Application Of Improved Affinity Propagation Clustering Algorithm In Financial Distress Prediction

With the accelerating pace of economic globalization, Chinese enterprises are facing more and more development opportunities and challenges. In order to predict the financial crisis, it is important for enterprises to get helpful information from a large number of data by using the effective financial distress prediction model. It is becoming more and more difficult to get the helpful information from the explosion of data in current time. Data Mining is the technology which can change the awkward situation. And the clustering analysis is one of the important methods of data mining. Clustering analysis has achieved good results in financial distress prediction area in recent years.This dissertation improves the affinity propagation clustering algorithm to overcome three defects. First, the improved affinity propagation clustering algorithm could deal with non-spherical data sets. Second, the number of clustering classes of improved affinity propagation clustering algorithm is more closed to the real number. Third, it reduces the sensitivity of the parameter P. And it has gotten a good result on UCI data sets.This dissertation combines the improved affinity propagation clustering algorithm with the fuzzy pattern recognition theory to build a new financial distress prediction model for T-2 period. This new model has been test on financial data of 30 normal and 30 abnormal listed companies. The model improved the predicting accuracy of T-2 period.