Hybrid Model For Text Classification

Abstract

Increasing popularity of social media sites (e.g., Twitter, Facebook) leads to the increasing amount of online data. Student's casual talks on these social media sites can be used to know their educational experiences i.e. their worries, opinions, feelings, and emotions about the learning process. However, the main challenge is to analyze this informal data. Because of the complexity of the student's posts on the social media sites, it requires human understanding. But, it is practically not possible to perform human interpretation on such increasing scale of data. Hence, there is a demand for pre-programmed data analysis methods. In this paper, a hybrid approach is used to analyze these large scale and informal data. We have considered engineering student's Twitter posts to know their concerns and difficulties in their learning experiences. We have carried out a qualitative analysis on the data of about 25,000 tweets of engineering student's educational problems and studied that many students face problems such as heavy study load, diversity problem, lack of social engagement, their negative emotions and lack of sleep. Considering these analysis, we have implemented a hybrid model by combining two most popular machine learning algorithms viz. Naïve Bayes classifier and Support Vector Machine to classify tweets regarding student's problems. The result shows that there is a significant reduction in training time and shows an increase in the classification accuracy when compared to individual Naïve Bayes classifier and Support Vector Machine.