TWO-STAGE CLASSIFICATION MODEL FOR SPAM EMAIL DETECTION

The rising usage of the Internet and social networks, online communication has become an essential part of our daily life. One of the widely used mediums for formal and business communication is e-mail, because of its open access, quickness and reliability. Email increase popularity, spam emails were rapidly rising, spam is one or more unsolicited messages that take the form of advertising or promotional materials like debt reduction plans, getting rich quick schemes, online dating, and health-related products, etc. Automatic spam detection is no new topic, business and companies are always looking to improve their user’s experience in order to protect their machines from potential damage; in case viruses were wrapped inside spam emails. The email spam definition is ambiguous since everybody has their views on it. At present, email spam is getting the attention of everyone. This study was conducted to design and implement a hybrid email spam detection system that combines the advantages of two-stage classification, machine learning methods, and dimensionality reduction approaches (Logistics regression and FNN) to address this issue of effective identifying and filtering out spam emails while reducing false positives. The Proposed system uses logistic regression as an effective first-stage classifier to rapidly filter out the bulk of non-spam emails while a neural network model was used as the second-stage classifier to accurately classify the remaining emails after using Principal component analysis (PCA) to do dimensionality reduction on the feature space of the remaining emails following the first-stage classification to improve performance and lower computing complexity. However, the study seeks to develop a spam detection system that can accurately identify and filter out spam emails while maintaining high efficiency and scalability with a primary objective to analyze the effectiveness of these algorithms in identifying and filtering spam emails. The results of this study demonstrated that both algorithms could successfully classify emails as spam or non-spam with notable accuracy. Logistic regression and FNN algorithms were evaluated based on their performance metrics, including precision, recall, and overall accuracy.