| --- |
| license: apache-2.0 |
| --- |
| Model Card: TERPredictor V1 |
| 📌 Model Name |
| TERPredictor V1 – A linear regression model for predicting the Total Expense Ratio (TER) of mutual fund Regular Plans. |
|
|
| 📖 Overview |
| TERPredictor V1 is a regression model trained to estimate the 'Regular Plan - Total TER (%)' of mutual funds based on various financial features. It uses a simple linear regression approach and achieves near-perfect performance on the test set. Due to the unusually high accuracy, this model is best suited for exploratory analysis and feature relationship interpretation, rather than generalization to unseen data. |
|
|
| 📊 Intended Uses |
| Expense Ratio Estimation: Estimate TER for new or hypothetical mutual fund structures. |
|
|
| Outlier Detection: Identify funds with unusually high or low TERs. |
|
|
| Feature Impact Analysis: Understand which components most influence TER. |
|
|
| 🧠 Model Architecture |
| Attribute Value |
| Model Type Linear Regression |
| Framework scikit-learn |
| Input Features 10 float64 columns |
| Target Variable Regular Plan - Total TER (%) |
| Identifier Dropped Scheme Name (object) |
| 📚 Training Details |
| Dataset Size: 1,622 samples |
|
|
| Train/Test Split: 1297 / 325 |
|
|
| Missing Values: None |
|
|
| Preprocessing: |
|
|
| Dropped identifier column (Scheme Name) |
|
|
| No normalization required due to linear model simplicity |
|
|
| 📈 Evaluation Metrics |
| Metric Value |
| Mean Squared Error (MSE) 0.000001 |
| R-squared (R²) 0.999999 |
| ⚠️ Note: These metrics suggest potential data leakage or a deterministic relationship between features and target. Use with caution. |
|
|
| 🚀 How to Use |
| python |
| from terpredictor import TERModel |
|
|
| model = TERModel.load_pretrained("your-huggingface-username/terpredictor-v1") |
| input_data = { |
| "feature_1": 0.12, |
| "feature_2": 0.03, |
| ... |
| } |
| predicted_ter = model.predict(input_data) |
| ⚠️ Limitations |
| Potential Data Leakage: Extremely high R² may indicate the target is directly derived from input features. |
| |
| Limited Generalization: Not recommended for predicting TER on unseen or structurally different funds. |
|
|
| No Feature Engineering: Model assumes raw features are sufficient. |
|
|
| 📄 License |
| MIT License |
|
|
| 👤 Author |
| Created by [Your Name or Organization] |
|
|
| 📚 Recommendations for Open-Sourcing |
| Include full training code and preprocessing steps |
|
|
| Provide detailed explanation of evaluation metrics |
|
|
| Add cautionary notes about performance anomalies |
|
|
| Consider publishing a cleaned or anonymized version of the dataset |
