| --- |
| license: apache-2.0 |
| --- |
| --- |
| license: mit |
| language: en |
| tags: |
| - sklearn |
| - text-classification |
| - psychology |
| - mbti |
| --- |
|
|
| # MBTI Personality Predictor |
|
|
| This repository contains scikit-learn models for predicting MBTI personality types from text. |
|
|
| ## Model Details |
|
|
| This system consists of a `TfidfVectorizer` and four separate `LogisticRegression` models, one for each of the MBTI dimensions: |
|
|
| * **Mind:** Introversion (I) vs. Extraversion (E) |
| * **Energy:** Intuition (N) vs. Sensing (S) |
| * **Nature:** Thinking (T) vs. Feeling (F) |
| * **Tactics:** Judging (J) vs. Perceiving (P) |
|
|
| ## Intended Use |
|
|
| These models are intended for educational purposes and to demonstrate building an NLP classification system. They can be used to predict an MBTI type from a block of English text. **This is not a clinical or diagnostic tool.** |
|
|
| ## Training Data |
|
|
| The models were trained on the [Myers-Briggs Personality Type Dataset](https://www.kaggle.com/datasets/datasnaek/mbti-type) from Kaggle, which contains over 8,600 entries of text from social media forums. |
|
|
| ## Training Procedure |
|
|
| Text was cleaned by removing URLs and punctuation, lemmatizing, and removing stopwords. The text was then vectorized using TF-IDF (`max_features=5000`, `ngram_range=(1, 2)`). Each `LogisticRegression` model was trained with `class_weight='balanced'` to counteract the natural imbalance in the dataset. |
|
|
| ### Evaluation Results |
|
|
| Average F1-Scores on the test set: |
| * **I/E Model:** Macro F1-Score: ~0.79 |
| * **N/S Model:** Macro F1-Score: [Add Your Score] |
| * **F/T Model:** Macro F1-Score: [Add Your Score] |
| * **J/P Model:** Macro F1-Score: [Add Your Score] |
|
|
| ## How to Use |
|
|
| ```python |
| import joblib |
| from huggingface_hub import hf_hub_download |
| |
| # Define the repo ID |
| repo_id = "YOUR_USERNAME/mbti-personality-predictor" |
| |
| # Download all the model files |
| vectorizer = joblib.load(hf_hub_download(repo_id=repo_id, filename="mbti_vectorizer.joblib")) |
| model_ie = joblib.load(hf_hub_download(repo_id=repo_id, filename="mbti_model_ie.joblib")) |
| model_ns = joblib.load(hf_hub_download(repo_id=repo_id, filename="mbti_model_ns.joblib")) |
| model_ft = joblib.load(hf_hub_download(repo_id=repo_id, filename="mbti_model_ft.joblib")) |
| model_jp = joblib.load(hf_hub_download(repo_id=repo_id, filename="mbti_model_jp.joblib")) |
| |
| # You can now use these objects for prediction... |
