Decision Tree and Naïve Bayes for Sentiment Analysis in Smoking Perception

The number of smokers is still significant in many countries. The trend of smokers has seemed to keep changing in time, especially in Malaysia. However, a smoke-free nation is one of the aims by 2045. Thus, there is a need to see the trend of smoking as it can be one the assistances to the relevant agencies to track the smoking trend. Perceiving the sentiment on smoking is now possible instead of only utilizing the traditional approaches like surveys, interviews, and questionnaires are popular methods for identifying current smoking trends. These approaches acquire time and cost, and fieldwork. In parallel with digitization and IR 4.0, the trend analysis methods are established in many forms, such as Twitter, web site, and Facebook. This paper's focal point is to see the utilization of the Decision Tree and Naive Bayes to perform sentiment analysis on smoking perceptions in three states in Malaysia. The methods are employed to classify the about 4500 tweets into positive, negative and neutral based on one-word and two-word search terms. The decision tree performs better when applied to the one-word search term data set, while Naïve Bayes is better for the two-word tweets search term data set. The finding demonstrates that this work is beneficial to obtain a quick and efficient result of the current smoker's perception, which will help the authorities develop new solutions to support the smoke-free generation initiative.