Resume Classification using Naive Bayes

A simple NLP project that classifies resumes into job categories using a Bag of Words representation and a Multinomial Naive Bayes model.

This notebook cleans resume text, extracts features with CountVectorizer, trains a MultinomialNB classifier, evaluates the model, and saves the trained pipeline with Joblib.

The dataset contains:

• 13,389 resume samples
• 2 columns: Category and Text
• 43 classes in total

1. Load the dataset
2. Explore category distribution
3. Clean resume text
4. Convert text to features with Bag of Words
5. Train a Naive Bayes classifier
6. Evaluate using accuracy, classification report, confusion matrix, and cross-validation
7. Save the model as a .pkl file

• Test Accuracy: 71.32%
• Cross-Validation Mean Accuracy: 68.35%

The trained model was tested with example resumes such as:

• Data Scientist resume → predicted as Data Science
• Frontend Developer resume → predicted as React Developer

• Resume Classification using Naive Bayes.ipynb — main notebook
• resume_classifier.pkl — saved trained model
• requirements.txt — Python dependencies
• images/ — result plots for GitHub README

# 1. Install dependencies
pip install -r requirements.txt

# 2. Open the notebook
jupyter notebook

If you are running in Google Colab, upload the dataset and then run the notebook cells.

This project uses:

• pandas
• numpy
• scikit-learn
• matplotlib
• seaborn
• joblib

This is a strong baseline NLP classification project, but it is not perfect yet. A few things can still be improved:

• Google Colab-specific code (from google.colab import files) should be removed or replaced for GitHub/local use.
• The dataset path is hardcoded as Dataset.csv, so make sure the file name matches in the repo.
• Some classes have weaker recall, so future improvement could include TF-IDF, better preprocessing, or a tuned linear model.

• Replace Bag of Words with TF-IDF
• Try Logistic Regression or Linear SVM
• Tune hyperparameters with GridSearchCV
• Add a simple web app or Streamlit demo

Add a license here if you plan to open-source the repository.