mjpsm / Kuumba-xgb-model

Kuumba XGBoost Regression Model

🧾 Overview

The Kuumba XGBoost Regression Model is part of the Soulprint Archetype family of machine learning models. It is trained to estimate the Kuumba score (a measure of creativity, innovation, and artistic contribution) from natural language text.

Kuumba, meaning creativity in Swahili, represents the archetype of innovation, artistry, and imaginative problem-solving within the Soulprint Village Map.

This model helps classify and measure the degree of Kuumba expressed in community activities, projects, and narratives.

📊 Training Data

• Dataset size: 1082 rows

• Input: Natural language text ("input")

• Output: Kuumba score ("output") between 0.0 and 1.0

• Dataset contributions by Frank (initial structure) and updated for alignment with Soulprint schema (input/output).

⚙️ Model Details

• Architecture: XGBoost Regressor

• Embeddings: all-mpnet-base-v2 from SentenceTransformers

• Hyperparameters:

  • n_estimators=500

  • learning_rate=0.05

  • max_depth=6

  • subsample=0.8

  • colsample_bytree=0.8

🧪 Performance

• MSE: 0.0154

• RMSE: 0.1241

• R² Score: 0.7431

✅ The model explains ~74% of the variance in Kuumba ratings. | ✅ Average prediction error ~0.12 on the 0–1 scale.

🚀 Usage

installation

pip install sentence-transformers xgboost joblib huggingface_hub

example interface

import joblib
from sentence_transformers import SentenceTransformer
from huggingface_hub import hf_hub_download

# -----------------------------
# 1. Download model from Hugging Face Hub
# -----------------------------
REPO_ID = "mjpsm/Kuumba_xgb_model"
FILENAME = "Kuumba_xgb_model.pkl"

model_path = hf_hub_download(repo_id=REPO_ID, filename=FILENAME)
model = joblib.load(model_path)

# Load embeddings
embedder = SentenceTransformer("all-mpnet-base-v2")

# -----------------------------
# 2. Run Prediction
# -----------------------------
text = "The youth created a new festival celebrating African art and innovation."
embedding = embedder.encode([text])
score = model.predict(embedding)[0]

print("Predicted Kuumba Score:", round(float(score), 3))

👥 Contributors

• Dataset: Frank (original dataset), Mazamesso (schema alignment + training)

• Model Training: Mazamesso (Coding in Color / MyVillage Project)

• Inspiration: Soulprint Archetype framework