User Location Prediction System on Twitter Using Machine Learning Technique

This paper presents a machine learning-based improve accuracy through efficient preprocessing, enhance usability via an interactive interface, and provide a scalable framework suitable for safety and analytics applications. framework for predicting the geographic location of Twitter users by analyzing their tweet content. Social media generates vast amounts of textual data, but the lack of explicit location information creates challenges for applications such as personalized services, emergency response, and safety monitoring. To address this gap, the proposed system integrates natural language processing techniques with supervised learning models. The workflow involves preprocessing tweets through cleaning, tokenization, and normalization, followed by feature extraction using the Term Frequency–Inverse Document Frequency (TF-IDF) method. A Multinomial Naïve Bayes classifier is then applied to categorize tweets according to their most probable location. A custom dataset of region-specific tweets was used to train and evaluate the model. The system further incorporates a user- friendly interface developed with Streamlit, enabling login, tweet input, and real-time prediction of both location and username. Experimental results indicate that the approach provides reliable classification accuracy, demonstrating its potential in enhancing applications related to security, personalized recommendations, and location-based analytics.