Research and Improvement of CHI Feature Selection in Sentiment Analysis
Abstract
Feature selection is a very important step in sentiment classification based on machine learning method. This paper will focus on the better CHI feature selection method and make improvements to overcome the shortcomings of the traditional method. The experimental results show that the improved IM-CHI feature selection method improves the F1 value of emotion classification by 4.9% in different feature dimensions, 4.1% in different classifiers, and 2.0% in data sets of different fields. It proves the effectiveness of this method in emotion classification.
- Publication:
-
Journal of Physics Conference Series
- Pub Date:
- April 2019
- DOI:
- 10.1088/1742-6596/1187/5/052022
- Bibcode:
- 2019JPhCS1187e2022D