Abstract: The quantitative research and analysis about the classification algorithm are not often sufficient to choose the suitable algorithm. In this paper, the K-Nearest Neighbor algorithm, SVM and decision tree are quantitatively analyzed about classification accuracy and running time by using different the parameters of the algorithm, the data noise and the number of nodes. Firstly, parameters effects of these algorithms are studied. Then, the optimal parameters are selected to analyze the influence of different noise on the classification accuracy. At last, the influence of the number of nodes on the classification accuracy and the running time is analyzed. The Scikit-learn module is used to simulate the content of the discussion. The experimental results clearly show the classification characteristics of these classification algorithms under different parameters, which provide guidance for the classification of the actual data.

Key words: parameter selection, classification feature, KNN, SVM, decision tree