Sentiment Analysis of Twitter Data on Procrastination Using Decision Tree Navie Base and Deep Learning

Abstract

Today's social networking sites are the exploration, so large amount of data is generated. Millions of people are sharing their veiws daily on microblogs such as Facebook , Twitter and Instagram. In this paper, explore Sentiment analysis to identify the polarity (Positive, Negative and Neutral) along with emotion for procrastination using machine learning and deep learning algorithms. By analysing related research papers and articles to identify related factors and attributes and suggest ‘how procrastination is cause of stress or not”. The approach used to solve the above problem includes preprocessing and analysis of tweets using data mining tools such as RapidMiner with plug-in Aylien and Indico API in Python language using twitter data.  The algorithm used at the backend of emotion API is LSTM (Long Short Term Memory). The proposed models were evaluated on two different tasks. First, a decision tree classifier was used to predict polarity on datasets of 300 and 500 samples, achieving accuracies of 72.22% and 68.67%, respectively. In contrast, a Naive Bayes classifier performed significantly worse on the same polarity task, yielding accuracies of 33.33% on the 300-sample dataset and 26.67% on the 500-sample dataset. Secondly, the models were assessed on their ability to distinguish between subjective and objective parts of tweets. For this task, the decision tree classifier achieved a precision of 86.67% when trained on 500 samples.

Keywords: Sentiment Analysis, Decision Tree, Naive Base, Machine Learning, Deep Learning