Spam Message Filtering using Naive Bayes Classifier

Overview

This repository contains code for building a spam message filter using the Multinomial Naive Bayes classifier. The project involves natural language processing (NLP) techniques, including text preprocessing, TF-IDF vectorization, and the application of the machine learning model.

Tools and Libraries

• Python
• Pandas
• NumPy
• Seaborn
• Matplotlib
• NLTK
• WordCloud
• Scikit-learn (Sklearn)

Dataset

The dataset used for this project is sourced from the 'spam.csv' file, containing labeled spam and ham (non-spam) messages.

Workflow

Data Cleaning and Exploration

• Removal of unnecessary columns and renaming.
• Exploratory Data Analysis (EDA) to understand message lengths, distribution, and visualization.

Text Processing

• Removal of punctuation and stopwords.
• Tokenization and lemmatization of words.
• Word Cloud visualization for both spam and ham messages.

Vectorization

• Bag-of-Words (BoW) vectorization using CountVectorizer.
• TF-IDF (Term Frequency-Inverse Document Frequency) vectorization.

Machine Learning Model

• Train-test split of the TF-IDF transformed data.
• Implementation of Multinomial Naive Bayes classifier.
• Evaluation of the model using classification reports on training and testing sets.

Results

• The model demonstrates high precision, recall, and F1-scores for 'ham' (non-spam) messages in both training and testing sets.
• While the model performs well on 'ham' messages, there is room for improvement in accurately classifying 'spam' messages.