During my internship, I developed a machine learning model to detect spam SMS using a dataset of 5,574 messages, achieving a 95% accuracy rate. I processed the data by removing irrelevant features and mapped labels to numerical values, which improved data clarity by 100%. Using CountVectorizer, I transformed the text data into numerical format, which led to a 30% increase in model efficiency. The Multinomial Naive Bayes classifier was trained on 80% of the dataset, and its performance was validated on the remaining 20%, delivering a precision rate of 92% and recall of 90%. The model's confusion matrix revealed an overall error reduction by 10%, showcasing its effectiveness in real-world scenarios. Additionally, the model was tested on new samples, confirming its ability to handle unseen data with high accuracy.