Update app.py

app.py

@@ -1,63 +1,81 @@
-import joblib
-import gradio as gr
-import numpy as np
-from datetime import datetime, timedelta
-
-# Load the saved model and encoders
-model = joblib.load("best_model.pkl")
-le_operation_type = joblib.load("le_operation_type.pkl")
-le_operator_id = joblib.load("le_operator_id.pkl")
-
-def predict_end_date(start_date_str, quantity, operation_type, operator_id, day_of_week):
-    try:
-        start_date = datetime.strptime(start_date_str, "%d-%m-%Y")
-    except ValueError:
-        return "Error: Start Date must be in DD-MM-YYYY format"
-
-    # Encode operation type and operator id using the saved encoders
-    try:
-        operation_code = le_operation_type.transform([operation_type])[0]
-    except ValueError:
-        return f"Unknown Operation Type: {operation_type}"
-
-    try:
-        operator_code = le_operator_id.transform([operator_id])[0]
-    except ValueError:
-        return f"Unknown Operator ID: {operator_id}"
-
-    X = np.array([[quantity, operation_code, operator_code, day_of_week]])
-    pred_days = model.predict(X)[0]
-
-    whole_days = int(pred_days)
-    fractional_day = pred_days - whole_days
-
-    total_seconds = fractional_day * 24 * 3600
-    hours = int(total_seconds // 3600)
-    minutes = int((total_seconds % 3600) // 60)
-
-    est_end_datetime = start_date + timedelta(days=whole_days, hours=hours, minutes=minutes)
-    est_end_date_str = est_end_datetime.strftime("%d-%m-%Y %H:%M")
-    duration_str = f"{whole_days} days, {hours} hours, {minutes} minutes"
-
-    return f"Estimated End Date & Time: {est_end_date_str}\nEstimated Duration: {duration_str}"
-
-# Get all options for dropdowns from the encoders
-operation_types = le_operation_type.classes_.tolist()
-operator_ids = le_operator_id.classes_.tolist()
-
-iface = gr.Interface(
-    fn=predict_end_date,
-    inputs=[
-        gr.Textbox(label="Start Date (DD-MM-YYYY)", value="01-01-2025"),
-        gr.Number(label="Quantity", value=1),
-        gr.Dropdown(operation_types, label="Operation Type"),
-        gr.Dropdown(operator_ids, label="Operator ID"),
-        gr.Slider(0, 6, step=1, label="Day of Week (0=Monday, 6=Sunday)", value=0),
-    ],
-    outputs="text",
-    title="Manufacturing Production Scheduling",
-    description="Predict estimated end date and duration from production inputs."
-)
-
-if __name__ == "__main__":
-    iface.launch()
+import joblib
+import gradio as gr
+import numpy as np
+from datetime import datetime, timedelta
+
+# Load the saved model and encoders
+model = joblib.load("best_model.pkl")
+le_operation_type = joblib.load("le_operation_type.pkl")
+le_operator_id = joblib.load("le_operator_id.pkl")
+
+def predict_end_date(start_date, quantity, operation_code, operator_code, day_of_week):
+    # Recreate engineered features
+    quantity_log = np.log1p(quantity)
+    quantity_log_squared = quantity_log ** 2
+    quantity_log_cubed = quantity_log ** 3
+    quantity_operation_interaction = quantity_log * operation_code
+    quantity_operator_interaction = quantity_log * operator_code
+    quantity_operator_interaction_squared = quantity_operator_interaction ** 2
+    operation_operator_interaction = operation_code * operator_code
+
+    # Reconstruct feature vector with same structure as training
+    X = pd.DataFrame([[
+        quantity_log,
+        quantity_log_squared,
+        quantity_log_cubed,
+        quantity_operation_interaction,
+        quantity_operator_interaction,
+        quantity_operator_interaction_squared,
+        operation_operator_interaction,
+        operation_code,
+        operator_code,
+        day_of_week
+    ]], columns=[
+        'Quantity_log',
+        'Quantity_log_squared',
+        'Quantity_log_cubed',
+        'Quantity_Operation_interaction',
+        'Quantity_Operator_interaction',
+        'Quantity_Operator_interaction_squared',
+        'Operation_Operator_interaction',
+        'Operation_Code',
+        'Operator_Code',
+        'DayOfWeek'
+    ])
+
+    # Predict
+    pred_days = model.predict(X)[0]
+
+    # Convert predicted days to end date
+    start_dt = datetime.strptime(start_date, "%d-%m-%Y")
+    end_dt = start_dt + timedelta(days=pred_days)
+
+    # Split days into days + hours + minutes
+    total_minutes = pred_days * 24 * 60
+    days = int(total_minutes // (24 * 60))
+    hours = int((total_minutes % (24 * 60)) // 60)
+    minutes = int(total_minutes % 60)
+
+    return f"Estimated end date: {end_dt.strftime('%d-%m-%Y')} ({days} days {hours} hours {minutes} minutes)"
+
+
+# Get all options for dropdowns from the encoders
+operation_types = le_operation_type.classes_.tolist()
+operator_ids = le_operator_id.classes_.tolist()
+
+iface = gr.Interface(
+    fn=predict_end_date,
+    inputs=[
+        gr.Textbox(label="Start Date (DD-MM-YYYY)", value="01-01-2025"),
+        gr.Number(label="Quantity", value=1),
+        gr.Dropdown(operation_types, label="Operation Type"),
+        gr.Dropdown(operator_ids, label="Operator ID"),
+        gr.Slider(0, 6, step=1, label="Day of Week (0=Monday, 6=Sunday)", value=0),
+    ],
+    outputs="text",
+    title="Manufacturing Production Scheduling",
+    description="Predict estimated end date and duration from production inputs."
+)
+
+if __name__ == "__main__":
+    iface.launch()