Kinara-xgb-model

Kinara Regression Model

The Kinara Regression Model is part of the Soulprint Archetypes framework developed under the MyVillage Project / Coding in Color initiative.
It predicts a Kinara score (0–1) for any input text, where Kinara reflects purpose, spirit, and vision — inspired by the Kwanzaa candleholder and symbolic elders such as Fannie Lou Hamer.

🔑 Archetype Background

According to the Soulprint Archetypes Document, Kinara is:

• Adjective: Guiding
• Description: Kinara is the keeper of purpose, spirit, and vision. They mentor, anchor, and align groups with mission and integrity.
• Catchphrase: Hold the light high.

📊 Model Details

• Type: XGBoost Regressor
• Embeddings: all-mpnet-base-v2 sentence embeddings
• Dataset: ~1,210 rows synthetic + curated data (balanced, strict-unique)
• Target: Kinara Score (continuous between 0 and 1)

Training Metrics

• MSE: 0.0086
• RMSE: 0.0927
• R² Score: 0.8856

These metrics indicate a strong fit with low error and high explanatory power.

🚀 Usage

First, install dependencies:

pip install xgboost sentence-transformers joblib
Then load and use the model:
import joblib
from sentence_transformers import SentenceTransformer

# Load model from Hugging Face
model = joblib.load("mjpsm/Kinara_xgb_model/Kinara_xgb_model.pkl")
embedder = SentenceTransformer("sentence-transformers/all-mpnet-base-v2")

# Example input
text = "During a heated family argument, I stayed calm and reminded everyone of our values."
X = embedder.encode([text])
pred = model.predict(X)

print(f"Predicted Kinara Score: {pred[0]:.2f}")

Example Predictions

• Input: “During a heated family argument, I stayed calm and reminded everyone of our values.” Output: 0.75

• Input: “I hesitated when a classmate asked for advice, and they left still confused.” Output: 0.24

• Input: “At a neighborhood gathering, I shared a story about unity that brought people closer.” Output: 0.75