🧠 Myers-Briggs Personality Prediction from Text

This project uses Natural Language Processing (NLP) and machine learning to predict a user's MBTI personality type based on their forum posts. The Myers-Briggs Type Indicator (MBTI) categorizes personalities into 16 types using 4 traits: Introversion/Extroversion, Sensing/Intuition, Thinking/Feeling, and Judging/Perceiving.

---

🚀 Project Overview

• Goal: Predict the MBTI type (e.g., INFP, ESTJ) from user-generated text
• Dataset: MBTI user forum posts from Kaggle + synthetic forum message data
• Tech stack: Python, scikit-learn, pandas, BeautifulSoup, Matplotlib, Seaborn, Plotly

---

🧬 Data Description

• mbti_1.csv: MBTI type (type) + concatenated forum posts (posts)
• Users.csv: User IDs and metadata (e.g., age, gender)
• ForumMessages.csv: Individual message-level data per user

---

🧹 Preprocessing Steps

• Cleaned HTML and URLs using BeautifulSoup and regex
• Removed special tokens and delimiters like |||
• Tokenized and vectorized text using:
  • TF-IDF for feature extraction
  • Truncated SVD for dimensionality reduction

---

🤖 Model Building

Tested multiple classifiers using scikit-learn pipelines:

• ✅ Multinomial Naive Bayes
• ✅ Logistic Regression
• ✅ ExtraTreesClassifier (ensemble)

Evaluation Metrics (via cross-validation):

• Accuracy
• F1-micro score
• Log loss

Best result:
🧠 Logistic Regression — F1 Score: ~0.656, Log Loss: ~1.3

---

📈 Learning Curve Analysis

Generated learning curves to assess:

• Model generalization
• Overfitting/underfitting patterns
• Impact of training set size on performance

---

📊 Visualization

• Distribution of MBTI types in the dataset
• Predicted MBTI types for new users (bar + pie charts)
• Trait mapping: e.g., "INFJ" → "Introversion Intuition Feeling Judging"

---

To install dependencies:

pip install -r requirements.txt