Analysis of Feature Selection Methods to Improve the Accuracy of SVM Algorithm in Classifying Complex Data

This research aims to improve the accuracy of the Support Vector Machine (SVM) algorithm in classifying primary tumor data through feature selection.Primary tumors, which can be benign or malignant, require accurate diagnosis for effective medical treatment.By utilizing primary tumor datasets from UCI Machine Learning, this research applies three feature selection methods: Gain Ratio, Chi-Square, and ReliefF.After going through the preprocessing and feature selection stages, SVM is used to classify the data.The results show that without feature selection, the SVM accuracy is 74.3363%.Applying Gain Ratio and Chi-Square feature selection increases accuracy to 75.5162%, while the ReliefF method provides the highest increase to 75.8112%.Thus, the feature selection method is proven to be effective in improving SVM accuracy, with ReliefF as the most superior method.This research emphasizes the importance of feature selection in medical data processing to increase the accuracy of classification results.