Deep Feature Weighting Based on Genetic Algorithm and Naïve Bayes for Twitter Sentiment Analysis
Reiza Adi Cahya, Dinda Adimanggala, Ahmad Afif Supianto · 2019
In recent years, users' have expressed their opinion about certain product or service, especially in microblogging sites like Twitter. The analysis of opinion or sentiment is important for organizations to identify correct measures to improve their products or services. To analyze sentiments, the Naive Bayes (NB) model is one of the more popular methods due to low computational time and understandability. However, NB assumes that each feature does not have dependencies with others, which may not be fulfilled in text analysis. To reduce the effect of this assumption, Genetic Algorithm (GA) is used as a feature weighting process. Features that are highly correlated with the class label and lowly correlated with other features are given more weights. Experiments are conducted on the Twitter Airline database to validate the proposed model. Result shows that average accuracy of weighted NB model is 73.72%, compared to standard NB model's 73.55%. It is concluded that feature weighted NB has minor advantage over standard NB model.