A Comprehensive Approach to SMS Spam Filtering Integrating Embedded and Statistical Features

Modern society relies heavily on mobile phones to communicate. One of the most valuable mobile phone services is SMS (Short Message Service), which simplifies communication greatly. There have been spammers who have misused this platform by sending inappropriate messages to users, provoking them and costing them money. Due to imbalanced data, unclear semantics, and the inability to extract sufficient features from short messages, SMS spam can be difficult to filter. While spam messages have been filtered so far using various methods, their accuracy is still a work in progress. This study uses embeddings and TF-IDF to provide more information from short text messages while improving SMS spam filtering accuracy. The proposed approach was tested on a real dataset. Experiments analyzing evaluation parameters demonstrate that this model is effective.