An Automatic Spam Detection System Based on Hybrid Scoring Models
Stefka Popova, Hristo Nenov, Velislav Kolesnichenko · 2024
Volume of spam everyone receives has significantly increased in the last years. Recognizing and clearing the spam manually by receiver become sometimes a challenging and time-consuming task. To solve this task automatic spam detection systems were developed using advanced machine learning algorithms. Such systems analyze each email at the time of its arrival thus preventing potential harm. The paper proposes an automatic system for spam detection based on three pre-trained models using Logistic Regression, Support Vector Machines and Random Forest algorithms. Each model estimates the email by content and header analysis. The final email classification is made by arbitrage of models’ decisions. The estimation is triggered in periods of few seconds processing different number of emails each time. Proposed system works for emails both in English and Bulgarian, and achieved very good accuracy of over 99.5% reducing classification error to less than 0.5%.