The research on determinants of capital structure based on fractal dimensionality reduction

This paper initially defines the influencing factors of small and medium-sized enterprises (SMEs)´ capital structure on the basis of the theoretical analysis and existing research findings. In order to remove those redundant attributes which are unimportant or related to each other, fractal dimensionality reduction is used for the attribute reduction on data sets collected from Guo Tai An unlisted companies database, and it finally generates 8 determinants comprising of total assets, age of firms and so forth. Furthermore, based on over 300 data records of empirical investigation, this paper utilizes the decision tree induction to classify and predict the firm´s financing situation, and approximately 85% accuracy implies its rationality. Relying on the machine learning algorithms, this paper analyzes the firm´s capital structure and makes efficient use of the statistic information among numerous data, which provides contributions to research in such field.