📩 SMS Spam Filter using Naive Bayes

This repository contains an R implementation of a Naive Bayes classifier for detecting SMS spam messages. The classifier uses text preprocessing techniques and a Document-Term Matrix (DTM) to train and evaluate the model. The dataset used is the publicly available SMS Spam Collection dataset from the UCI Machine Learning Repository.

Table of Contents

1. Project Overview
2. Dataset
3. Installation and Requirements
4. Usage
5. Code Explanation
6. Results

Project Overview

The goal of this project is to create a machine learning model that classifies SMS messages as either "ham" (non-spam) or "spam." The implementation uses the Naive Bayes algorithm, which is well-suited for text classification tasks.

Dataset

The dataset used for this project is the SMS Spam Collection, which can be downloaded from the UCI Machine Learning Repository. It contains 5,574 labeled SMS messages with the following columns:

• Label: ham for non-spam messages and spam for spam messages.
• Message: The SMS text message.

Save the dataset as data/sms_spam.txt in the repository folder.

Installation and Requirements

Prerequisites

• R (version 4.0 or later)
• RStudio (optional but recommended)

Required Libraries

Install the following R libraries:

install.packages(c("dplyr", "readr",  "caret", "e1071", "tm"))

Usage

1. Clone this repository:

   git clone https://github.com/alexanderk001/sms-spam-filter.git
   cd sms-spam-filter

2. Place the dataset in the data/ directory as sms_spam.txt.

3. Open the R script in RStudio and run it step by step.

4. Results, including accuracy and confusion matrix, will be printed in the console.

Code Explanation

1. Loading Data

The dataset is loaded using the readr package, and the labels (ham/spam) are converted into factors for modeling.

2. Text Preprocessing

The text data is cleaned using the following steps:

• Convert to lowercase.
• Remove punctuation and numbers.
• Remove stop words.
• Strip extra whitespace.

3. Document-Term Matrix (DTM)

The cleaned messages are converted into a DTM, which represents the frequency of terms in each document (message). Sparsity is reduced by keeping only frequently used terms.

4. Binary Conversion

The DTM is converted to binary values:

• 1 if a term is present.
• 0 if a term is absent.

5. Training and Testing

The dataset is split into training (80%) and testing (20%) sets. A Naive Bayes classifier is trained on the training set, and predictions are made on the test set.

6. Evaluation

The performance of the classifier is evaluated using:

• Confusion Matrix: Shows the number of true positives, true negatives, false positives, and false negatives.
• Accuracy: Proportion of correctly classified messages.

Results

Example output:

Confusion Matrix and Statistics
          Reference
Prediction ham spam
      ham  963   15
      spam   2  134

Accuracy : 98.47%

Key Metrics:

• Accuracy: 98.47%
• Sensitivity: 99.79%
• Specificity: 89.93%