MAP Misconception Detection – Classical Models
This repository contains classical machine learning components used for the Kaggle MAP (Misconception Annotation Project) competition. We developed two independent pipelines using TF-IDF vectorization and Logistic Regression for:
- Task 1: Category classification
- Task 2: Misconception label prediction
📁 Included Files
🔹 Task 1 (Category Classification)
| File | Description |
|---|---|
tfidf_task1.pkl |
TF-IDF vectorizer fitted on student answers for Task 1 |
logreg_task1.pkl |
Trained logistic regression model for predicting category labels |
🔹 Task 2 (Misconception Detection)
| File | Description |
|---|---|
tfidf_task2.pkl |
TF-IDF vectorizer fitted on student answers for Task 2 |
logreg_task2.pkl |
Trained logistic regression model for predicting misconception labels |
🔹 Label Encoders
| File | Description |
|---|---|
label_encoder_category.pkl |
LabelEncoder used to convert category labels into integers |
label_encoder_miscon.pkl |
LabelEncoder used to convert misconception labels into integers |
🧠 Usage
To use the models in your code:
import joblib
# Load Task 1 components
tfidf1 = joblib.load("tfidf_task1.pkl")
clf1 = joblib.load("logreg_task1.pkl")
# Load Task 2 components
tfidf2 = joblib.load("tfidf_task2.pkl")
clf2 = joblib.load("logreg_task2.pkl")
# Load label encoders
le_category = joblib.load("label_encoder_category.pkl")
le_miscon = joblib.load("label_encoder_miscon.pkl")
🛠️ Requirements
These models were trained using:
- scikit-learn
- joblib
```bash
pip install scikit-learn joblib
📜 Licens: MIT License. You are free to use, modify, and distribute this model with proper attribution.
✍️ Author: Godspower Maurice (PythonCreate) Created for the MAP Math Misconceptions @ Kaggle NLP competition — 2025.