Research On The Prediction Model Of Regional Human Capital Based On GA-PSO Optimizing BP Neural Network

This paper establishes a prediction model based on GA-PSO optimizing BP neural network, and then empirically analyses China's regional human capital by using the model.Firstly, the index evaluation system of regional human capital investment is established, selecting China's regional data to form the Original Sample Set According to original index or original index generated.Secondly, the BP neural network prediction model based on GA-PSO optimizing is established, and the best value and threshold for the training of BP neural network can be obtained by GA-PSO. Using model, it can be successfully used to make predictive simulation and analysis.There are four steps:①Determining the structure of BP neural network, providing the Primitive experimental result of optimization by variation ability and global searching ability of GA.②Strengthening the local search, improving convergence speed, performing fine searching based on PSO. The best weights and thresholds for the training of BP neural network can be obtained.③The above results are used to train normally of BP neural network as initial weights and thresholds in order to predict level of regional human capital.Finally, the predictive simulation of the established model is carried on in order to detect the feasibility of the model. There are empirical researches of lateral prediction,vertical prediction based on Matlab2007a and some significant conclusions are drawn, which is helpful to formulate policy.The main creative point:1.According to the characteristics of GA,PSO,BP neural network, them are connected in series to set up the prediction model of regional human capital. This paper puts forward the concept of mixing degree, which blends GA with PSO and makes easy to analyze the Proportion of Populations.2.Prediction Method for the level of regional human capital is divided into lateral prediction and vertical prediction.It could predict one year's level of regional human capital, as well as any one year's level of regional human capital by training with continuous and previous several years' datum.3.Increasing noise samples for the training of prediction model by means of extending datum, which could meet the demand of sufficient data for the training and avoid being trapped in local optimum.4.Combined skills of algorithms are used in the prediction model in comparison to traditional GA and PSO.Based on the information above, there are four chapters in this paper.In Chapter 1: introduction part;In Chapter 2: preparation knowledge, theoretical Basis of BP neural network,GA and PSO;In Chapter 3: the prediction model of regional human capital based on GA-PSO optimizing BP neural network;In Chapter 4: Simulation Analysis for the level of regional human capital based on prediction model.The results of the simulation and the example validations indicate that the prediction model of regional human capital based on GA-PSO optimizing BP neural network is feasibility and superiority.