Improve terminology and update app descriptions

- Replace 'training' with 'historical context' (zero-shot forecasting)
- Update 'Actual Test Data' to 'Actual Values'
- Change 'Train/Test Split' to 'Forecast Window'
- Update app title and description to reflect multi-model support
- Add comprehensive About section with feature list
- Clarify slider help text for backtesting
- Make ERCOT info message conditional

app.py

@@ -20,8 +20,11 @@ def fetch_data(source_name, days_back=180):
     return fetch_data_with_fallback(source_name, days_back)
 
 # Streamlit app interface
-st.title("Electricity Market Price Forecasting with Chronos-2")
-st.write("This demo uses **Chronos-2** to forecast electricity prices from ERCOT (Texas) market data.")
+st.title("⚡ Electricity Market Price Forecasting")
+st.write("""
+This demo uses **Amazon Chronos** pretrained models for zero-shot time series forecasting on electricity market data.
+Select a model, choose your data source, and evaluate forecasting performance with backtesting on real ERCOT prices.
+""")
 
 # Model selection
 available_model_names = ModelConfig.get_model_names()
@@ -66,7 +69,8 @@ else:
         default_data.strip(),
         height=150
     )
-    st.info("💡 Live data from ERCOT's Day-Ahead Market (DAM SPP) - averaged across all settlement points per day")
+    if "ERCOT" in data_source_used:
+        st.info("💡 Live data from ERCOT's Day-Ahead Market (DAM SPP) - Daily average prices across all settlement points")
 
 try:
     time_series_data = process_input(user_input)
@@ -74,34 +78,34 @@ except ValueError:
     st.error("Please make sure all values are numbers, separated by commas.")
     time_series_data = []  # Set empty data on error to prevent further processing
 
-# Select the number of days for testing (forecasting on known data)
+# Select the forecast window for backtesting
 max_test_days = min(64, len(time_series_data) - 10) if len(time_series_data) > 10 else 1
 prediction_length = st.slider(
-    "Select Test Window (Days to Forecast & Compare)", 
+    "Forecast Horizon (Days to Backtest)", 
     min_value=1, 
     max_value=max_test_days, 
     value=min(14, max_test_days),
-    help="The last N days will be used as test data. The model will forecast these days and compare with actual values."
+    help="The model will use historical context to forecast the last N days, then compare predictions with actual values to evaluate performance."
 )
 
 # If data is valid, perform the forecast
 if time_series_data:
-    # Split data into train and test
-    train_length = len(time_series_data) - prediction_length
-    train_data = time_series_data[:train_length]
-    test_data = time_series_data[train_length:]
+    # Split data into context (historical) and test
+    context_length = len(time_series_data) - prediction_length
+    context_data = time_series_data[:context_length]
+    test_data = time_series_data[context_length:]
     
     # Create timestamps
     end_date = datetime.now()
     start_date = end_date - timedelta(days=len(time_series_data) - 1)
     all_dates = pd.date_range(start=start_date, periods=len(time_series_data), freq='D')
-    train_dates = all_dates[:train_length]
-    test_dates = all_dates[train_length:]
+    context_dates = all_dates[:context_length]
+    test_dates = all_dates[context_length:]
     
-    # Create a DataFrame for training
+    # Create a DataFrame with context for the model
     context_df = pd.DataFrame({
-        'timestamp': train_dates,
-        'target': train_data,
+        'timestamp': context_dates,
+        'target': context_data,
         'id': 'ercot_prices'
     })
 
@@ -128,12 +132,12 @@ if time_series_data:
 
     # Plot the historical and forecasted data with dates
     plt.figure(figsize=(14, 7))
-    plt.plot(train_dates, train_data, color="royalblue", label="Training Data", linewidth=2)
-    plt.plot(test_dates, test_data, color="green", label="Actual Test Data", linewidth=2, marker='o', markersize=4)
+    plt.plot(context_dates, context_data, color="royalblue", label="Historical Context", linewidth=2)
+    plt.plot(test_dates, test_data, color="green", label="Actual Values", linewidth=2, marker='o', markersize=4)
     plt.plot(test_dates, median, color="tomato", label="Forecast", linewidth=2, linestyle='--', marker='s', markersize=4)
     plt.fill_between(test_dates, low, high, color="tomato", alpha=0.3, label="80% Prediction Interval")
-    plt.axvline(x=train_dates[-1], color='gray', linestyle=':', linewidth=1, alpha=0.7)
-    plt.text(train_dates[-1], plt.ylim()[1]*0.95, ' Train/Test Split', fontsize=10, color='gray')
+    plt.axvline(x=context_dates[-1], color='gray', linestyle=':', linewidth=1, alpha=0.7)
+    plt.text(context_dates[-1], plt.ylim()[1]*0.95, ' Forecast Window', fontsize=10, color='gray')
     plt.xlabel("Date")
     plt.ylabel("Price ($/MWh)")
     plt.title(f"ERCOT Electricity Price Forecast - {prediction_length} Day Test Window")
@@ -169,5 +173,14 @@ if time_series_data:
         st.dataframe(comparison_df, use_container_width=True)
 
 # Note for comments, feedback, or questions
-st.write("### Notes")
-st.write("For comments, feedback, or any questions, please reach out to me on [LinkedIn](https://www.linkedin.com/in/javadbayazi/).")
+st.write("### About")
+st.write("""
+**Features:**
+- 🤖 Multiple Chronos models (Chronos-2 and T5 variants)
+- 📊 Real-time ERCOT electricity market data
+- 🎯 Backtesting with error metrics (MAE, RMSE, MAPE)
+- 📈 Visual comparison of forecasts vs actual values
+- 🔧 Modular architecture for easy extension
+
+For questions or feedback, reach out on [LinkedIn](https://www.linkedin.com/in/javadbayazi/).
+""")