Ebook: Machine Learning and Artificial Intelligence
Machine learning (ML) and artificial intelligence (AI) applications are now so pervasive that they have become indispensable facilitators which improve the quality of all our daily lives.
This book presents the proceeding of MLIS 2022, the 4th International Conference on Machine Learning and Intelligent Systems, held as a virtual event due to the continued uncertainty caused by the Covid-19 pandemic and hosted in Seoul, South Korea from 8 to 11 November 2022. The aim of the annual MLIS conference is to provide a platform for the exchange of the most recent scientific and technological advances in the field of machine learning and intelligent systems, and to strengthen links in the scientific community in related research areas. Scientific topics covered at MLIS 2022 included data mining, image processing, neural networks, natural language processing, video processing, computational intelligence, expert systems, human-computer interaction, deep learning, and robotics. The book contains the 20 papers selected for acceptance after a rigorous peer review process from the more than 90 full papers submitted. Selection criteria were based on originality, scientific/practical significance, compelling logical reasoning and language, and the 20 papers included here all provide either innovative and original ideas or results of general significance in the field of ML and AI.
Providing an overview of the latest research and developments in machine learning and artificial intelligence, the book will be of interest to all those working in the field.
During the recent pandemic, evidence from Machine Learning (ML) and Artificial Intelligence (AI) applications encouraged researchers by providing a new angle from which to fight the novel corona virus outbreak, as well as better facilitating our daily life.
The International Conference on Machine Learning and Intelligent Systems (MLIS) is an annual conference that aims to provide a platform for the exchange of the most recent scientific and technological advances in the field of machine learning and intelligent systems, and to strengthen links in the scientific community in related research areas. MLIS 2022 was initially scheduled to be held in a hybrid event from 8th–11th November; live in Seoul, South Korea, and online via MS TEAMS, but was changed to a fully virtual conference due to the continued uncertainty caused by the global pandemic. Scientific topics covered at MLIS 2022 include data mining, image processing, neural networks, natural language processing, video processing, computational intelligence, expert systems, human-computer interaction, deep learning, and robotics.
This volume presents the proceeding of the 4th International Conference on Machine Learning and Intelligent Systems (MLIS 2022). Rigorous peer review was conducted on each paper by at least two reviewers, and 20 accepted papers were selected from the more than 90 full papers submitted based on originality, scientific/practical significance, compelling logical reasoning and language. We believe that the 20 papers selected all provide either innovative and original ideas or results of general significance in the field of ML and AI.
On behalf of the conference organizers, we would like to take this opportunity to express our sincere thanks to the guest editor and reviewers for their tremendous efforts and dedication to the conference and to the authors for their relevant contributions to the conference, as well as our colleagues from IOS Press for their support and their tireless efforts in the publication of the conference proceedings. We believe that with such support and contributions, future MLIS Conferences will also scale new heights.
Prof. Jon-Lark Kim
Sogang University, South Korea
This paper investigated the attitudes of 702 college students toward the implementation of fully online learning during the COVID-19 pandemic. Toward this goal, responses of the students were collected and analyzed through hierarchical cluster and sentiment analyses using the R software. Hierarchical cluster analysis revealed hopeful and apprehensive attitudes toward online learning. Advantages of online learning emerged as positive sentiments while challenges and their impact on mental health emerged as negative sentiments. It is concluded that online learning is a promising platform of learning provided that its shortcomings are addressed. Implications to teaching are offered.
When it comes to the digital age, computers and technological devices have grown more pervasive than in the past. As of now, technology has been considered to have a major role in every aspect of life. In this case, it is the health and cognitive senses of individuals. In terms of human perception, it is often associated with both long-term and short-term memory or active memory, which are recognized patterns by automated systems. When creating user interfaces, it is necessary to account for the cognitive strengths and limitations of humans. People can get an overview of large datasets if appropriately presented in a proper manner, which can help increase the efficacy and ability of information management. However, one of the most frequent data management issues is inadequate design and data collecting, particularly in the focus theme of elderly health and wellbeing. Infographics are a formal way to help people interpret complex health information more easily and avoid possible mistakes in data management. In the healthcare industry, the system of augmented reality (AR) has been shown to facilitate the procedures of healthcare using AR booklet and mobile applications for elderly-health-related content and healthcare management for older individuals.
In the context of digital informatization, the Internet is changing the way of human existence. The rapid development of the Internet has promoted the use of smartphones in people’s daily lives, and at the same time, a large number of applications running on different operating system environments have appeared on the market. Predicting the duration of application usage is crucial for the management planning of related companies and the good life of users. In this work, a dataset containing time series of user application usage information is considered and the problem of “application usage” forecast is being addressed. The dataset used in this work is based on reliable and realistic user records of the usage of the application. Firstly, this paper investigates suitable forecast models for application development on the applied user usage time dataset, which includes neural network algorithms and ensemble algorithms, among others. Then, an Explainable Artificial Intelligence Approach (SHAP) is introduced to explain the selected optimal forecast models, thus enhancing user trust of the forecasting models. The forecast results show that the ensemble models perform better in the time series dataset of user application usage information, especially LightGBM has more obvious advantages. Explanation results show that the frequency of use of the target variables, category and lagged nature are important features in the forecast of the application dataset.
This paper continued the study of the complexity behavior of the satisfiability analysis of hard random formulas given in the conjunctive normal form (CNF). In the 3-CNF formula, each clause contains three literals of logical variables. The number of logical variables in the formula is N. In this paper, the SAT solver is improved by introducing equality reduction and pure literal identification procedures. The solver improvement reduced the exponent (with base 2) from N/20.86 to N/21.41 with an R=4.6 ratio of the number of clauses to the number of variables. The results show that the efficiency of the pure literals identification procedure decreases as R increases. An important part of the study is an empirical estimation of the algorithmic complexity of the SAT problem with large number of variables. The proposed method gives a convenient lower bound on the complexity of analysis for random 3-CNF formulas. We estimated the algorithmic complexity for the range N=256÷8192. The exponential dependence of complexity on N for random 3-CNF formulas at a fixed value of R is demonstrated in this range.
Recent studies on fuel cell design have showed that the use of simulation tools is beneficial in terms of saving time and money. Current density management is still a key research problem for several technologies, including Direct Borohydride Fuel Cell (DBFC). This paper describes a systematic machine learning technique for estimating the cell current density for DBFC as a function of various input factors. Artificial Neural Networks (ANN) and Decision Tree Regressor (DTR) are two popular machine learning models that were trained and evaluated for the current density simulation using a conducted fuel cell experiments presented in previous research. The ANN model performed the better than the DTR model in the simulation, with a mean absolute error of 3.00015 for training and 5.57614 for testing. The simulation exhibits very small error values, indicating that the suggested approaches accurately mirror real-world DBFC process.
The key goal of this work is to explore interactions and discursive exchanges between social users, to extract information towards decision support. We analyzed customer-generated data on Facebook, during a period of a ten-day strike, of a well-known airline company. The main goal was to check service and responsiveness of the airline, and also to develop indicators that might enable reviewing and reinforce strategies to be used in customer service response to strike events. The authors aim to investigate the possibility of structuring data, collected from OSN’s, incorporating human interaction and network structure, using SNA to study the network from a duo fold manner: the web discourse, which depends on the transmission of information; and the interaction among social users, as information disseminators. Our work intends to determine whether social users and their interactions are consistent with the creation of indicators for decision support.
With the help of the space-to-depth and depth-to-space modules, we provide a convolutional neural network design for depth estimation. We show designs that down sample the spatial information of the picture utilizing space-to-depth (SD) as opposed to the widely used pooling methods (Max-pooling and Average-pooling). The space-to-depth module may shrink the image while maintaining the spatial information of the image in the form of additional depth information. This technique is far superior to Max-pooling, which diminishes the image’s information and features. We also suggest a lightweight decoder step that builds a high-resolution depth map out of many low-resolution feature maps using the depth-to-space (DS) module. The suggested architecture effectively learns depth estimation with high processing speed and accuracy. We trained and evaluated our suggested model on NYU-depthV2 dataset and attained low error values (RMSE=0.342) and high delta accuracies (δ3=0.996) at a fast-processing speed (25Fps).
Rolling bearings are treated as important machinery power components, faults of rolling bearings affect machinery operation, so an intelligent fault diagnosis method is very useful of safety operation in rolling bearings. This paper proposes a novel fault diagnosis method based on improved Adaptive Deep Convolution Neural Networks algorithm to realize fault recognition for rolling bearings. First, the Continuous Wavelet Transform (CWT) method is applied to the time-frequency decomposition of vibration signals and extract feature information images for training and testing. Second, to further improve self-learning ability of the Adaptive Deep Convolution Neural Network (ADCNN) in feature images, the Multiple Channels ADCNN method is proposed to classify different fault image types for the rolling bearing. Finally, fault images corresponding to different health states of the rolling bearing are applied to the proposed method, the experiment proves that the proposed method has a better performance for fault recognition in rolling bearings.
Condition monitoring becomes an integral part of the industrial manufacturing system to ensure a safe working environment and reduce the cost of maintenance. Involving deep learning techniques in fault diagnosis methods not only increases the accuracy and reliability of the system but also reduces the operation time and hassle of the manual feature extraction process. In this paper, a complete framework for fault classification is introduced by using the vibration signals of bearings containing normal and faulty conditions. Firstly, the frequency spectrums of the time-series signals are generated with FFT and transformed the 1-D signal into 2-D images with the recurrence plots (RP) algorithm. Finally, a deep CNN model is designed to classify the bearing conditions with the extracted high-level features from the RP-based image dataset. The images show a distinct pattern in every bearing condition and the CNN model can achieve 99.24% accuracy to classify three different bearing conditions. The image classification-based fault diagnosis approach is automated and eliminates the disadvantages of the manual feature extraction process. The generated images with RP were also trained with three predefined CNN models to verify the effectiveness of the fault patterns. Finally, the comparative analysis demonstrates that the proposed method outperforms other researchers’ approaches both in terms of classification accuracy and computational cost.
In this work, the wavelet transformation (WT) under the context of convolution neural network (CNN) is developed and applied for breast cancer detection. The main objective is to investigate the effectiveness of the WCNN pooling architecture when compared to other two famous pooling strategies; max and average pooling, particularly targeting at the features extraction and classifying the phases of breast cancer by avoiding the under and overfitting problems. It is discovered in this work that the combination of WT and CNN outperforms the traditional and typical CNNs (with 96.49% of accuracy 95.81% of precision, 96.73% of recall and 96.23% of F measure).
This work aims to numerically investigate the performance of the multiquadric (MQ) radial basis function in more general formats for image reconstruction applications. Desired features, i.e., accuracy and shape parameter sensitivity, of each form is numerically compared and explored. The famous Lena image is damaged using two levels of damage: 20% and 40%, in a Salt-and-Pepper manner. It has been discovered in this work that β=3/2 produces reasonably good accuracy and is least affected by the change in shape parameter while keeping both the CPU time and the condition number reasonably acceptable. This finding is promising and useful for further applications of MQ in more complex contexts.
Under the architecture of a neural network, this work proposes and applies three multiquadric radial basis function (MQ-RBF) interpolation schemes; The Common Local Radial Basis Function Scheme (CLRBF), The Iterative Local Radial Basis Function Scheme (ILRBF), and The Radius Local Radial Basis Function Scheme (RLRBF). The schemes are designed to perform locally to overcome drawbacks normally encountered when using a global one. The famous Franke function in two dimensions is numerically tackled. It is revealed in this work that all three local methods outperform the traditional MQ interpolation in terms of both CPU-time and condition number, while the accuracy is overall acceptable, particularly when the number of nodes increases. This finding indicates their potential for dealing with bigger datasets and more complex problems.
Each year, many people pass away in car accidents. Drunk driving, distracted driving, and fatigued driving are the leading causes of car accidents. Some studies have noted that the features relevant to drunk driving and fatigued driving are driver’s alcohol consumption, smoking, and carbon monoxide (CO) concentration in the car. Hence, it is important and necessary to detect these factors during driving to avoid driver drowsiness and prevent car collisions. In this paper, we develop an IoT-based preventative system to monitor in-vehicle air quality and check for driver head bending. To ensure users can easily deploy the proposed system, we used low-cost sensors and tiny IoT devices to run the system. The suggested system uses low-cost gas sensors including MQ-2, MQ-3, and MQ-7 to track air quality in the car using Arduino D1. Additionally, we developed a Banana Pi camera setup to run a customized Teachable Machine model to ascertain whether or not the driver is paying attention to the road. We showcase the proposed system on the ThingSpeak cloud platform. Finally, in the proposed system, the buzzer and speaker will sound alarm warnings to the driver and occupants if the gas concentration in the car exceeds a specified value or driver distraction is detected.
This article presents an intelligent system that can create personalized tours in Tokyo for individual travelers. Such a system provides tourist spots selection and recommends tour plan according to user’s preference. Using six designed features reflecting tourist spots’ characteristics as well as user’s interests, tourist spots are selected based on matching degree of the feature vectors of related spots and that of user’s pre-input interests. Given user’s starting point, an optimal tour schedule is then proposed considering necessary transportation and moving time, opening-hours, and sightseeing time needed. The system is designed to be user-friendly and interactive, allowing user to make unlimited adjustments directly on system proposed tour plan, and offering emergency help in the event of injury, illness, or unexpected tour changes.
The purpose of this paper is to propose and study the structure of wavelet transformation (WT) and convolution neural networks (CNN). To get more insights into its effectiveness, three WCNN architectures are designed and tested against one another seeking which model provides the best performance in breast cancer detection using histopathological images. The Breast cancer histopathological database (BreakHis) is used for this task.
This article presents an education tool that offers junior learners, such as junior high school students. basic knowledge of machine learning through designed games or quizzes. Learners can study basic ideas of k-NN classification, linear regression, and k-means clustering when enjoying their game playing. The system can also run on a smartphone for the convenience of learners. With such a tool, learners are expected to gain a better understanding of fundamentals of AI and machine learning and increase their interests in future development of AI applications.
Frost causes damage to crops. Predicting the frost occurrence in advance is highly valuable for practitioners taking possible frost-prevention measures. In our previous study, we have applied machine learning to forecast frost occurrence using two methods. To better support user with a potential trend of frost occurrence in a future period, we propose an integrated system of frost forecast taking the advantages of the two methods.
A large number of sensors based on Internet of Things (IoT) technology are now widely deployed in artificial intelligence, health care monitoring, air quality monitoring, and other fields. The sensors require high power consumption for real-time monitoring data. Some studies have suggested using solar energy for the primary power source to operate sensors. However, due to uncertain climate change, solar energy supply cannot always provide sufficient voltage to operate sensors. Consequently, some abnormal behavior events frequently occur in the IoT system using solar energy. Abnormal detection is a typical imbalanced learning problem due to the very rare amount of abnormal events. Under such data with skewed class distribution, classic classification models fail to provide reliable classification results with abnormal events. Under this condition, in this paper, deploying solar energy supply, we developed an IoT-based system using Arduino Microcontroller and Banana Pi, in which, the SMOTE-PSO algorithm is utilized to improve classification accuracies on abnormal event data in our system. Finally, two types of SVM kernel functions are used to verify classification capability in the developed IoT system.
Most group activity recognition models focus mainly on spatio-temporal features from the players in sports games. Often they do not pay enough attention to the game object, which heavily affects not only individual action but also a group activity. We propose a new group activity recognition model for sports games that incorporates players’ motion information and game object positional information. The proposed method uses a transformer encoder for temporal feature extraction and a ’simple’ conventional convolutional neural network for extracting spatial features and fusing them with the relative ball position-embedded features. The experimental results show that our model achieved comparable results to state-of-the-art methods on the Volleyball dataset by using only one transformer encoder block and the ball position.
Understanding one’s own behavior is challenging in itself; understanding a group of different individuals and the many relationships between these individuals is even more complex. Imagine the amazing complexity of a large system made up of thousands of individuals and hundreds of groups, with countless relationships between those individuals and groups. However, despite this difficulty, organizations must be managed. Indeed, ultimately the organization’s work is done by people, individually or collectively, alone or in combination with technology. Therefore, organizational behavior management is the central task of management work – it involves understanding the behavior patterns of individuals, groups, and organizations, predicting what behavioral reactions will be elicited by various managerial actions and finally applying this understanding. Undeniably, society’s work is often done by organizations, and the role of management is to make organizations do that work. Without it, our entire society would quickly stop operating. Not only would the products you have come to know and love swiftly to evaporate from store shelves; food itself would suddenly become scarce, having drastic effects on huge numbers of people. To this end, the term Technology- Enhanced Learning is used to support workers’ learning about technology; the gap between what is understood to be satisfactory and the current level of knowledge of the workforce is addressed by a Logic-programming-based Social Computing Framework entitled An Entropic Approach to Knowledge Representation and Reasoning, which relies on computational structures built on Artificial Neural Networks and Cases -based Thinking, as well as predictions and/or assessments, to empower the level of knowledge of the employees, here in technology, later in other areas.