شبکه عصبی

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.

In this research, a hybrid approach based on fuzzy logic and artificial neural network is presented to predict the failure of machines in order to increase productivity. The subject of this research is one of the factories of the automobile industry named Diaco Ide Aria, which operates in the field of automobile parts production. Preventive maintenance requires correct prediction of breakdowns and accidents, equipment and machines so that productivity can be increased by timely and correct maintenance of machines as well as fixing defects and breakdowns. To model the multi-layer perceptron fuzzy-neural network (MLP), first, 100 failures and stops were collected in a period of 15 months and then entered into MATLAB software. The obtained results show that the implementation of fuzzy-neural network and the prediction of machine failure time has reduced the duration and cost of repairs. Therefore, the working time and accessibility of the machines increased and ultimately increased the productivity by 57%, also, the accuracy of the developed neural-fuzzy model was estimated at 94%.

The purpose of this study is to combine two methods of data envelopment analysis and neural network in order to provide an optimal model for ranking inefficiency factors in the Iranian banking industry. First, through the study of theoretical foundations and interviews with banking experts, efficiency evaluation indicators in the banking industry were identified and finalized. In order to evaluate the efficiency of the units in the statistical population of the study, data envelopment analysis technique was used, especially the modified goal programming data envelopment analysis model, which was identified from 32 managements, 3 efficient managements and 29 inefficient managements. Then, the branches of inefficient management were evaluated and using the information of inefficient branches, the neural network matrix was prepared to identify the causes of inefficiency and the results were analyzed with different neural network models. The model with the lowest mean square error will be selected as the optimal model to determine the inefficiency factors. As a result, the self-organized mapping model with hyperbolic tangent transfer function and 0.9 momentum training rule was selected. By analyzing the sensitivity of this method, the indicators of provincial liquidity share, personnel distribution and operating costs were selected as the most important factors of inefficiency.

The purpose of this study is to explain the impact of green supply chain on economic performance, emphasizing the mediating role of green innovation, environmental management and quality management in companies listed on the Tehran Stock Exchange.

In this research, the multivariate perceptron neural network approach and modeling of variance-based structural equations in SPSS26 and SmartPls3.3.3 software have been investigated.

The results show that green supply chain management affects economic performance and increases green innovation, environmental management and quality management of economic performance. The establishment of a green supply chain has led to the observance of environmental requirements, and by observing environmental requirements, labor productivity is improved through specialized training of employees.

 In previous studies have not considered the pros and cons of the relationship between environmental management and labor productivity. In this study, according to the selection of companies listed on the Tehran Stock Exchange during the covid-19 period and the statistical sample selected by systematic elimination method and available sampling, these views were examined. Also, in the research method, most researches have only fitted the model with structural equations and regression equations, while in this research, the proposed model fits with multilayer perceptron neural network method and variance-based structural equations and finally to evaluate the performance prediction comparison model. Economically, the root mean square error index is used

The proposed model is a time-cost-quality trade-off model with three objective functions: minimizing project completion time, minimizing total project cost, and maximizing total quality of activities in a PERT network with multi-mode activities.

After presenting the appropriate mathematical model, based on the design of the experiments, the possible levels of each decision variable were determined. Then, using the simulation process, random values ​​of decision variables and response variables were obtained each time, and by using neural networks, we established a neural network model. To solve this model, two algorithms NSGA-II and MOPSO were used.

To evaluate the efficiency of the model, the designed model was implemented in the maintenance department of Abtin Ardakan Steel Company. According to the results, it is found that the NSGA-II algorithm has better performance than the MOPSO algorithm.

In this paper, a model was presented that by eliminating unrealistic assumptions and taking into account the realities of the project is closer to reality than the models presented in this field and has more application in practice.

The supply and demand management and planning of the blood supply chain have complexities considering the uncertainty in demand and supply. The supply uncertainty often results from the random demand as well as unpredictable and irregular behavior of the people involved in blood donation as the only source of blood supply.  In addition, the perishability and short lifespan of blood and blood products are the problems that should be considered.  Adequate and healthy blood supply plays an essential role in health systems.  Therefore, the prediction of demand to prevent endangering the patients’ health due to the inventory shortages on one hand, and increasing inventory which increases wastages and government expenses on the other hand, have attracted the attention of researchers.  Accordingly, the aim of the present study is to predict the demand for blood bank products using neural networks.  Neural networks have the ability to properly predict the future based on certain trends of the past by adjusting some parameters.  To collect the necessary data for this research, the statistics, databases and computer networks of the blood transfusion network of Zanjan province have been selected. For data analysis, the existing facilities and functions of the MATLAB software have been used. Based on previous demands of blood products, the research findings demonstrate that the best neural network model to predict demand has two delays and five neurons in the hidden layer.  The results also show that the error values in all three blood products are close to each other but have different values.

"Taguchi" method is a conventional method for quality control in offline mode. This method is applied to design and select the best level of parameters for designing a better method to make high quality products. Taguchi method is one-response and it is a disadvantage for it. In the real world applications, there are several problems with some indicators of quality. Therefore, Taguchi method is not appropriate for optimizing multi-response problems and we need to an engineering and optimizing method to judge the best combination of parameters. On the other hand, due to some uncontrollable factors or the impossibility of experimental conditions, only some of experiments are implemented and a large number of them are incomplete. In this paper, to simulate the remaining experiments the Back-Propagation neural network is used. To overcome one-response problem in Taguchi method, the data envelopment analysis (DEA) is used. Since the results obtained from the neural network are uncertain, DEA model with interval grey data is used. To implement this approach and to identify effective factors, the wear characteristics of composite material PBT, the combined approach based on Taguchi method, neural network and DEA is used and the results will be analyzed.

A few years after the implementation of computerize maintenance management systems (CMMS), showing the analysis results of the collected data in these systems is necessary and important for continuing the usage of these systems. Equipment downtime and cost data can identify patterns of equipment faults and equipment costs. It can also show deficiencies in the implementation of maintenance systems. Detection of failures and important factors of failures such as the type of mission, geographical conditions, quality of parts and soon helped to classify and determine the cost norms based on the type of mission. These analyses can help managers to make appropriate decisions more accurately and make proposals for improving the systems.Data mining can provide a view of equipment readiness and the factors that affecting them. This view can be used to estimate the costs of maintenance and repair of equipment. Knowing the time of certain failures can be used to plan maintenance program. The issues are useful for improving knowledge and culture in the implementation of equipment. These analyses caused better and more accurate maintenance and usage of equipment.In this paper, we introduce some data mining applications in maintenance and repairs activities. It shows some hidden rules discovered from data of equipment maintenance systems in a case. Clustering models, neural networks, decision trees, visualization and descriptive statistics are used in this case study and some of results introduce in this paper. The findings show that there are certain pattern depending on the type of mission and how to use the equipment.

In many real commercial and administrative issues, relations are often complex, non-routine and are not predictable by conventional methods. Considering the importance of feasibility studies in making decision to start a production activity, forecasting in marketing studies is very important. There are lots of tools and techniques which are used for having an exact prediction, neural networks can be utilized for forecasting with high degrees of accuracy. The purpose of this article is to demonstrate the preference of using neural networks in forecasting nonlinear processes in comparison with conventional techniques and also to increase its accuracy by using economic parameters such as inflation and exchange rates. As a case study, this paper uses the production rate data of PET bottles from 1379 to 1392, then the production rate of 1393 is predicted by using artificial neural networks and nonlinear models. For validating the model, indexes MAPE and MSE obtained from these methods are compared. The result shows the preference of using the neural network for prediction in comparison with time and exponential series techniques, due to the lowest error in forecasting

The increasing consumption of electricity in Iran is one of the greatest concerns of the government. Using the subsidy-based pricing system is one of the main reasons of improper pattern of residential electricity consumption that has imposed great cost over the government due to the increased number of consumers and their improper way of consuming electricity. In this paper، we analyze the factors that affect residential electricity demand using artificial neural network (ANN) and predict the amount of electricity consumption in 2006 (the end of the year in which subsides are being removed) by definition of five different price scenarios. The per-capita residential electricity consumption is considered as a dependent variable of the model. Electricity price، GDP per capita، macroeconomic fluctuations and a variable representing weather temperatures are used as explanatory factors. The proposed model has a good explaining capability (R=0. 996) and with predicting independent variables up to 2016، the dependent variable were predicted using procedures like time series and ARIMA. The achieved results show that the price factors have limited role in defining the pattern of residential electricity consumption. So small changes in electricity price will not reduce the electricity consumption and committing scenarios with gradual changes in price will not lead to the reduction of electricity consumption. Therefore، it is necessary for the government to commit scenarios with significant increase of prices in order to correct the pattern of residential electricity consumption; otherwise، the electricity demand will increase uncontrollably due to the increasing population and consumption.

Banks as financial institutions must estimate the credit risk of their debtors. This is the basis of pricing a loan, determining appropriate interest rates and determining the mortgage required to each borrower. Since the continuity of bank activities largely depends on the amount of credit losses in a particular period, banks should consider the credit quality of their loan portfolio as a collection of debts.In this paper, the calculation of credit risk of corporates applying for loans has been investigated. Using fuzzy analytic hierarchy process, effective criteria for credit risk have been analyzed.The neural network is used to extract an open box model that describes the relationship between effective criteria and the credit risk of the companies who apply for a loan. Neural network model has been run with historical data. Observations have been based on 174 corporate who had taken out a loan from a major Iranian bank named Mellat (All loans had been made during 2005 to 2008).The output of the model can predict credit risk of a corporate by at least 84% accuracy.

During the past decade in Iran, Information Technology (IT) has had a deep impact on economy. Many companies have invested on IT and it’s complementary. Despite a vast percentage of firm’s budget is spent on IT continues increasingly, there are evidence which has been illustrate on failure of firms in obtaining the benefits of these expenditures within expected period. This fact nevertheless has caused what may be named as productivity paradox. The ICT productivity paradox (contradictory of research on IT effects on performance in organizations with what the researchers expected) and its complementary assets (by performing such investments, it can be ensure to impact of IT on organizational performance) besides the growing of ICT in new product development project, is faced organizations with one serious question:” in a NPD (New Product Development) project, what is the investment priorities on ICT complementary in order to make the maximum performance level of a project?”. To find the answer, in this paper we present an extensive case study within research strategic units and new product development departments of Iran Khodro, and conceptual research model has been analyzed using neural networks technique. The results for a specific project show that a resource development (such as feasibility study for each product and doing financial assessment) and human resources will have the most important role in increasing project performance in three field of financial, management, and general. The management shall focus on leading and inspiring the team members of projects in order to gain project objectives, meanwhile as it can be possible the project members must be single task. Training the team and making the culture of using advanced technology is one of the other aspects of human resource activity as an IT complementary. In fact by investing on these complementary factors and using their beneficiary besides the IT, each related company would be imagining of performance increment.