Influence of Feature Selection Techniques for Social Media Data Analysis (Text and Image)
Yogesh Gupta · 2025
Social media is an application where a significant amount of data has been contributed by the users. Therefore, social media has been used in various data driven application development. The data available in social media can be images, text or videos. Researchers have developed many approaches to process such data to get insights and to make accurate decisions. Feature selection plays an important role in this analysis. Therefore, this paper develops various feature selection techniques including TF-IDF, Bi-gram, POS tagging for text data and Sobel operator, LBP for images. All the experiments are performed on two datasets obtained from Kaggle for social media-based images and text. Further, the extracted features are classified and used in artificial neural network for training and validations. Based on the experimental results, it is proved that TF-IDF based features are superior to other implemented feature selection techniques for classifying the social media text. And the combination of Sobel, LBP and Color obtains better accuracy for image-based classification.