The Choice of Kernel Function for One-Class Support Vector Machine

Support vector machines (SVM) are mainly used in classification problem, like character recognition, face recognition, pedestrian detection, text classification and other fields. Isolation forest can be used to anomaly detection (outlier detection), which requires relatively small samples from large data sets. SVM has several kernel functions including RBF, linear, polynomial and radial. In this paper, the experiments on different real-world data sets are performed to demonstrate the importance of kernel choice and the corresponding sensitivity of the algorithm. This paper mainly explores the performance of various main kernel functions in different data sets in OCSVM, and compares the final effect of OCSVM and isolation forest in processing the same data set. Furthermore, this paper also compares the OCSVM method using RBF kernel with adaptive boosting.