https://github.com/rohitdusane/spam-classification-model
A Python-based machine learning model for spam detection leveraging TF-IDF vectorization and multiple classifiers, including Naive Bayes, Logistic Regression, and Random Forest. This project demonstrates preprocessing techniques, model training, and performance evaluation for classifying SMS messages as spam or ham.
https://github.com/rohitdusane/spam-classification-model
data-science flask mlflow natural-language-processing scikit-learn spam-detection text-classification
Last synced: 2 months ago
JSON representation
A Python-based machine learning model for spam detection leveraging TF-IDF vectorization and multiple classifiers, including Naive Bayes, Logistic Regression, and Random Forest. This project demonstrates preprocessing techniques, model training, and performance evaluation for classifying SMS messages as spam or ham.
- Host: GitHub
- URL: https://github.com/rohitdusane/spam-classification-model
- Owner: RohitDusane
- Created: 2025-08-21T05:13:26.000Z (10 months ago)
- Default Branch: main
- Last Pushed: 2025-08-21T09:21:41.000Z (10 months ago)
- Last Synced: 2025-08-21T10:34:55.060Z (10 months ago)
- Topics: data-science, flask, mlflow, natural-language-processing, scikit-learn, spam-detection, text-classification
- Language: Jupyter Notebook
- Homepage: https://spam-ham-classification-ml-project.onrender.com
- Size: 12.5 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Spam Classification Model with Multiple Machine Learning Algorithms
## GOAL
The goal of this project is to develop a **spam classification model** that can classify SMS messages as either **spam** or **ham** (not spam). The project uses multiple machine learning models like **Logistic Regression**, **Random Forest**, **Decision Tree**, **XGBoost**, and **Multinomial Naive Bayes** to obtain the final model, which is then deployed using a **Flask app** on **Render** for real-time predictions.
## PURPOSE
The purpose of this project is to explore different machine learning models and their performance in spam classification tasks, then deploy the final model in a Flask web application so that it can classify messages dynamically via API requests.
## DATASET
The dataset used in this project is the **SMS Spam Collection Dataset**, available at:
[SMS Spam Collection Dataset on Kaggle](https://www.kaggle.com/datasets/uciml/sms-spam-collection-dataset)
This dataset contains labeled SMS messages with two classes: **spam** and **ham**.
## DESCRIPTION
This project uses **TF-IDF (Term Frequency-Inverse Document Frequency)** vectorization to convert SMS text messages into numerical features. We then apply multiple machine learning models (Logistic Regression, Random Forest, Decision Tree, XGBoost, Multinomial Naive Bayes) and compare their results. The model with highest **F1-Score** metrics is used (**Multinomial Naive Bayes**) to create a `final predictive model`. The model is then deployed in a Flask app, making it available for real-time predictions.
## WHAT I HAD DONE
- Acquired and loaded the dataset.
- Cleaned and preprocessed the text data (lowercasing, removing stop words, stemming).
- Applied **TF-IDF** vectorization to transform text into numerical features.
- Implemented multiple machine learning algorithms (Logistic Regression, Random Forest, Decision Tree, XGBoost, Multinomial Naive Bayes).
- Compared the models using **F1-Scorring** and choose the highest scored model to create the final predictive model.
- Developed a **Flask API** to serve the model and deployed it on **Render**.
## WORKFLOW OF YOUR PROJECT FILES
1. **Data Acquisition**: Loaded the SMS dataset and displayed the first few rows.
2. **Data Preprocessing**: Cleaned and tokenized the text data.
3. **TF-IDF Vectorization**: Converted the cleaned text data into numerical features using TF-IDF.
4. **Model Training**: Trained multiple models (Logistic Regression, Random Forest, Decision Tree, XGBoost, Naive Bayes).
5. **Model Evaluation**: Evaluated the models using **accuracy**, **precision**, **recall**, and **F1-score**.
6. **Flask App**: Developed a Flask web application to make predictions.
7. **Deployment**: Deployed the Flask app on **Render** for real-time usage.
## DIAGRAMS
- **Data Preprocessing Flow**:
- **WORD CLOUD**:
## STATE YOUR PROCEDURE AND UNDERSTANDING FROM YOUR WORK
1. **Preprocessing**: The text data was cleaned by removing unwanted characters, converting all text to lowercase, and stemming using **SnowballStemmer**. We also removed **stopwords** to enhance feature quality.
2. **TF-IDF Vectorization**: I chose **TF-IDF** for feature extraction because it captures the importance of words within documents in relation to the entire corpus, making it a strong technique for text classification tasks.
3. **Model Selection**: I experimented with multiple classifiers and found that using a **Multinomial NB** increased prediction f1-score (highest).
4. **Deployment**: I used **Flask** to wrap the final model into an API and deployed it on **Render** for real-time access.
### What I learned:
- Different models can have varying levels of performance on the same dataset.
## DETAILED EXPLANATION OF SCRIPT, IF APPLICABLE
1. **Data Preprocessing**:
- Text is tokenized, stopwords are removed, and stemming is applied using **NLTK**.
2. **TF-IDF Vectorization**:
- **TfidfVectorizer** from **Scikit-learn** is used to convert text into numerical features.
3. **Model Training**:
- Trained the following models:
- **Logistic Regression**
- **Random Forest**
- **Decision Tree**
- **XGBoost**
- **Multinomial Naive Bayes**
4. **Model Evaluation**:
- Evaluated models using **classification_report** for accuracy, precision, recall, and F1-score.
5. **Model Performance Comparison**
The following table summarizes the performance metrics of different machine learning models used in the spam classification project, along with the best hyperparameters for each model:
| **Model** | **Train Accuracy** | **Test Accuracy** | **Precision (SPAM)** | **Recall (SPAM)** | **F1-Score (SPAM)** | **Best Params** |
|---------------------------------------|--------------------|-------------------|----------------------|-------------------|---------------------|---------------------------------------------------------------------------------|
| **Multinomial Naive Bayes (MNB)** | 0.990 | 0.988 | 0.984 | 0.924 | 0.953 | `{'alpha': 0.1}` |
| **Logistic Regression (LRC)** | 0.999 | 0.986 | 0.976 | 0.917 | 0.945 | `{'C': 0.1, 'max_iter': 100, 'penalty': None, 'solver': 'saga'}` |
| **XGBoost (XGB)** | 0.990 | 0.968 | 0.916 | 0.826 | 0.869 | `{'learning_rate': 0.5, 'max_depth': 3, 'n_estimators': 100}` |
| **Random Forest Classifier (RFC)** | 0.972 | 0.956 | 1.0 | 0.659 | 0.795 | `{'max_depth': 20, 'min_samples_split': 5, 'n_estimators': 100}` |
6. **Flask App**:
- Created a REST API using **Flask** to serve the model and make predictions.
## USAGE
The issue of **spam classification** is important in filtering out unwanted messages in many domains like **SMS**, **emails**, and **social media messages**. This solution provides a reliable, real-time method for detecting spam messages based on a variety of machine learning models.
## USE CASES
1. **SMS Spam Detection**: Automatically filter SMS messages that are spam.
2. **Email Spam Filtering**: Classify emails as spam or not.
3. **Real-time Applications**: Use in customer service to filter out spammy inquiries.
## LIBRARIES USED
- **Pandas**: For data manipulation.
- **NLTK**: For natural language processing (tokenization, stopwords, stemming).
- **Scikit-learn**: For model building, evaluation, and vectorization.
- **XGBoost**: For boosting-based classification.
- **Flask**: For creating the API and serving the model.
- **Matplotlib/Seaborn**: For data visualization.
## ADVANTAGES
- **Real-time predictions** for spam classification via the Flask API.
- **Multiple model comparison** to choose the best-performing algorithm.
- **Scalable**: Can be extended to classify emails or other types of messages.
## DISADVANTAGES
- **Imbalanced dataset**: If the dataset is highly imbalanced, some models may perform poorly.
- **Training time**: Training models like **XGBoost** and **Random Forest** can be time-consuming.
## APPLICATIONS
1. **SMS Filtering**: For mobile operators or messaging services to filter spam messages.
2. **Email Filtering**: To block spam emails in email clients like Gmail, Yahoo, etc.
3. **Customer Service**: Spam detection in automated customer support systems.
4. **Social Media**: Detect spam or malicious messages on platforms like Twitter, Facebook, etc.
## RESEARCH
I have researched various spam detection techniques and found that using **TF-IDF** with models like **Multinomial Naive Bayes, Logistic Regression** and **XGBoost** yields great results for spam classification tasks.
I explored using multiple models and evaluated their performance to choose the best approach for this problem.
## SCREENSHOTS
1. **Prediction Interface**:
- **Input 1**: "Congratulations! You've won a free iPhone!"
- **Input 2**: "Hey, are we still on for dinner tomorrow?"
2. **Model Evaluation**:
## CONCLUSION
This project successfully builds and deploys a **spam classification model** using multiple machine learning algorithms, evaluated through performance metrics, and deployed via **Flask** on **Render**. The **Multinomial Naive Bayes** ensures a higher F1-Score amongest the models. The real-time API implementation allows for efficient and easy deployment for production environments.
## REFERENCES
1. **Scikit-learn Documentation**: https://scikit-learn.org/stable/
2. **NLTK Documentation**: https://www.nltk.org/
3. **Kaggle Dataset**: https://www.kaggle.com/datasets/uciml/sms-spam-collection-dataset
4. **Flask Documentation**: https://flask.palletsprojects.com/
## YOUR NAME
ROHIT R DUSANE
GitHub: ['MY GITHUB'](https://github.com/RohitDusane)
## DISCLAIMER, IF ANY
This project is for educational purposes only. The model is based on the **SMS Spam Collection Dataset**, which may not reflect real-world spam classification challenges in production systems.