def plot_advanced_forecasts(test_data, forecasts_dict, series_idx=0):
"""Advanced plotting with multiple models and uncertainty bands"""
fig, axes = plt.subplots(2, 2, figsize=(15, 10))
fig.suptitle('Advanced GluonTS Forecasting Results', fontsize=16, fontweight="bold")
if not forecasts_dict:
fig.text(0.5, 0.5, 'No successful forecasts to display',
ha="center", va="center", fontsize=20)
return fig
if series_idx < len(test_data.label):
ts_label = test_data.label[series_idx]
ts_input = test_data.input[series_idx]['target']
colors = ['blue', 'red', 'green', 'purple', 'orange']
ax1 = axes[0, 0]
ax1.plot(range(len(ts_input)), ts_input, 'k-', label="Historical", alpha=0.8, linewidth=2)
ax1.plot(range(len(ts_input), len(ts_input) + len(ts_label)),
ts_label, 'k--', label="True Future", alpha=0.8, linewidth=2)
for i, (name, forecasts) in enumerate(forecasts_dict.items()):
if series_idx < len(forecasts):
forecast = forecasts[series_idx]
forecast_range = range(len(ts_input), len(ts_input) + len(forecast.mean))
color = colors[i % len(colors)]
ax1.plot(forecast_range, forecast.mean,
color=color, label=f'{name} Mean', linewidth=2)
try:
ax1.fill_between(forecast_range,
forecast.quantile(0.1), forecast.quantile(0.9),
alpha=0.2, color=color, label=f'{name} 80% CI')
except:
pass
ax1.set_title('Multi-Model Forecasts Comparison', fontsize=12, fontweight="bold")
ax1.legend()
ax1.grid(True, alpha=0.3)
ax1.set_xlabel('Time Steps')
ax1.set_ylabel('Value')
ax2 = axes[0, 1]
if all_forecasts:
first_model = list(all_forecasts.keys())[0]
if series_idx < len(all_forecasts[first_model]):
forecast = all_forecasts[first_model][series_idx]
ax2.scatter(ts_label, forecast.mean, alpha=0.7, s=60)
min_val = min(min(ts_label), min(forecast.mean))
max_val = max(max(ts_label), max(forecast.mean))
ax2.plot([min_val, max_val], [min_val, max_val], 'r--', alpha=0.8)
ax2.set_title(f'Prediction vs Actual - {first_model}', fontsize=12, fontweight="bold")
ax2.set_xlabel('Actual Values')
ax2.set_ylabel('Predicted Values')
ax2.grid(True, alpha=0.3)
ax3 = axes[1, 0]
if all_forecasts:
first_model = list(all_forecasts.keys())[0]
if series_idx < len(all_forecasts[first_model]):
forecast = all_forecasts[first_model][series_idx]
residuals = ts_label - forecast.mean
ax3.hist(residuals, bins=15, alpha=0.7, color="skyblue", edgecolor="black")
ax3.axvline(x=0, color="r", linestyle="--", linewidth=2)
ax3.set_title(f'Residuals Distribution - {first_model}', fontsize=12, fontweight="bold")
ax3.set_xlabel('Residuals')
ax3.set_ylabel('Frequency')
ax3.grid(True, alpha=0.3)
ax4 = axes[1, 1]
if evaluation_results:
metrics = ['MASE', 'sMAPE']
model_names = list(evaluation_results.keys())
x = np.arange(len(metrics))
width = 0.35
for i, model_name in enumerate(model_names):
values = [evaluation_results[model_name].get(metric, 0) for metric in metrics]
ax4.bar(x + i*width, values, width,
label=model_name, color=colors[i % len(colors)], alpha=0.8)
ax4.set_title('Model Performance Comparison', fontsize=12, fontweight="bold")
ax4.set_xlabel('Metrics')
ax4.set_ylabel('Value')
ax4.set_xticks(x + width/2 if len(model_names) > 1 else x)
ax4.set_xticklabels(metrics)
ax4.legend()
ax4.grid(True, alpha=0.3)
else:
ax4.text(0.5, 0.5, 'No evaluation\nresults available',
ha="center", va="center", transform=ax4.transAxes, fontsize=14)
plt.tight_layout()
return fig
if all_forecasts and test_data.label:
print("š Creating advanced visualizations...")
fig = plot_advanced_forecasts(test_data, all_forecasts, series_idx=0)
plt.show()
print(f"\nš Tutorial completed successfully!")
print(f"š Trained {len(trained_models)} model(s) on {len(df.columns) if 'df' in locals() else 10} time series")
print(f"šÆ Prediction length: 30 days")
if evaluation_results:
best_model = min(evaluation_results.items(), key=lambda x: x[1]['MASE'])
print(f"š Best performing model: {best_model[0]} (MASE: {best_model[1]['MASE']:.4f})")
print(f"\nš§ Environment Status:")
print(f" PyTorch Support: {'ā
' if TORCH_AVAILABLE else 'ā'}")
print(f" MXNet Support: {'ā
' if MX_AVAILABLE else 'ā'}")
else:
print("ā ļø Creating demonstration plot with synthetic data...")
fig, ax = plt.subplots(1, 1, figsize=(12, 6))
dates = pd.date_range('2020-01-01', periods=100, freq='D')
ts = 100 + np.cumsum(np.random.normal(0, 2, 100)) + 20 * np.sin(np.arange(100) * 2 * np.pi / 30)
ax.plot(dates[:70], ts[:70], 'b-', label="Historical Data", linewidth=2)
ax.plot(dates[70:], ts[70:], 'r--', label="Future (Example)", linewidth=2)
ax.fill_between(dates[70:], ts[70:] - 5, ts[70:] + 5, alpha=0.3, color="red")
ax.set_title('GluonTS Probabilistic Forecasting Example', fontsize=14, fontweight="bold")
ax.set_xlabel('Date')
ax.set_ylabel('Value')
ax.legend()
ax.grid(True, alpha=0.3)
plt.tight_layout()
plt.show()
print("\nš Tutorial demonstrates advanced GluonTS concepts:")
print(" ā¢ Multi-series dataset generation")
print(" ā¢ Probabilistic forecasting")
print(" ā¢ Model evaluation and comparison")
print(" ā¢ Advanced visualization techniques")
print(" ā¢ Robust error handling")
