Feature selection methods for mining social media

Abstract

People can share their thoughts and opinion through Social Media which can easily widespread. So many public issues and political views are also discussed on social media. HIV/AIDS is also one of the important topics discussed. This work aims to classify HIV/AIDS related twitter data. Since the twitter data is highly dimensional, it is essential to do reduce dimensionality of the data to attain better classification results. Tweets are collected using keyword search and necessary pre-processing steps are carried out. Then feature extraction methods such as Bag of Words (BOW) model and TF-IDF are implemented. Singular Value Decomposition (SVD) and Principal Component Analysis (PCA) techniques are used for dimensionality reduction. Finally, classification is carried out and the results are discussed.