You can deploy this model and start earning money today!

Moongirl99/
ensemble-models

Name: ensemble-models
Availability: InStock

Views

No views yet

Huntrix Ensemble Models

Model Description

This repository contains the best performing ensemble models for the Huntrix incident classification task.

Performance Metrics

Overall Validation Score: 0.5905
Overall Test Score: 0.5566
ST1 Validation Score: 0.7440
ST2 Validation Score: 0.4369

Model Architecture

ST1: Hazard Category & Product Category prediction using LinearSVC
ST2: Hazard & Product prediction using LinearSVC + SGDClassifier with hierarchical approach

Hyperparameters

json

{
  "TFIDF_MAX_FEATURES_ST1": 80000,
  "TFIDF_NGRAM_RANGE_ST1": [
    1,
    2
  ],
  "SVM_C_ST1": 2.0,
  "TFIDF_MAX_FEATURES_ST2": 50000,
  "TFIDF_NGRAM_RANGE_ST2": [
    1,
    2
  ],
  "SVM_C_HAZARD_ST2": 3.0,
  "MIN_GROUP_SIZE": 10,
  "SGD_ALPHA_PRODUCT": 3e-05,
  "SGD_EPOCHS": 30,
  "CONF_GAP": 0.15
}

Files

preprocessor_st1.pkl: TF-IDF vectorizer and metadata encoder for ST1
preprocessor_st2.pkl: TF-IDF vectorizer for ST2
label_encoders.pkl: Label encoders for all target variables
prior_product.pkl: Prior product distribution by category
st1_hazard_category_svm.pkl: ST1 hazard category classifier
st1_product_category_svm.pkl: ST1 product category classifier
st2_hazard_global.pkl: ST2 global hazard classifier
st2_hazard_by_hcat.pkl: ST2 hazard classifiers grouped by hazard category
st2_product_global.pkl: ST2 global product classifier
st2_product_by_pcat.pkl: ST2 product classifiers grouped by product category

Usage

python

import pickle
from huggingface_hub import hf_hub_download

# Download and load models
preprocessor_st1 = pickle.load(open(hf_hub_download(repo_id="Moongirl99/ensemble-models", filename="models/preprocessor_st1.pkl"), 'rb'))
# ... load other models similarly

Training Details

Training time: 22.0 seconds
Tested configurations: 32
Successful configurations: 32
Date: 2026-03-03