MBTI Personality Prediction using NLP

This project predicts MBTI (Myers-Briggs Type Indicator) personality types from user-generated text using Natural Language Processing (NLP) and machine learning techniques.

More about MBTI on wikipedia and also in 16personalities.com.

📌 Features

• Predicts the MBTI type of the user based on three questions.
• Recommends movies and books for the user based on their types.

🧪 Model Pipeline (For existing logistic regression)

Code for model training: google colab file

1. Text Preprocessing

   • Converting to lowercase.
   • Removing URL, special characters.
   • Lemmatization
   • Stopword removal

2. Vectorization

   • Converting text into TF-IDF(Term Frequency and Inverse Document Frequency) features for numerical feature extraction.

3. Classification
   Logistic regression model of scikit-learn

4. Accuracy with Logistic regression

   • Oversampling: 91%, with 84% on MBTI 500 dataset.
   • Augmentation: 89%
   • Using class weights: 67%
   • Undersampling: 56%

📚Dataset used

• MBTI dataset
• Movie dataset
• Books dataset
• MBTI testing dataset

🚀Project workflow

📦 API Usage

The API returns the top three MBTI types along with their confidences, as predicted by the model.

Endpoint: POST https://personality-predictor-mtm1.onrender.com/predict
Payload:

{
  "text": "I love quiet evenings and meaningful conversations."
}

Response:

{
  "confidences": [
    0.3015382561625205,
    0.1649483711317476,
    0.10491362966713379
  ],
  "predictions": [
    "INFP",
    "INFJ",
    "INTP"
  ]
}

🔧 Tech Stack

• Python (Flask) for deploying the model. Hosted via onrender.
• scikit-learn for model building and accuracy analysis.
• React with Javascript, for building the frontend. Hosted via Vercel.
• NodeJS for Backend- authentication, fetching movies/books of particular types. Hosted via onrender.
• MongoDB atlas for database, storing user info and movies/books data.