Research On Improved Affinity Propagation Clustering Algorithms And Its Applications For Evaluation Of Listed Companies Performance

Clustering is an important research direction in the field of data mining. Affinitypropagation algorithm, as a kind of unsupervised clustering method, has manyadvantages such as high productivity, stabilization and that do not have the pre-fixednumber of clusters, etc. In the view of affinity propagation algorithm has manyadvantages, some theoretical study and applications study based on affinitypropagation algorithm are made in the paper. The major contents could besummarized as follows:(1) In order to overcome the effect of noisy and less important attributes, thecoefficient of variation is introduced into the affinity propagation algorithm, and theoriginal similarity calculation formula is improved. An improved affinity propagationalgorithm with coefficient of variation (CVAP) is proposed in this paper. In order toverify the validity of the improved algorithm, the CVAP algorithm is applied in thefield of listed companies’ evaluation in this paper, the experimental results show thatthe improved algorithm has better clustering performance than the original algorithm.(2) In order to further study on feature weighting, fruit fly optimization algorithmis adopted to optimize the weights of feature in this paper. So an improved AffinityPropagation algorithm is proposed in this paper, namely FOAP (Fruit FlyOptimization Affinity Propagation Algorithm). Weighted data can better reflect thedata indicators in the evaluation system of real action and the similarity between datasamples. The experimental results show that the method of performance evaluation oflisted companies is scientific and effective.(3) The traditional performance evaluation models of listed companies selectfinancial data which is usually at some point. It does not consider different time spaninformation which is contained in the data. In order to solve this problem, this articleintegrates multivariable panel data and the affinity propagation clustering algorithm toevaluate the performance of listed companies. In addition, in order to fully extract theinformation of listed companies’ financial data, this paper adopts a semi-supervisedlearning strategy. And a semi-supervised affinity propagation algorithm which fusesmultiple panel data is proposed, namely SMAP (Semi-supervised AffinityPropagation Algorithm Fuses Multiple Panel Data, SMAP). The experimental resultsshow that the results of this model are more accord with actual situation. It can provide effective decision-making basis for the government and financial regulators totimely and accurately identify problems and formulate economic policies andregulatory measures.