Regional Economic Development Forecast Based on Improved Deep Convolution Neural Network

In recent years, the regional economic development forecast across various industries is essential for informed policymaking. However, the regional economic development consists significant challenges such as complex dependencies and long-term forecast stability. To overcome these challenges, an Improved Deep Convolutional Neural Network (DCNN) method is employed for forecasting annual Gross Domestic Product (GDP); primary, tertiary and secondary growth values. In this process, the original industrial data are preprocessed with a Minimum-Maximum (Min-Max) normalization approach and correspondingly BoxCox technique is used to attain normal distribution by eliminating the skewness. After that, panel data spanning of consecutive years is designed to track multiple entities. Further, the proposed Improved DCNN method is employed to forecast the economic development by capturing temporal dependencies from panel data and predicts the future GDP growth index effectively. From the results, the proposed Improved DCNN attains higher results in terms of R-Squared (RA2)value is 0.96 for all 31 provinces than the existing DCNN method.