یادگیری ماشین

Nowadays, one of the biggest challenges in the field of statistical process control (SPC) is how to effectively handle the abundance of big data to evaluate the quality of processes and products. Image data, which is increasingly utilized in the manufacturing and service industries, poses a significant component of this big data. Images offer a cost-effective means of swiftly generating a large volume of data within just a few seconds. Machine vision systems (MVSS) are extensively employed across various industries for obtaining information pertaining to dimensions, geometric features, surface defects, surface finish, as well as the differentiation between conforming and nonconforming products. Consequently, researchers are placing greater emphasis on utilizing statistical process control methods for analyzing image data to detect process variations and defective products, among other goals. This research contribution is highly attractive to practitioners seeking to leverage digital tools for quality management due to its diverse range of potential applications in addressing real-world issues (e.g., Quality 4).Notably, the research effectively integrates machine learning, traditional statistical methods, and image processing within the framework of image-based statistical process monitoring. Choosing appropriate types of images should consider their respective strengths and weaknesses. Binary images are well-suited for monitoring geometric features, grayscale images are suitable for assessing product surfaces, and multi-spectral images prove useful when color represents a critical quality characteristic.This paper presents a systematic overview accompanied by a conceptual classification scheme based on content analysis methodology. The objective is to analyze and categorize prior research that has explored various aspects of statistical process monitoring applied to image data across different industries. The focus is specifically on reliable scientific sources, without any constraints on time limitations. Moreover, drawing from 64 relevant papers in this field, the paper highlights research gaps and provides directions to inspire future studies.

Supplier selection is a crucial aspect of supply chain management. Traditionally, multi-criteria decision-making methods and experts' experience have been the go-to approaches for this process. However, in today's highly competitive business environment, making decisions quickly and accurately has become paramount. Consequently, innovative data-driven technologies and machine learning methods have gained significant importance. Surprisingly, the combination of simulation and machine learning has received limited attention in research endeavors. This study evaluates supplier performance based on specific characteristics utilizing a combination of simulation and machine learning techniques. The research investigates its applications in data-driven decision support for supplier selection. We tackled the supplier selection challenge by simulating the problem using Arena software. The dataset generated from the simulation served as input for our machine learning model. We employed different algorithms, namely Decision Tree (DT), K-Nearest Neighbor (KNN), and Logistic Regression (LR), to analyze the data. Our research demonstrates the remarkable effectiveness of machine learning algorithms in supplier selection. Based on the results, the DT algorithm with 99% accuracy, the LR algorithm with 98% accuracy, and the KNN algorithm with 96% accuracy assign orders to suppliers with the highest probability of delivering them on time. Our approach proves invaluable in analyzing the supplier base and identifying critical suppliers or combinations thereof, minimizing disruptions caused by adverse supplier performance. These findings underscore the potential of integrating advanced computational methods to significantly enhance decision-making processes within supplier selection in supply chain management. Our analysis highlights the pivotal role of combining simulation and machine learning techniques, offering a robust framework for improving supplier selection methods in the fast-paced and competitive landscape of modern industries. This approach not only improves existing methods but also promises a new era in supply chain management. The synergy of simulation and machine learning can revolutionize how businesses choose strategic suppliers and ensure speed and accuracy in decision-making processes.

Couplings are widely used in the industry and this equipment are always subject to defects and failures due to continuous rotation. Vibration analysis is a suitable technique for failure analysis and failure detection of rotating equipment. The purpose of this research is to analyze the failures that occurred in a coupling, whose data was collected in normal state and three failure states with four sensors connected to the coupling. For this purpose, two different types of feature extraction have been used, and seven machine learning algorithms and one deep learning algorithm have been used to classify situations. In this research, the performance of each of the implemented algorithms and the importance of extracted features have been investigated, and the role of sensors and their importance to reduce the number of sensors have been investigated. From the results of this research, we can point out the high importance of the features of the frequency domain in the accuracy of the implemented models, as well as the high efficiency of two sensors for classification.

The purpose of this research is to predict the daily demand for seven yogurt products from the Kale Amol Dairy Products Company, leveraging external influencing factors such as weather conditions, specific calendar days, and product prices. This forecasting is particularly crucial due to the perishable nature of the products, which have a high rate of deterioration.

This study employs a public dataset comprising 12 months of demand history from various food retail stores. Initially, the research uses the classic time series forecasting technique, specifically the Seasonal Auto-Regressive Integrated Moving Average with eXogenous factors (SARIMAX). Subsequently, it implements more advanced machine learning methods, including Long Short-Term Memory (LSTM) neural networks and Convolutional Neural Networks (CNN). The performance of these models is evaluated and compared using accuracy measurement methods such as Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE).

The findings reveal that the CNN method outperforms other forecasting methods in terms of accuracy. Additionally, the LSTM model also demonstrates good performance, although it is not as superior as the CNN method.

Originality/Value:

 This research contributes to the existing literature by focusing on the demand forecasting of perishable goods, which has double importance due to their high deterioration rate. It also highlights the growing interest and effectiveness of advanced machine learning techniques, particularly CNN and LSTM, in improving the accuracy of demand forecasts. This study provides valuable insights for businesses in the dairy industry aiming to enhance their economic profit and competitiveness through better demand forecasting.

This research introduces a nonlinear multi-objective optimization model that is designed to simultaneously optimize profit and customer satisfaction in production systems. The investigated problem includes optimization in complex and uncertain conditions of production, which is faced with resource and time limitations. The proposed model provides optimal solutions for managers by using non-linear objective functions and detailed analysis of operating conditions. This fuzzy logic is combined with machine learning algorithms such as neural networks and reinforcement learning to create an intelligent and flexible model that effectively adapts to sudden changes in dynamic environments. This model uses the combination of non-dominant fourth sorting genetic algorithms (NSGA-IV) and variable selection network (VSN) in a hybrid framework and provides an advanced and multi-faceted approach to solving complex multi-objective optimization problems. Pareto-optimal results obtained from this model indicate its efficient and optimal performance. The proposed model can be used as a practical and strategic source for managers and decision makers in optimizing production and improving customer satisfaction in uncertain and dynamic conditions.

Job satisfaction plays a crucial role in enhancing productivity and reveals intriguing insights that impact the operational effectiveness of organizations. Due to the importance of maintenance units, special attention should be paid to their employees. This study employs a machine learning approach to enhance the performance and job satisfaction of maintenance units through the focus on health, safety, environment, and ergonomics (HSEE). A standardized questionnaire is developed for on HSEE data. Within the neural-fuzzy inference network, inputs such as health and safety protocols, environmental data collection, and its reliability is assessed using Cronbach's alpha coefficient. Subsequently, various adaptive neuro fuzzy inference system (ANFIS) models are utilized to predict job satisfaction based factors, and ergonomics are considered, while job satisfaction serves as the output. Following the selection of the optimal model, individual efficiency levels are assessed and scrutinized based on the calculated error. The findings suggest that enhancing employee job satisfaction relies on prioritizing the enhancement of ergonomics and the work environment.

In this paper, the problem of finding profitable pairs by automatically limiting the search space of pairs using machine learning techniques and integrating an unsupervised learning algorithm, OPTICS, to pair identification and selection in pair trading is discussed. In addition, to optimize the portfolio consisting of pairs of assets and allocate optimal capital to them, a genetic-based algorithm to increase the Sharpe ratio is used. The proposed technique for automatic clustering is better than the conventional methods of searching for pairs of assets used by investors and leads to a higher average rate of return on investment and a higher Sharpe ratio for portfolios in trading using selected pairs of clusters. These calculated evaluation criteria for the portfolio were improved after using a bi-objective optimization genetic algorithm. This study was simulated using intraday price data of a group of stocks in the Tehran Stock Exchange between the years 2015 to 2020 and taking into account the transaction costs.

This research presents an application-oriented approach for developing machine learning models that consider the trade-off between model accuracy, processing speed, and efficient resource utilization, focusing on applications such as wearable smart systems.

A set of models is developed based on the Abstraction and Decision Fusion Architecture (ADFA), and then, using a multi-criteria decision-making approach, the appropriate models for the intended application are identified. The proposed methodology has three main phases: 1) developing models based on the ADFA, 2) defining evaluation criteria, and 3) selecting models using the Fuzzy Analytic Hierarchy Process (FAHP).

The experimental results of this research demonstrate the effectiveness of this approach in developing suitable machine learning models for applications related to wearable devices, such as smart glasses.

Originality/Value: 

This research introduces three innovations: 1) the use of ADFA for developing models for the classification of Persian handwritten characters, 2) defining a new abstraction for summarizing handwritten character images, and 3) developing a fuzzy multi-criteria decision-making approach for mapping the developed models in the ADFA to real-world applications.

Given the competitive market of the insurance industry, customer retention is one of the most important goals of insurance brokers. As a matter of fact, attracting a new customer as well as establishing a relationship with the insurer and gaining his trust requires a lot of money. However, the cost of attracting new customers is much more than retaining existing customers. Accordingly, marketing strategies have shifted from product-oriented and many companies have turned to customer relationship management.Companies and organizations have found that retaining their current customers as their most valuable asset is highly important. Therefore, the strategy of insurance companies is to first retain existing customers and then attract new customers. In this regard, identifying the effective factors in customer turnover is essential. In this paper, data mining methods are used to predict the factors affecting customer dissatisfaction. Based on the empirical results, it has been determined that the customer attraction channel, purchase history and place of insurer are important factors affecting customers dissatisfaction in the insurance industry, respectively.

Selection of the best stocks for the portfolio as well as allocating the optimal amount of capital per stock in the portfolio are serious challenges in investing in the stock market. The use of machine learning capacities in the process of optimal capital allocation among portfolio assets has received less attention and usually, the same weight is assigned to portfolio stocks or traditional risk assessment methods are used to divide capital between portfolio stocks. The common disadvantage of these methods is that they all use simple and inflexible mechanisms to estimate the performance of a set. The purpose of this paper is to show for the first time, that machine learning can be used to create a more effective mechanism for estimating performance, which leads to a more efficient allocation of capital to portfolio stocks.

Our proposed framework, uses two predictive models based on machine learning. In the first step, stocks historical information is used in a return forecasting model, then based on the predicted returns, the appropriate stocks of the portfolio are selected. In the second step, a separate forecasting model predicts portfolio returns by taking into account both the forecasted returns in the first model and the expected risk of the stocks. At the end based on the predicted return of the numerous random portfolios, the appropriate weight for each asset is selected.

Comparing the returns of adjusted portfolios with this model and adjusted portfolios with other portfolio optimization methods shows the superiority of the proposed model.

Originality/Value: 

In this paper, by using machine learning models, the process of selecting the appropriate stock of the portfolio and allocating capital among the candidate stocks is done optimally.

Today, maintenance and repair have become very important in the manufacturing industry. An ecient solution to prevent downtime is to predict equipment failure. Therefore, accurate and correct prediction of breakdown events in the eld of predictive maintenance can be very useful. In general, each prediction will be accompanied by a certain amount of error, which in various ways tries to control this error or limit it to a reasonable amount. In this thesis, a framework has been proposed that species when the system under review will need maintenance and repairs to prevent downtime as much as possible. Therefore, the main purpose of this thesis is to design and implement an ecient combination structure to accurately predict failure events using both standard statistical standard models and machine learning in predictive maintenance. The literature review results indicate that the use of these methods in recent years has led to extensive advances in the eld of providing accurate forecasts and subsequently improved the level of decisions made by managers and decisionmakers. The proposed model is used to predict failure events in benchmark data related to the truck air pressure system. Finally, the performance of the proposed model is compared with other data-driven techniques individually and in combination, which includes logit models, support vector machines, and multilayer perceptron neural networks. According to the numerical values obtained from the nal analysis, the results indicate that the backup vector machine model has higher prediction accuracy than other single models, and also the results indicate the eciency and eectiveness of the proposed parallel combination structure compared to the use of models individually and in series combination in modeling and forecasting issues. The parallel hybrid model improved the accuracy of predictions by an average of 11% in test data and 7% in training data. Therefore, due to the greater accuracy in combining classical statistical models and machine learning in parallel, the use of this combined method to improve the accuracy of predictions in the eld of predictive maintenance is recommended for future studies.

This paper proposes several innovative approaches to model evaluation after obtaining the reinforcement learning model of scheduling with predictive maintenance. To train this model, its reward and loss function must be determined according to the conditions of the workshop environment. One of the innovations of this paper is to provide a definition of the reward function for the issue. This learning model is examined in different modes of work entry into the workshop and the results obtained from other scheduling methods show better outputs. The predictive maintenance model is evaluated by four learning methods and the quality of these models is examined. By selecting and adding the best machine failure model to the scheduling reinforcement learning model, the instant tasks entered into the workshop are assigned to the machines. By comparing the proposed method with the previous ones, the best performance is found and shown.

The emergence of severe acute respiratory syndrome Coronavirus 2 (SARS-CoV-2) in China on December 2019 has led to a global outbreak of COVID-19, which spread worldwide and became an international public health issue. People all over the world must fight in this unexpected battle and the part each individual plays, is important. The health care system has done an excellent job, and the government has done various actions to help the community control the virus spreads as well. On the other hand, in most cases people help to improve the situation alongside the policies. However, the role of information technology in helping various social institutions to fight COVID-19 is hidden and is not well appreciated. The purpose of this study is to discover the hidden role of information technology (IT) that ultimately helps to control the epidemic. Research has shown that strategies for using potential technologies can be beneficial. These IT strategies can also be tailored either to control the epidemic or to support community exclusion during the epidemic which in turn, helps control the spread of infection. This article sheds light on the impact of various technologies that help health care systems, government and public in different aspects to fight COVID-19. In addition to the technologies implemented, this paper also deals with the potential unexpected technologies that can be effective in controlling epidemic conditions for future applications. It has also tried to provide IT-based solutions to deal with the disease epidemic as well.

The emergence of severe acute respiratory syndrome Coronavirus 2 (SARS-CoV-2) in China on December 2019 has led to a global outbreak of COVID-19, which spread worldwide and became an international public health issue. People all over the world must fight in this unexpected battle and the part each individual plays, is important. The health care system has done an excellent job, and the government has done various actions to help the community control the virus spreads as well. On the other hand, in most cases people help to improve the situation alongside the policies. However, the role of information technology in helping various social institutions to fight COVID-19 is hidden and is not well appreciated. The purpose of this study is to discover the hidden role of information technology (IT) that ultimately helps to control the epidemic. Research has shown that strategies for using potential technologies can be beneficial. These IT strategies can also be tailored either to control the epidemic or to support community exclusion during the epidemic which in turn, helps control the spread of infection. This article sheds light on the impact of various technologies that help health care systems, government and public in different aspects to fight COVID-19. In addition to the technologies implemented, this paper also deals with the potential unexpected technologies that can be effective in controlling epidemic conditions for future applications. It has also tried to provide IT-based solutions to deal with the disease epidemic as well.

The emergence of severe acute respiratory syndrome Coronavirus 2 (SARS-CoV-2) in China on December 2019 has led to a global outbreak of COVID-19, which spread worldwide and became an international public health issue. People all over the world must fight in this unexpected battle and the part each individual plays, is important. The health care system has done an excellent job, and the government has done various actions to help the community control the virus spreads as well. On the other hand, in most cases people help to improve the situation alongside the policies. However, the role of information technology in helping various social institutions to fight COVID-19 is hidden and is not well appreciated. The purpose of this study is to discover the hidden role of information technology (IT) that ultimately helps to control the epidemic. Research has shown that strategies for using potential technologies can be beneficial. These IT strategies can also be tailored either to control the epidemic or to support community exclusion during the epidemic which in turn, helps control the spread of infection. This article sheds light on the impact of various technologies that help health care systems, government and public in different aspects to fight COVID-19. In addition to the technologies implemented, this paper also deals with the potential unexpected technologies that can be effective in controlling epidemic conditions for future applications. It has also tried to provide IT-based solutions to deal with the disease epidemic as well.

Thanks to computer networks and advancements in electronic science, cloud computing networks with off-site accessibility make it possible for users to communicate with one another. Cloud computing in relation to your software is still in its infancy. When resources are not in use, the total cost of going to the cloud is almost zero. So it's no surprise that scientific and industry research is moving toward cloud computing. When it comes to distributed systems and multi-agent applications in cloud computing networks, security issues are a must-have requirement. In order to avoid these concerns, intrusion detection systems as well as a series of security frameworks have been introduced. In this research, we try to study the methods presented intelligently, intrusion detection systems and proposed frameworks in the cloud computing network.

Blood donation has an important and critical role to preserve the health and survival of human life. In today's world, despite the enormous scientific advancements and the great developments in medical sciences, adequate supply of healthy blood is one of the challenges and concerns of the medical community in the world. Preserving and supplying the volume of blood required in blood banks of each region, and the diverse blood groups with the connections between them, with assuming that the number of blood groups are rarer; makes the prediction and planning of blood donation more and more complicated and important during the time. The use of data mining in hospitals and blood transfer centers databases helps in the discovery of relations, so that they can have a future prediction based on the past information. Accordingly, they have better diagnosed and successful cure various illnesses and show the patterns of new injuries. In this paper, we try to use data mining and machine learning techniques in decision making levels at mentioned field, to use this mechanism for prediction that how much blood will be donate to blood transfusion centers and blood banks in different period time, to estimate and supply the required blood volume of blood banks in different areas. In this regard, we use several classification algorithms in supervised learning for the prediction, including decision tree algorithms, KNN, SVM and MLP, these algorithms are implemented to predict and results of accuracy are presented.