increment, this is coming about to be a significant justification for environmental change. A significant supporter of these emanations is the transport sector. This sector alone is assessed to represent about 16.2% of the absolute worldwide CO2 outflows. The focal point of this paper is to examine and audit diverse AI and ML techniques that can be utilized to decrease these CO2 outflows caused because of the transport sector. The paper talks about an ML approach to deal with anticipated CO2 discharges brought about by vehicles. Utilizing the aftereffects of this model, the neighbourhood overseeing bodies can get ready for a superior public vehicle foundation if necessary. Region savvy examination can assist the overseeing bodies with managing the progression of the public vehicles in various locales of the city which will, thus, bring about a decrease of CO2 discharges. The model has shown promising results by achieving an RMSLE of 0.71 and accuracy of about 99.8% using Machine Learning models like Lasso Regression, Multiple Linear Regression, XGBoost, Support Vector Regressor (SVR), Random Forest, and Ridge Regression.

ABSTRACT. With size of datasets varying non-uniformly in sample size and feature length, to optimize the feature set usually different methods such as filter, wrapper methods are used. However, with different machine learning techniques though either feature reduction is used, or feature extraction is used, both have its own merits and de-merits. The proposed work proposes a hybrid model that tries to combine the feature extraction and feature reduction techniques thereby using both linear and non-linear techniques to take the best parts of both methods. After the initial ensemble is created still the feature set is further optimized by using the concept of entropy and information gain. Using mutual information, on further analysis the best non-redundant feature sets are selected after considering a specific threshold and using this as a testing tool the datasets are again analyzed to check the working accuracy. The model performance is found to be effective even using reduced feature sub-set. Also, it has been found apart from excelling in classification accuracy, the model has been successful in maintaining the range of the metric irrespective of the input size.

ABSTRACT. The energy sector of any country is the powerhouse for its economy. All the other sectors and domains of any country rely heavily on the energy industry. Along with strengthening and boosting the economies worldwide, this sector is one of the major contributors to climate change. The amount of emissions generated by the energy sector have accelerated this disastrous change. India is advancing as a developing nation with incredible speeds and its energy sector is considered to be one of the most assorted sectors with the varied number of resources and reserves it holds. The objective of this research is to study the non-renewable sources of India’s energy sector and predict future temperature forecasts by analyzing the emissions of these non-renewable sources on the temperature. A combination of ARIMA and ARIMAX modelling have been employed to first complete the dataset and then predict future values. The results show an alarming rate of increase in the future and consequently the paper discusses the mitigation methods and reforms at governmental levels. The geographical potential of Indian states for harnessing energy from cleaner sources has also been discussed.

ABSTRACT. The statistics presented by the World Health Organization attributes depression to be a primary cause of concern globally, leading to suicide in majority of the cases if left undetected. Nowadays, Social media is a great point for its users to express their opinions through text, emoticons, photos or videos thus reflecting their sentiments and moods. This has created an opportunity to study social network for understanding the mental state of the users. Studies show that depression generally has an impact on the writing style and corresponding language use. In addition, user persona on social media can also provide us a lot of information about the mental state of the user. The primary aim of our research is to study user’s persona and posts on Twitter and identify the attributes that may indicate depressive symptoms of online users. We used machine learning approaches and natural language processing techniques for training our data and evaluating the efficiency of our proposed method. We proposed a two-level depression detection in which the social media features, personality trait and sentiment analysis of user’ biography provide us an opportunity to identify suspected depressed users. We combined these attributes with other Linguistic features (N-Gram+TF-IDF) and LDA and achieved an accuracy of 89% using Support Vector Machine classifier. According to our research, proper feature selection and their combinations help in achieving better improvement in performance.

ABSTRACT. The life of every human being in this world is hampered by Covid 19 virus. Everyone is familiar with do’s and don'ts in social environment. The main objective to keep away from Covid 19 is to wear the mask and maintain social distancing. Our AI based algorithms highlighted in this research paper are checking the mask unavailability and are measuring social distancing. We estimate social distancing parameters using estimation of actual distance based on perspective distance without using camera calibration. OpenCV libraries are used to detect people in 2D plane while deep learning-based approach is used for mask unavailability check. It uses combination of 3 methods such as mapping of distances, diagonal ratio comparison, overlapping of bounding box drawn around detected human being for social distancing violation. Every method has pros and cons. AI based methodology has been used to assign weightage to above various algorithms depending on the scenarios. Distance between human beings can be calculated and created real time alert in case of violation. This algorithm has tested on various video streams to check the performance and feasibility of it.

ABSTRACT. A major problems writers and poets face is the writer’s block. It is a condition in which an author loses the ability to produce new work or experiences a creative slowdown. The problem is more difficult in the context of poetry than prose, as in the latter case authors need not be very concise while expressing their ideas, also the aspects such as rhyme, poetic meters are not relevant for prose. One of the most effective ways to overcome this writing block for poets can be, to have a prompt system, which would help their imagination and open their minds for new ideas. A prompt system can possibly generate one liner, two liner or full ghazals. The purpose of this work is to give an ode to the Urdu/Hindi poets, and helping them start their next line of poetry, a couplet or a complete ghazal considering various factors like rhymes, refrain, and meters. The result will help aspiring poets to get new ideas and help them overcome writer’s block by auto-generating pieces of poetry using Deep Learning techniques. A concern with creative works like this, especially in the literary context, is to ensure that the output is not plagiarized. This work also addresses the concern and makes sure that the resulting odes are not exact match with input data using parameters like temperature and manual plagiarism check against input corpus. To the best of our knowledge, although the automatic text generation problem has been studied quite extensively in the literature, the specific problem of Urdu/Hindi poetry generation has not been explored much. Apart from developing system to auto-generate Urdu/Hindi poetry, another key contribution of our work is to create a cleaned and preprocessed corpus of Urdu/Hindi poetry (derived from authentic resources) and making it freely available for researchers.

ABSTRACT. Development of an open domain question answering system that can automatically find and give the answer for any question in the given context is always an open challenge for the researcher. This paper discusses about a web based application which accepts a media file from the user and generates a span of context for the same. Then the users can pose any open domain question which gets mapped with the context and an appropriate answer is provided as output. SQuAD 2.0 dataset is used for training the model. The pre-trained models of ALBERT with different parameter sizes are implemented to perform the media file content context matching with the input question and the most optimized one is selected as an output. To measure the performance of the system, Exact Matching (EM) score and F1 score are calculated.

ABSTRACT. Surround View Monitoring System is an emerging ADAS (Advanced Driver Assistance System) feature that assists the driver while parking the vehicle safely and allows driver to monitor the blind spot areas around the vehicle by allowing driver to see the top view of the 360-degree surrounding of the vehicle. The main aim of this project is to help the driver in the driving process which increases car safety and road safety. The objective of this project is to provide 360 degree surrounding view of the vehicle, to obtain information about the possible obstacles present in the blind spot area of the vehicle and to assist the driver while parking the vehicle. In this system, there are four fisheye cameras mounted around a vehicle to cover the whole surrounding area. After correcting the distortion of four fisheye images and registering all images on a planar surface, a flexible stitching method will be developed to smoothen the seam of adjacent images away to generate a high-quality result. In the post-process step, a unique brightness balance algorithm will be proposed to compensate the exposure difference as the images are not captured with the same exposure condition.

ABSTRACT. In this work Sky Computing approach has been proposed for “Health Analytics”. Most of Health data of ailments are stored by hospitals in private clouds. In several cases when a query on an ailment is raised by a doctor it requires extensive querying of data on the internet. This may not be always reliable as obtained from the internet. Cloud Computing provides a solution but in several cases datasets required to address a query may not be always available on the same cloud. In such cases it is necessary to obtain datasets from multiple clouds. The Sky computing approach reduces query processing time by 30%, compared to the conventional manual cloud approach.

ABSTRACT. Recent studies confirm that several journals exchange intentionally biased citations to inflate their Journal Impact Factors (JIF) mutually. It includes excessive self-citations, stacking, cartels, cabals, and rings. Microscopically, the key entities are authors, editors, and publishers. Identifying coordinated citation manipulation is complex because multiple dynamics are involved. Also, such behavior varies largely across disciplines. Hence, there is still a lack of automated algorithms to detect them readily. Nevertheless, the real problem arises when authentic journals with identical citation patterns (natural biases) are associated similar to abnormal patterns. Thus, our prime objective in this paper is to understand all reasons behind naturally occurring citation biases. This paper proposes a novel generalized methodology to detect such journals with irregular JIF inflations using community-based analysis. First, we model large-scale time-series citation data of 1,606 journals in a network structure. Next, we detect communities from the resultant temporal network using a multi-layered modularity maximization algorithm. Broadly, we obtain four communities- Self-Citation (SC), Pairwise Mutual-Citation (MCP), Group Mutual-Citation (MCG), and Uni-directed Citation (UC). Macroscopically, we define the underlying community dynamics using network parameters. The promiscuity of the SC class is the highest, 0.90, and cohesion strength of MCG class is highest at 0.71. Microscopically, we present a case-by-case analysis from real-world data. An abrupt change in a donor's publication rate and sudden inflation in the recipient journal's JIF is a characteristic feature. Other features leading to natural biases include narrow domain specialization, publisher's impact, citations from newly published or review journals, overlapping author sets, and author self-citations. Consequently, future studies must carefully consider all these factors before modeling any citation anomaly detection algorithm.

ABSTRACT. Multiple Wi-Fi Connectivity Combined to deliver end-to-end network solutions for the wide range of network expansion. Wi- Fi Networks makes use of additional USB Wi-Fi Network adapter to combine the existing Wi-Fi networks to improve the performance of networks . It provides solution to combine Wi-Fi networks and answering the following Network related problems like Resources sharing, Traffic and Security Management in Networks. It increases the performance and efficiency of available networks by managing the various factors like detecting the available Wi- Fi connectivity along with bandwidth for file transfer and measuring the speed and accuracy of using the network resources. Multiple Wi-Fi Networks Combined to find the optimal solution in sharing of network recourses between various network Wi-Fi Connectivity . By making use of 5G LTE AER router we can able to combine more than one Wi-Fi network and by making use of this application to enhance the performance in combine the available network.

ABSTRACT. Aircraft Maintenance is a complex task requiring highly trained engineers. Facilitating the Aircraft maintenance through providing proper tools and equipment is essential in ensuring good maintenance work. While there are a plethora of Augmented Reality applications, the best-suited applications for Aircraft Maintenance are to be selected and evaluated for their usefulness. In this paper, ARCaddy, an AR application suite consisting of a set of openly available AR applications is presented. The ARCaddy is packaged with relevant and tested AR applications. ARCaddy is evaluated for its usefulness by the feedback from the Aircraft Maintenance Engineers.

ABSTRACT. It is a well-accepted fact that today every individual have to perform so many activities simultaneously which leads to hectic lifestyle. This hectic lifestyle has resulted to an increase in mental stress among people. Meditation has been one of the most widely and effective ways in reducing stress of a person. Apart from stress reduction, meditation also helps to increase the concentration level, control emotions, reduce sleeplessness and enhances the ability of a person to face challenges. Though meditation has several benefits, self-assessing the impact of meditation on stress level has been difficult and challenging. To address this problem, in this paper, we have designed and developed a Virtual Reality Assisted Meditation Therapy (VRAMT) system. During meditation, this system measures the brainwaves and classifies them to track the state of mind as whether calm or stressed. Based on the classification result, the system provides virtual reality feedback in real-time to the user through which the user can self-assess one’s state of mind.

ABSTRACT. Many researches admonish for the burgeoning global warming above pre-industrial levels which has catastrophic consequences for human generation. There are various sectors in industry which contribute to the increase in carbon emissions. Industries such as textile, mining, steel etc emit huge amounts of carbon di-oxide ()every day. is termed to be the predominant greenhouse gas in global warming. It’s levels are increasing without much flora compensating for it. It ultimately leads to climate change. Similar to other industries, AI has also started increasing its contribution to the deterioration of the environment. In this study, we focus on elucidating about the negative aspects of Artificial Intelligence towards climate change. Calculations of carbon emissions have become a necessity in order to limit the usage of the resources and increase awareness among people regarding the implications of global warming. Thus, the paper also discusses a real time carbon emission calculator used to calculate carbon emissions of personal computers.

ABSTRACT. Covid-19: a pandemic situation in the world gives a turning point to the education system and it becomes e-education. E-learning is an emerging trend in the digital era and empowerment of this trend is necessary. Traditional education systems trying to adopt this new method of teaching and learning. But only teaching and learning are not sufficient in the education system. We have to focus on learners and the environmental impact on them. The traditional education system is unable to resolve all the issues that arise due to obstacles such as understanding ability, thinking, mood, concentration, etc. Proposed research work focusing on designing, developing, and modeling of soft computing decision-making model for solving real-life problems and learners capability in the education system. This research work uses Fuzzy Inference System (FIS) which is one of the applications in MATLAB software, for analyzing learner's results from the obtained scores and other factors related to the environment. It also predicts the learner which is helpful in e-learning.

ABSTRACT. Biophilia hypothesis, suggests humans possess an innate tendency to seek social connections with all forms of life. Connection is understood as a core human need. Same is applicable for a population in any of the organization or establishment where conversation and communication are fundamental ingredients of connectedness. Inspired by this, our work uses Organizational Network Analysis (ONA) which is an established method for studying modern insight to enable strategies for exchange of ideas and information within an Organization. Existing relationship(s) amongst employees and effect of organizational culture on the underlying social network within an organization are visualized. we further analyze the impact of employee demographics & organizational hierarchy on the network of employee connects. This paper discusses various network findings, including, overall network view, relations between employees, shortest paths between any two employees, degree of centrality, eigenvector centrality and most influential employee and statistically relates these findings to employee demography and analyzes impact of organization structure along with human behavior. Python NETWORKX package has been used for all processing. There is wide applicability of the work within a large and growing organization, global establishments. HR Function may make use of ONA for their diversity and inclusion goals and these findings can help them in aligning the same to arrive at more realistic insights

ABSTRACT. Today whenever there is a critical requirement of an expert healthcare advice or a requirement of a specialist doctor not much means are available to verify the creditability of the doctor. (modified the abstract) If we are able to maintain the doctors creditability on a system than it becomes an asset of record which will be available in a secure and tamper proof manner. On each successful handling of a case healthcare agencies may upload the data of the success and there can be a mechanism designed to update the doctor’s creditability by assigning a token. Tokens may be categorized depending on the success rate for which a parameter in medical terms may be fixed. For this Ethereum blockchain may be used. The Smart Contract would execute every time to update the tokens of the doctors registered. As more and more Health care agencies become a part of the system it would provide a standard platform to make a selection of a doctor based on the severity of the diseases.

ABSTRACT. Mass transportation connects people with education, employment, and community resources. However, navigating public transportation can be difficult for the visually impaired and physically challenged members of society. There seems to be a rapid proliferation of technology. There has been an urgent need to develop improved methodologies to assist people with disabilities in gaining access to public transportation. While there are numerous navigation systems, some of them rely on Global Positioning System technology, which is useful in outdoor environments but ineffective in indoor environments. Beacons are emerging sensors that are increasingly being used for indoor positioning in shopping malls and airports. They make use of Bluetooth Low Energy technology, which is widely supported by today's smartphones. In this paper, we propose a proof of concept for a mobile application that uses Bluetooth Low Energy beacons to assist people with special needs. This app would provide a simple voice interface for navigating within stations, buses, and trains. As part of our research, we intend to use existing beacons and deploy new beacons in some regional traffic centers. We intend to analyze ‘ridership data’ in order to extract contextual data about the commuter's current environment in order to provide commuters with a cognitive solution and assistance during travel. Our research's ultimate goal is to improve public transportation for visually impaired people.

ABSTRACT. Many recent technologies increase the generation of data and its usages. There are concerns to store, retrieved large data, processing multiple queries and services simultaneously. The cube format of the data and dimensional databases can ease the process of retrieval and modelling the data efficiently and effectively. This study suggests few efficient ways to address the concerns using the concept of a data warehouse and analytical operations. It also offers the design aspect of a Hybrid analytical system by linking different functionalities under a Layered Architecture style. The desired inputs are selected from warehouses, later consolidated to form incremental subsequent higher-level data. This style supports a Hybrid system to provide trust by linking across many data sources of the distributed warehouse systems. It enables the ELT services other than the normal ETL operations to handle large data to support the data lake. The suggestive functionalities engine is used to produce data patterns. The merit of the PDC tree is incorporated to provide some possible parallel operations. The findings are applied to a case study of data modelling to predict a potential future epidemic. Such a system generates several reports to help the users or the authority for handling such an epidemic in better efficient ways.

ABSTRACT. In this paper, symptoms and medical treatment data of 130 COVID-19 patients have been collected from a leading Hospital in Kolkata. After necessary de-identification and data wrangling, a thorough exploratory data analysis has been performed. Further, it has been investigated if the drug Remdesivir affects early discharge. A decision tree-based model was subsequently built to predict the length of stay of a patient, based on demographics and health parameters. It is observed that Remdesivir cannot be concluded to be more effective than alternative treatments. It is observed that diabetes significantly increases the length of stay of a patient. It may be noted that such a study has not been conducted earlier for COVID-19 patients in India. This study will be beneficial for the healthcare community & pharmaceutical companies as there is a lot of conflicting views and an acute dearth of information about the disease and its treatment.