A hybrid FCM-CNN method to cluster and forecast financial performance of listed companies

Abstract

Most methods for evaluating a company’s financial performance currently focus on scoring, when there is a large amount of data, it is difficult to distinguish the company’s financial status. To cluster and predict the financial performance of companies, a hybrid model based on the fuzzy C-means clustering algorithm (FCM) and convolutional neural network (CNN) is proposed in this paper. Pearson correlation analysis was first performed on the indicators to ensure that they are not correlated with each other and to avoid indicator redundancy. The entropy method determined the weight of each index and ensured the high validity of the selected indicators. Then, FCM clustering was carried out, and the performance of each company was clustered according to the indexes after data preprocessing with clustering labels. The processed data and labels were introduced into CNN to predict the level. The empirical study showed that the FCM-CNN model was superior to other machine learning models, which proved that this model has better clustering and forecasting ability, and could be applied to the prediction of corporate financial performance.