Evaluation and detection of fraud in the supply chain process based on the integrated approach of ANN-Big data

In recent decades, competition to provide superior value to customers has shifted from competition between companies to competition between supply chains. Appropriate design of the supply chain with regard to the economic, social and environmental dimensions at the strategic, tactical and operational levels guarantees the survival and sustainable development of enterprises active in each part of the supply chain. One of the important issues in the supply chain is the existence of fraud and the risks associated with it throughout the chain. Also, according to the amount of scientific and experimental research conducted in recent years, the problem of detecting and evaluating fraud using computational methods alone It is an important topic for research. The most important challenges of fraud detection and evaluation are the unavailability of real data sets, the existence of asymmetric data sets, the large size of data sets, the dynamic behavior of fraudsters and the dispersion of fraudulent events, which are the result of these two Uncertainty and ambiguity in decision making. Therefore, the aim of this research is to provide a model to detect and evaluate fraud in the supply chain process based on the ANN-Big data integrated approach. In terms of its purpose, this research is a developmental and applied research. First, due to the creation of a comprehensive model, taking into account the various sources of decision-making regarding the legitimacy or fraud of electronic card transactions, and the requirements of big data analysis, it tries to develop theoretical knowledge in this field, this research can be considered Considering the nature of a mathematical analytical research, on the basis of this research, in terms of approach, it is a quantitative research. The inter-industry standard process of data mining or CRISP DM (Chapman et al. 2000) has been used as the analysis methodology. Clustering models and K-means algorithms, Kohonen neural network, anomaly detection based on clustering, hidden Markov model, parallel processing method of reduction mapping and fuzzy Dempster cipher fusion model are used. The analysis was done using MATLAB and R software. The proposed model has been able to show superior performance compared to other models in terms of output and execution time.