Data mining applied in food trade network

The proposed study deals with the design and the development of a Decision Support System (DSS) platform suitable for the global distribution system (GDS). Precisely, the prototype platform combines artificial intelligence and data mining algorithms to process data collected into a Cassandra Big Data system. In the first part of the paper platform architectures together with all the adopted frameworks including Key Performance Indicators (KPIs) definitions and risk mapping design have been discussed. In the second part data mining algorithms have been applied in order to predict main KPIs. The adopted artificial neural networks architectures are Long Short-Term Memory (LSTM), standard Recurrent Neural Network (RNN) and Gated Recurrent Units (GRU). A dataset with KPIs has been generated in order to test the algorithms. All performed algorithms show a good matching with the generated dataset, thus proving to be the correct approach to predict KPIs. The best performances in terms of Accuracy and Loss are reached by using the standard RNN. The proposed platform represents a solution to increase the Knowledge Base (KB) for a strategic marketing and advanced business intelligence operations.