📩 SMS Spam Classifier

A machine learning-based web application that classifies SMS messages as either "Spam" or "Ham" (legitimate) with high precision. The application is built using Python and deployed via Streamlit, utilizing a Multinomial Naive Bayes classifier trained on the SMS Spam Collection dataset.

🚀 Project Overview

Unwanted spam messages are a nuisance and a potential security threat. This project aims to filter these messages out by analyzing their text content. The system uses Natural Language Processing (NLP) techniques to preprocess text and a machine learning model to predict the category of the message.

• Input: Raw SMS text.
• Output: Classification (Spam/Not Spam).
• Key Metric: The model was selected specifically for its 100% Precision score, ensuring that legitimate messages are not accidentally classified as spam.

🛠️ Tech Stack

• Language: Python 3.x
• Web Framework: Streamlit
• Machine Learning: Scikit-learn (MultinomialNB, TF-IDF)
• Natural Language Processing: NLTK (PorterStemmer, Stopwords, Tokenization)
• Data Manipulation: Pandas, NumPy
• Visualization: Matplotlib, Seaborn, WordCloud

📂 Project Structure

• SMS-Spam-detection.ipynb: Jupyter Notebook for EDA, preprocessing, and model training.
• app1.py: Main Streamlit application script.
• spam.csv: Dataset used for training.
• vectorizer.pkl: Saved TF-IDF Vectorizer (generated by notebook).
• model.pkl: Saved Naive Bayes Model (generated by notebook).
• requirements.txt: List of project dependencies.
• README.md: Project documentation.

⚙️ How It Works

1. Model Training (SMS-Spam-detection.ipynb)

The notebook covers the entire data science pipeline:

• Data Cleaning: Dropping empty columns, renaming features, and handling duplicates.
• Exploratory Data Analysis (EDA): Analyzing message lengths, class distribution (imbalanced), and visualizing frequent words using WordClouds.
• Preprocessing:
  • Lowercasing text.
  • Tokenization (breaking text into words).
  • Removing special characters, punctuation, and stopwords.
  • Stemming (reducing words to their root form).
• Vectorization: Using TfidfVectorizer (max features=3000) to convert text into numerical data.
• Model Selection: After testing multiple algorithms (Logistic Regression, SVM, Random Forest, etc.), Multinomial Naive Bayes was chosen for offering the best balance of accuracy (~97%) and precision (1.0).

2. The Application (app1.py)

The Streamlit app provides a user interface for the model:

1. Loads the pre-trained vectorizer.pkl and model.pkl files.
2. Accepts user input via a text area.
3. Preprocesses the input using the same pipeline defined in training.
4. Displays the prediction ("Spam" or "Not Spam").

🔧 Installation & Run

1. Clone the repository:

   git clone [https://github.com/yourusername/sms-spam-classifier.git](https://github.com/yourusername/sms-spam-classifier.git)
   cd sms-spam-classifier

2. Create a virtual environment (Recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

   Note: Ensure you download the necessary NLTK data:

   import nltk
   nltk.download('punkt')
   nltk.download('stopwords')

4. Run the App:

   streamlit run app1.py

📦 Requirements

streamlit
pandas
numpy
scikit-learn
nltk
matplotlib
seaborn
wordcloud