Research And Application Of Adaptive Spectral Clustering Algorithm Based On Arificial Immune

Social insurance audit is beneficial to the national economy and the people’s livelihood. The effective audit measures can found whether the social insurance is illegal in timely. And it can safeguard the legitimate rights of workers. The current audit means mainly have artificial audit and computer audit. The main audit method of artificial audit is artificial sampling audit, the artificial audit is lag and time consuming, while sampling audit exists blindness and lugs. Computer audit mainly uses audit system, although the audit efficient is improved to some extent, while the accuracy is low. In face of the massive social security audit data, the traditional manual audit method is not enough.Data mining can find useful knowledge from massive data, so the researchers began to use data mining methods to analysis audit data. There are many methods in data mining; clustering analysis has strong scalability, weak dependence on domain knowledge, less affected by noise and other advantages. So this paper uses clustering to analyze social security audit data.Although traditional clustering algorithm is easy to achieve and to use, it is easy to fall into local optimal solution, having a poor effect in non convex space. Spectral clustering algorithm is based on spectral graph theory, it can map data to spectral, and it has better robustness. So this paper uses the spectral clustering algorithm to solve the problem of traditional clustering algorithm.This paper analyzes the spectral clustering algorithm, then proposes that although the traditional clustering algorithm can solve some problems, but it still needs to input cluster number K value manually. The determination of the clustering number is crucial, the optimization of the traditional is urgent. In addition to the optimization need of the traditional spectral clustering, this paper researches the artificial immune system deeply. This paper proposes an adaptive spectral clustering algorithm based on artificial immune system. The algorithm simulates the antibody’s clone and variation. The antibody recognizes antigen through the first immune response and second immune response. This paper uses some data to test behind realizing the algorithm. And then this paper compared the result with traditional spectral clustering and genetic algorithm. The result verifies the feasibility and stability of the algorithm.This paper finds that it still needs make further improvement after analyzing the characteristic of the social security audit data. So this paper makes further improvement based on artificial immune spectral clustering algorithm. According to the contribution of the data attribute is different, we join expert knowledge and propose an adaptive spectral clustering algorithm based on semi supervise. This paper does some pretreatment on social security audit data, includes attribute selection, data consolidation, data filling, data classification and so on. Then this paper analyzes the social security audit data, gets the results of whether the index is illegal. At last this paper compares the results with the results that expert tagged and get the accuracy. By the contrasts with the other two algorithms this paper gets that the algorithm has a higher stability. Finally, the paper sums up some rules and the results are basically consistent with the policy, further validation the advantage of the algorithm.