Feature Extraction and Classification of Text Data by Combining Two-stage Feature Selection Algorithm and Improved Machine Learning Algorithm

Abstract

Efficient text classification is crucial for information processing due to the generation of massive text data. However, the uneven distribution and redundancy of text data often result in poor classification performance. To address this issue, a two-stage feature selection algorithm is proposed using the fusion of information gain and maximum correlation minimum redundancy algorithm. To improve SVM performance in text data classification, an improved SVM algorithm based on Fourier hybrid kernel function is proposed. The study found that the proposed improved algorithm achieved an accuracy of 0.82 on the IMDB dataset using only 40 feature subsets. Even when the number of features exceeded 390, the F1 value of the proposed algorithm remained 1% to 2% higher than that of other algorithms. The improved algorithm performed best when the feature dimension was around 400. The proposed algorithm, which combines the Fourier hybrid kernel function with a two-stage feature selection algorithm based on the information gain and maximum correlation minimum redundancy algorithm, achieved a 1%~3% higher F1 value and increased the number of correctly classified texts by 20 to 45. These results demonstrate the effectiveness of the algorithm as a classification tool for processing large-scale text data, which is significant for information retrieval and data mining.Povzetek: Predstavljena sta dvodstopenjski algoritem za izbiro značilk in izboljšani algoritem strojnega učenja za povečanje točnosti klasifikacije besedilnih podatkov. Združujeta informacijski dobiček in metodo minimalne redundance ter maksimalne korelacije (MRMR) z izboljšano SVM.