mirzafathir/usd-idr-xgboost

USD/IDR Exchange Rate Forecasting — XGBoost Model

MAE: 37.9 IDR | MAPE: 0.224% | R²: 0.984

Trained on daily USD/IDR data from 2000–2026 using 27 macroeconomic and technical features.

Usage

import joblib, json, numpy as np

xgb = joblib.load("xgboost_model.pkl")
cfg = json.load(open("model_config.json"))

# X shape: (n_samples, 27)
preds = xgb.predict(X)  # returns log-returns

Back-transform to price level

# Recursive forecast
prices = [current_price]
for ret in preds:
    prices.append(prices[-1] * np.exp(ret))

Files

File	Description
xgboost_model.pkl	Main XGBoost model (log-return target)
xgboost_q10.pkl	Quantile 10% (lower CI)
xgboost_q90.pkl	Quantile 90% (upper CI)
feature_scaler.pkl	StandardScaler for features
model_config.json	Feature columns + config
evaluation_results.csv	Test set metrics

For the full interactive dashboard: HF Spaces