Create README.md
Browse files
README.md
CHANGED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# 🧠 TabTransformer Multitask Model for Churn, Tenure, and LTV Prediction
|
| 2 |
+
|
| 3 |
+
This model is a multitask `TabTransformer` implemented in PyTorch, designed to perform:
|
| 4 |
+
|
| 5 |
+
- **Binary classification** for customer **churn**
|
| 6 |
+
- **Regression** for customer **tenure**
|
| 7 |
+
- **Regression** for customer **LTV (Lifetime Value)**
|
| 8 |
+
|
| 9 |
+
It is saved as a pickle file: `model.pkl` and includes all custom layers (e.g., positional encoding).
|
| 10 |
+
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
## 🧩 Model Architecture
|
| 14 |
+
|
| 15 |
+
- Tabular input with:
|
| 16 |
+
- `x_num`: Numerical features (projected into latent space)
|
| 17 |
+
- `x_cat`: Categorical features (embedded + transformer)
|
| 18 |
+
- Transformer-based attention over categorical embeddings
|
| 19 |
+
- Multi-head output for multitask predictions:
|
| 20 |
+
- `Churn`: Sigmoid activation for binary classification
|
| 21 |
+
- `Tenure` and `LTV`: Linear regression heads
|
| 22 |
+
|
| 23 |
+
---
|
| 24 |
+
|
| 25 |
+
## 🧪 How to Use
|
| 26 |
+
|
| 27 |
+
### 1. Install Dependencies
|
| 28 |
+
|
| 29 |
+
```bash
|
| 30 |
+
pip install torch pandas
|