Agent-Based Pattern Mining of Discredited Activities in Public Services

Along with the applications of computer and database techniques, a large even huge amount of data has been produced and collected in the field of public service. Many valuable regulations hide in the raw data. It is necessary that new technique is studied to help the manager make proper decision. We propose a data mining model which is based on multi-agent technique to mine the discredit activities in the water sale public service. The multilayer feed-forward neural network (MFNN) trained by the improved backpropagation (BP) algorithm and decision tree algorithm have been employed in the model. Five agents have been developed to mine the discredit patterns and evaluate the users´ credit according to the model. A case study of Daqing oilfield water supplying payment system has been implemented for verifying this model