Module 105 — Deliverables for quantitative models (evidence pack)

import json
import os
from datetime import datetime

import numpy as np
import matplotlib.pyplot as plt

# -----------------------------
# Create a run id and folders
# -----------------------------
run_id = datetime.now().strftime("run_%Y%m%d_%H%M%S")
base = os.path.join("artefacts", run_id)
plots_dir = os.path.join(base, "plots")
os.makedirs(plots_dir, exist_ok=True)

# -----------------------------
# Example "config snapshot"
# -----------------------------
config = {
    "problem": "toy_forecast",
    "target": "y_{t+1}",
    "horizon": 1,
    "split": {"type": "time_holdout", "train_frac": 0.7},
    "baseline": "last_value",
    "model": "linear_lags",
    "seed": 0,
}

with open(os.path.join(base, "config_snapshot.json"), "w") as f:
    json.dump(config, f, indent=2)

# -----------------------------
# Generate a tiny dataset + metrics
# -----------------------------
rng = np.random.default_rng(config["seed"])
T = 300
t = np.arange(T)
y = 0.02 * t + 0.6 * np.sin(2*np.pi*t/20) + rng.normal(0, 0.8, T)

# Baseline: last value
y_hat = np.roll(y, 1)
y_hat[0] = np.nan

# Split
idx = np.arange(T)
valid = ~np.isnan(y_hat)
idx = idx[valid]
y_v = y[valid]
y_hat_v = y_hat[valid]

split = int(config["split"]["train_frac"] * len(idx))
test_y = y_v[split:]
test_hat = y_hat_v[split:]

rmse = float(np.sqrt(np.mean((test_y - test_hat)**2)))
mae = float(np.mean(np.abs(test_y - test_hat)))

metrics = {"test_rmse": rmse, "test_mae": mae, "n_test": int(len(test_y))}

with open(os.path.join(base, "metrics.json"), "w") as f:
    json.dump(metrics, f, indent=2)

# -----------------------------
# One plot
# -----------------------------
plt.figure(figsize=(8, 4))
plt.plot(test_y[:80], label="y (test)")
plt.plot(test_hat[:80], label="baseline last value")
plt.title("Deliverable plot: baseline forecast on test window")
plt.xlabel("Time index (test)")
plt.ylabel("Value")
plt.legend()
plt.tight_layout()
plt.savefig(os.path.join(plots_dir, "baseline_test_overlay.png"), dpi=150)
plt.close()

# -----------------------------
# Short report (assumptions + limitations)
# -----------------------------
report = f"""# Evidence pack: {run_id}

## Summary
This run evaluates a toy baseline forecast for a synthetic series.

## Config
- Target: {config["target"]}
- Horizon: {config["horizon"]}
- Split: {config["split"]}

## Metrics (test)
- RMSE: {rmse:.4f}
- MAE: {mae:.4f}
- n_test: {len(test_y)}

## Assumptions
- Synthetic data is stationary enough for demonstration.
- Baseline uses y_t to predict y_(t+1).

## Limitations
- Not a trading evaluation.
- No regime stability checks included.
"""

with open(os.path.join(base, "report.md"), "w") as f:
    f.write(report)

print("Created evidence pack at:", base)
print("Files:", os.listdir(base))