Research And Application Of Mining Algorithm Based On Social Security Data

With the development of social security informationization,more and more data have been accumulated in social security management platforms.From the social security data,useful potential laws and patterns are excavated,social security risks are forecasted,and policies that are conducive to the social security are increasingly valued by the government.Because of the development of data mining technology,the combination of data mining technology and social security data can exactly meet the urgent needs of the Humanities and Social Affairs Bureau.To this end,the Humanities and Social Affairs Department of Jiangmen City has set up the Human Society Big Data Project to combine the technology and capabilities of schools and enterprises to explore the value of social security.This article based on research work and some results of the project.We first study the Verhulst algorithm,the BP neural network algorithm,the particle swarm optimization algorithm and the random forest classification algorithm systematically,and compare the advantages and disadvantages of each algorithm.Based on this,a particle swarm optimization grey Verhulst-BP neural network combination forecasting algorithm is proposed.The algorithm is applied in the pension insurance fund forecast.Through the modeling experiments of Decision-Tree classification algorithm,BP neural network classification algorithm and random forest classification algorithm,an algorithm model for unbalanced data classification and prediction is established,and the model is applied to the instability analysis and prediction of the enterprise.The main research results we have obtained are:1.A gray Verhulst-BP neural network combination forecasting algorithm based on particle swarm optimization is proposed and compared with the grey Verhulst algorithm.Experiments show that the algorithm is effective.2.Applying the improved algorithm to the forecast of pension fund,this algorithm has better robustness and higher prediction accuracy than BP neural network algorithm,gray Verhulst algorithm and grey Verhulst-BP neural network algorithm.3.Compareing the classification effects of Decision Tree Classification algorithm,BP neural network classification algorithm and random forest classification algorithm in non-equilibrium data.Analyzing the evaluation result,and it was confirmed that the random forest classification algorithm has better performance in non-equilibrium data classification and prediction.4.A random forest classification algorithm was used to construct an enterprise instability analysis and prediction model.Experiments show that the random forest classification algorithm has a good effect in the identification of unstable enterprises.5.We have developed a corporate instability analysis and forecasting platform and displayed the forecast results on the platform.