| license: mit | |
| tags: | |
| - gender-classification | |
| - advertising | |
| - lightgbm | |
| # Gender Prediction Model for Targeted Advertising | |
| ## Model Description | |
| LightGBM classifier trained on user behavior data to predict gender (Male/Female) for ad targeting. Achieves **82.3% accuracy** and **0.817 F1-score**. | |
| ## Features | |
| - **User Agent parsing** (browser, OS, mobile/tablet) | |
| - **Geographic data** (country, region, timezone) | |
| - **Referer vector components** (10-dimensional embedding) | |
| - **Time features** (hour, weekday, weekend) | |
| ## Performance | |
| | Metric | Score | | |
| | --------- | ----- | | |
| | Accuracy | 0.823 | | |
| | Precision | 0.805 | | |
| | Recall | 0.828 | | |
| | F1-Score | 0.817 | | |
| ## Usage | |
| ```python | |
| import joblib | |
| model = joblib.load('gender_classifier.pkl') | |
| imputer = joblib.load('imputer.pkl') | |
| # Preprocess data using same steps as training | |
| # Then predict | |
| ``` | |