RATE OF INTEREST PREDICTING SYSTEM APPLYING NEURAL NET TO ECONOMETRIC MODEL

摘要

PURPOSE: To simplify the construction of an econometric model and to improve the maintainability of the model by applying a neural net to the model. CONSTITUTION: In a model construction supporting device 10, the construction of an econometric model is supported by a visual correlation network graphic and the support is extended to a neural net. As for the extended neural net, the actual record data file 70 in the past is used as educator data and is learned by a model learning device 13. The neural net is verified by utilizing the same past actual record data by a model verifying device 20. The verified neural net is located at a rate of interest predicting device 32. A simulation executing device 30 performs a control such as the data input/output to this neural net and performs a rage of interest predicting processing. In this case, an exogenous variable scenario to be required for a prediction is estimated from the past economy index data in time series by an exogenous variable time series predicting device 31 and is automatically generated. The generated scenario is stored in a storage part 50, becomes the input of the rate of interest predicting device 32 and is finally stored in a rate of interest prediction data file 80.