Spam detection using machine learning

In the modern digital age, where the quantity of spam emails has skyrocketed, classifying emails as spam is an essential responsibility. The study's aim is to identify an effective spam classifier that can distinguish and separate spam emails, SMS, and YouTube from authentic ones with precision. Three distinct datasets—YouTube, SMS, and Gmail—make up the dataset used for this study. It chooses the optimal algorithm based on optimal detection accuracy and precision. Tokenization, stop word removal, stemming, and feature extraction are a few NLP techniques we employed to preprocess the text input and extract pertinent features. To identify the optimal model for spam categorization, we will assess a number of machine learning (ML) methods, including Naive Bayes, Support Vector Machines (SVMs), K-Nearest Neighbors (KNN), Logistic Regression, and Random Forests. Evaluation criteria, including precision, recall, and F1-score, will be used to gauge the classifier's accuracy. Our results show that the Random Forest model works better than other models in identifying spam on YouTube and SMS, with the greatest accuracy scores of 95% and 97%, respectively, while Logistic Regression, Naive Bayes, and SVM perform better with the highest accuracy scores of 98% in identifying spam on emails. The research will also enhance the development of NLP and ML methods for the classification of spam.