Research on forecast of GDP based on process neural network

For the multivariate forecast of Gross Domestic Product (GDP), the common features of traditional forecast methods are difficult to express the time cumulative effects in real forecast, and on the other hand, the factors influencing GDP have very typical timing characteristics. Therefore, in consideration of increasing GDP forecast accuracy, process neural network (PNN) was used into the GDP forecast. Making use of the feature of time-varying input function in PNN, the time and space cumulative effect of GDP influence factors was adequately considered into the forecast, and penalty factor was introduced to PNN training to improve BP algorithm. The GDP forecast model of Heilongjiang Province was established based on the above improved algorithm and it was compared and analyzed with the traditional method. The result shows that the PNN model has higher accuracy.