ABSTRACT. In this paper, for minimally repairable items from heterogenous populations, a new multi-stage procedure with periodic replacements is proposed. Between consecutive periodic replacements, the failures are minimally repaired and the information on the numbers of minimal repairs in relevant intervals of time becomes a decision parameter in the corresponding optimization problems. For instance, for the 2-stage policy, an item is replaced at the optimally obtained time if the number of minimal repairs at this time is larger than the optimally obtained value. Otherwise, the replacement is postponed until the optimally obtained time. Theoretical results and numerical illustrations justify the proposed approach.

ABSTRACT. In manufacturing industry, data-based condition monitoring is conducted to maintain an effective operation of manufacturing process. Particularly, the degradation of production tools and equipment can significantly impact the quality of generated product and yield through process operation. To prevent system failures and reduce operational costs, condition-based maintenance (CBM) techniques based on observed degradation have been developed. To perform an effective maintenance, it is important to collect abundant degradation data to secure a wealth of information, and to detect the point of transition from a normal to an abnormal state. This study proposes an image sensor data-based comprehensive analysis for diagnosing the status of manufacturing process based on degradation modeling and condition monitoring. To effectively model inherent variability and uncertainty of image degradation, change-point spatio-temporal process (CP-STP) is developed. By estimating the parameter set via Markov chain Monte Carlo (MCMC) method, the spatial and temporal diffusion of images is described with the proper change-point. Based on the modeling result, CBM scheme is established to identify the time for repair of replacement. Through the application of proposed methods, this research contributes to the early detection of critical degradation points and the strategic planning of maintenance, thereby improving the reliability and efficiency of manufacturing process.

ABSTRACT. We consider a system in which the degradation of two components follows a Wiener process with positive drift, and a failure occurs when the sum of these weighted degradation amounts, which is called the degradation level, exceeds a threshold value given in advance. The deterioration level is periodically observed, and if preventive maintenance is necessary, we choose whether to replace one or two components. Also, in the case that one component is replaced, we decide which is more appropriate. We investigate various properties of the optimal maintenance policy by modeling the system via a Markov decision process(MDP). Then, we theoretically show some structural results on the optimal maintenance policy which are consistent with the relevant studies in this area. Also, we provide the significance that the weight is introduced through numerical analysis.

ABSTRACT. The fifth-generation mobile network (5G) is widely recognized as one of the biggest market opportunities in the coming years. Despite the commercialization of 5G services by carriers, many consumers are still hesitant to switch from 4G to 5G. To address this emerging issue, this study applies the push-pull migration model to develop a second-order reflective-formative model to investigate users to switch their subscription from 4G to 5G within a telecom carrier. A total of 172 respondents were collected and analyzed using partial least squares technology. Our findings suggest that pushing has a greater effect than pulling, indicating that users are generally satisfied with the incumbent 4G service. However, 5G becomes more attractive when carriers put more effort into promoting its benefits. The results suggest that carriers need to focus more on promoting the benefits of 5G to encourage users to switch.

ABSTRACT. Aiming for “sustainable development” based on the idea of gradual development with consideration for environmental conservation is an important issue in all fields. The field of information and communication technology is no exception, and in order to promote sustainable development, there is increasing interest in electricity consumption, especially from the perspective of energy issues in data centers. Data centers have costs of servers, infrastructure, power requirements, and networking. It is reported that amortized cost of servers roughly goes up to 45%. Therefore, reduc- ing the power costs of servers leads to a reduction in power consumption. On the other hand, today’s virtualization technology has advanced to the point where network functions can also be virtualized, so-called Virtual Network Functions (VNFs), which improve energy efficiency by automatically scaling virtual resources. In this paper, we consider hybrid systems in which the VNFs and legacy network equipment coexist. We evaluate the performance of the hybrid systems by using queueing models with setup and delayed-off policy.

ABSTRACT. Vacation queueing models are useful for those systems in which the server wants to utilize his idle time for different purposes: maintenance or computer security actions, saving energy etc. In classical models, the server is unvailable to customers during a vacation period and the server stop completely the original work and cannot come back to the regular busy period until the vacation period ends. Here we consider a model with two types of vacations (called differentiated, for example long and short) and in which the server is available to customers but at different rates than normal mode (working vacations). We obtain the joint probability distribution of the server state and the queue size. Several mean performance measures of interest are derived . Numerical illustrations are given in order to show the effect of some parameters on these performances measures.

ABSTRACT. To protect the environment of Earth and maintenance of all living things, we have to make the development of reliability theory, as information technologies have advanced rapidly. This chapter proposes three new schemes of mean deviation, redundant x units, and scalability. Mean deviation shows good examples of statistical applications to reliability engineering, redundant x units is a mathematical extension of a standard redundant system with $n$ units, and scalability is a property in the context of system design and structure to handle a growing of work. Three schemes would be interested in some researchers and be much useful for theoretical development and practical application in fields of reliability, statistics and others.

ABSTRACT. Financial service providers such as banks, trust companies, and securities firms are focused on understanding the environment in which their corporate clients operate, and on identifying changes in their needs related to the securities finance business in a timely and qualified manner. Financial service providers are exploring efficient sales support measures while implementing effective sales support by their headquarters (hereinafter referred to as sales support), rather than relying solely on corporate sales representatives to understand the changing needs of corporate clients. Sales support can be broadly classified into two categories: relatively high-cost sales support, such as accompanying corporate sales representatives on visits, and relatively low-cost sales support, such as providing information using some information communication environments. The purpose of this study is to explore effective sales support measures to sustain contact relationships between corporate clients of financial service providers and corporate sales representatives of financial service providers. Specifically, Sales support has a set annual budget, and measures to minimize sales support costs are to be selected while taking this annual budget into account. In this paper, we consider the tactics of selecting sales support measures according to the rate of change in the effect of contacts with corporate clients.

ABSTRACT. We consider optimal execution times with forward and backward recoveries for two kinds of modules to keep running in computer systems. Fault-tolerant computing techniques are crucial for maintaining reliability, particularly in IoT devices and so on. Almost all computer systems are made of some modules. Even if a module causes an error, it may be fixed by re-executing it. Especially, focusing on forward and backward recovery techniques, we obtain optimal execution times using forward and backward recovery techniques. We assume two modules: One of modules whose error rate is large, and when errors occur, a process executes forward recovery. The other module whose error rate is small, and when errors occur, a process executes backward recovery in which the same process is re-executed. And, we assume that the above system executes the following. The process is a constant process, and it is executed successive tasks with processing times Y_k (k=1, 2, ..., N). The system executes checkpoints at each end of tasks which needs overhead time.

ABSTRACT. Generally, attacks are monitored by multiple security tools such as Firewall, IDS and WAF. There are host-based tools which enable server monitoring by installing software. In terms of host-based tools, if the check is performed frequently, the overhead of system processing becomes larger. Therefore, it is necessary to perform checking under appropriate management policy. This paper formulates a stochastic model for a server system with multiple security tools to check and monitor multiple types of attacks. Cyber attacks are performed so that they circumvent server monitoring. Therefore, random checking is one of the effective ways to detect cyber attacks. This paper assumes that cyber attacks are checked at both random and periodic time, and the cost between attack occurrence and its detection increases with the time non-linearly. In this model, we consider type II error. The total expected cost until cyber attacks are detected is derived and an optimal policy which minimizes it is discussed. Finally, numerical examples are given.

ABSTRACT. This paper considers a modified reliability model of High-performance and Flexible Protocol with Hybrid ARQ. We have already discussed some reliability models of a window flow control scheme based on High-performance and Flexible Protocol. That is, when a server transmits data packets to a client and packet loss has occurred multiple times, it attributes the packet loss occurrences to network congestion and transmits the data packets at longer interval in order to mitigate congestion. These models presumed that when packet loss has occurred multiple times, the data transmission interrupts and the retransmission of all data restarts from the beginning. In this paper, we consider a stochastic model that when packet loss has occurred multiple times, only error packets are retransmitted at longer interval in order to mitigate congestion. That is, when the transmission fails, only error packets are retransmitted and a packet transmission interval is extended. Moreover, when the retransmission has failed again, the server checks it again and the connection is made newly. We derive the mean time until packet transmissions succeed, and discuss an optimal policy analytically.

ABSTRACT. This study addresses the threat of the Asian hornet (Vespa velutina), a significant predator of honeybees, to Taiwan's agricultural ecosystem. We developed an advanced real-time detection system for the Asian hornet using the improved YOLOv7tiny model, enhanced with Space-to-Depth (SPD) and Squeeze-and-Excitation (SE) attention mechanisms for precise identification. This research utilizes data from Roboflow, showed significant improvements in accuracy, precision, recall, F1 score, and mAP. This system holds potential for integration with surveillance cameras, offering a vital tool for protecting honeybees and maintaining ecological balance in agriculture.

ABSTRACT. This study not only proposes a valuable model for predicting missing words in classical Japanese literature but also suggests the potential of this model to be instrumental in repairing the literature. It could significantly advance the field of Natural Language Processing research in the context of historical literature. In recent years, Natural Language Processing has been applied to artificial intelligence programs such as ChatGPT and literary works. However, Natural Language Processing research in Japan has mainly focused on modern Japanese, and research in Japanese classical literature has yet to progress enough. Our research takes a novel approach by attempting to forecast missing words in classical Japanese literature. It creates several language models based on three pieces of classical literature using the Skip-gram of fastText. We employ LOOCV (leave-one-out cross-validation) to validate each model’s accuracy. The results highlight significant differences between the modern language model and our proposed models, which we attribute to the historical context. Next, the experiment demonstrates the efficiency of our model creation method in predicting a missing word. The results of the experiments show that our proposed method can predict words similar to a missing word.

ABSTRACT. This study focuses on the detection of impurities in grain-based spirits, specifically vodka and sorghum liquor, contained in transparent glass bottles. Common contaminants such as sediments, insects, and cotton fibers are identified using machine vision and deep learning technologies. Through the integration of advanced image processing techniques and YOLOv8 object detection, this study demonstrates the potential for enhancing impurity detection in liquor production, ensuring superior product quality and safety standards.

ABSTRACT. Machine learning techniques have enabled the analysis of large volumes of data. However, the pursuit of a model with the ability to accurately predict mortality is a global objective. The research data was collected from Taiwan's Shin-Kong Wu Ho-Su Memorial Hospital. In order to identify the important features, this study used logistic regression (LGR), random forests (RF), and eXGBoost (XGB) algorithms, combined with feature selection methods and grid search, to compare the prediction model. This study aims to develop a mortality prediction model based on a grid search for patients undergoing chronic hemodialysis (CHD) by leveraging artificial intelligence techniques to enhance predictive performance by incorporating more potent risk factors and comorbidities. The XGB model, after undergoing hyperparameter tuning with GSCV, attained an accuracy of 71.1% and an area under the ROC curve (AUC) of 0.759. This performance surpassed that of the models built with LGR, which achieved 66.9% accuracy and an AUC of 0.729, as well as RF models, which achieved 64.7% accuracy and an AUC of 0.741. It can be used to detect CHD patients with a mortality risk and administer suitable interventions to provide appropriate advice to healthcare professionals.