نظام ذكي يساعد علي الحمايه من الرسال الضاره في البريد

Spam Email Classification Project

This repository contains a Jupyter Notebook for classifying SMS or email messages as spam or ham (non-spam) using classic machine learning and natural language processing techniques. The notebook includes the full pipeline from data loading and cleaning to feature engineering, model training, hyperparameter tuning, resampling for class imbalance, ensemble modeling, and evaluation.

The dataset used is the SMS Spam Collection Dataset from the UCI Machine Learning Repository. It contains labeled text messages as either ham or spam.

Main steps in the notebook:

Data loading and initial exploration

Data cleaning (deduplication, label encoding)

Exploratory Data Analysis (EDA) with count plots and pie charts

Text preprocessing using NLTK (tokenization, sentence/word/character counts)

Feature extraction with CountVectorizer

Train/test splitting

Model benchmarking with LazyPredict

Hyperparameter tuning using GridSearchCV (BernoulliNB)

Addressing class imbalance with RandomUnderSampler

Model evaluation using confusion matrix, classification report, accuracy, precision, recall, F1-score, ROC curve, and AUC

Ensemble modeling with VotingClassifier (combining Decision Tree, BernoulliNB, RandomForest)

Best results achieved with BernoulliNB (after RandomUnderSampler):

Accuracy: ~98.65%

F1-score: ~95.21%

Precision: ~97.20%

Recall: ~93.29%

Requirements:

Python 3.x

Jupyter Notebook

pandas

numpy

matplotlib

seaborn

scikit-learn

nltk

lazypredict

xgboost

imbalanced-learn

Install dependencies:

pip install pandas numpy matplotlib seaborn scikit-learn nltk lazypredict xgboost imbalanced-learn

How to run:

Clone the repository

Install the requirements

Open the notebook in Jupyter

Run all cells to reproduce results

License: MIT

Acknowledgements: UCI Machine Learning Repository for the dataset, scikit-learn, NLTK, LazyPredict, XGBoost