03062025_V2_UMAP_Embedding_Classifier

This repository contains two final AutoGluon TabularPredictor models (binary and multi-class) built using UMAP-reduced embeddings from the Alibaba-NLP/gte-large-en-v1.5 model.

Key Details

• UMAP for Binary Classification: Best n_components tuned via Optuna = 11.
• UMAP for Multi-class Classification: Best n_components tuned via Optuna = 43.
• Data: 112 technical questions with tiering classifications (0–4).
• Performance Metrics:
  • Binary: Accuracy ≈95.65%, F1 ≈0.97, ROC AUC ≈0.91.
  • Multi-class: Accuracy ≈56.52%, F1 ≈0.59, ROC AUC ≈0.74.

Usage

1. Loading the Models:

   from autogluon.tabular import TabularPredictor
   binary_predictor = TabularPredictor.load("binary_final_model")
   multi_predictor = TabularPredictor.load("multiclass_final_model")
   

2. Preprocessing: Generate embeddings for your input text using the Alibaba-NLP/gte-large-en-v1.5 model and apply the UMAP transformation with the provided reducer files (umap_reducer_binary.joblib and umap_reducer_multi.joblib).

3. Prediction: Use predict() and predict_proba() to obtain predictions.

License

This project is licensed under the Apache-2.0 License.

Contact

For questions or collaboration, please contact LeiPricingManager.