Improvement of Private Investment Estimation and Forecasting in Ambiguous and Complex Environments by a Flexible Neural Network-Fuzzy Regression Algorithm

This study presents an integrated algorithm to improve the private investment forecasting in ambiguous and complex environments. The unique nature of this algorithm lies in the integration of the Artificial Neural Network (ANN), Fuzzy Linear Regression (FLR), and Conventional Linear Regression (CLR) approaches. Hence, it can be easily applied to certain, ambiguous, or complex and nonlinear environments due to its flexibility. This algorithm is applied to forecast private investment in Iran by considering annually data of private investment from 1971 to 2003. In this study, private investment in Iran is viewed as the resultant of standard economic indicators, which are amount of savings, public investment, rate of interest, and variation in gross national income. According to the proposed algorithm, six well-known FLR, ANN, and CLR approaches are applied to the collected data. The sensitivity analysis of Mean Absolute Percentage Error (MAPE) results indicated that ANN outperforms its rivals due to the nonlinear and complex nature of private investment function in Iran. However, the superiority of FLR over CLR is referred to unstable and ambiguous economic status of Iran. The algorithm of this study may be easily used for other data sets. This is the first study that presents an ANN-FLR-CLR algorithm for private investment forecast capable of handling complexity, non-linearity, and ambiguity.

Forecasting; Private Investment; Sensitivity Analysis; Artificial Neural Network; Fuzzy Linear Regression; Complexity; Ambiguity