mjpsm/Idea-Difficulty-XGB

🧾 Overview

This model predicts the difficulty of a business idea as Low, Medium, or High.
It is part of the Entrepreneurial Readiness series of tabular classifiers (alongside Skill Level, Risk Tolerance, and Confidence).

The model was trained with XGBoost on a 2,000-row synthetic dataset of structured features that capture common difficulty drivers.

---

📥 Input Features

Feature	Type	Range	Definition
capital_required	int	1–10	How much upfront capital is needed (1 = minimal, 10 = very high)
technical_complexity	int	1–10	How technically difficult the product/service is to build or maintain
market_competition	int	1–10	How crowded the target market is with competitors
customer_acquisition_difficulty	int	1–10	How difficult it is to acquire and retain customers
regulatory_hurdles	int	1–10	The degree of legal/regulatory challenges
time_to_mvp_months	int	1–60	Estimated time to Minimum Viable Product launch (in months)
team_expertise_required	int	1–10	Level of specialized expertise/team members required
scalability_requirement	int	1–10	Degree to which scaling is required for success

Target label:

• Low = Idea is relatively easy to execute
• Medium = Moderately challenging
• High = Difficult, requiring significant resources and expertise

---

📊 Performance

• Accuracy: 0.9733
• Macro F1: 0.9733
• Log Loss: 0.0584

Confusion Matrix (rows = true, cols = predicted):

	High	Low	Medium
High	100	0	0
Low	0	96	4
Medium	2	2	96

---

🚀 Quickstart (load from the Hub)

# Load directly from: mjpsm/Idea-Difficulty-XGB
from huggingface_hub import hf_hub_download
from xgboost import XGBClassifier
import pandas as pd, json

REPO_ID = "mjpsm/Idea-Difficulty-XGB"
model_path = hf_hub_download(REPO_ID, "xgb_model.json")

clf = XGBClassifier()
clf.load_model(model_path)

# IMPORTANT: Use the same feature names/order as training
FEATURES = [
    "capital_required","technical_complexity","market_competition",
    "customer_acquisition_difficulty","regulatory_hurdles",
    "time_to_mvp_months","team_expertise_required","scalability_requirement"
]

row = pd.DataFrame([{
    "capital_required": 7,
    "technical_complexity": 9,
    "market_competition": 6,
    "customer_acquisition_difficulty": 8,
    "regulatory_hurdles": 7,
    "time_to_mvp_months": 18,
    "team_expertise_required": 5,
    "scalability_requirement": 9
}], columns=FEATURES)

pred_id = int(clf.predict(row)[0])

# If label_map.json is NOT uploaded, default to alphabetical LabelEncoder order:
CLASSES = ["High","Low","Medium"]  # update if you publish label_map.json
print("Predicted Idea Difficulty:", CLASSES[pred_id])

# OPTIONAL: If you later upload 'label_map.json', prefer this:
# lm_path = hf_hub_download(REPO_ID, "label_map.json")
# label_map = json.load(open(lm_path))
# inv_map = {v:k for k,v in label_map.items()}
# print("Predicted Idea Difficulty:", inv_map[pred_id])