Update app.py

app.py

@@ -1,29 +1,42 @@
 import streamlit as st
 import pandas as pd
+import pickle
 from prophet import Prophet
 from datetime import datetime, timedelta
 import numpy as np
+import plotly.graph_objects as go
+
+# Load the trained models (optional, for initialization or fallback)
+@st.cache_resource
+def load_model():
+    try:
+        with open('model.pkl', 'rb') as f:
+            models = pickle.load(f)
+        return models
+    except FileNotFoundError:
+        return None
 
 # Prepare data for Prophet
 def prepare_prophet_data(usage_series):
     end_date = datetime.now()
     start_date = end_date - timedelta(days=len(usage_series) - 1)
     dates = [start_date + timedelta(days=i) for i in range(len(usage_series))]
+    
     prophet_df = pd.DataFrame({
         'ds': dates,
         'y': usage_series
     })
     return prophet_df
 
-# Train or update Prophet model with user-provided usage series
+# Train or update Prophet model with reduced sensitivity
 def train_model_with_usage(usage_series):
-    print("Training with changepoint_prior_scale=0.01, usage:", usage_series)  # Debug
     prophet_df = prepare_prophet_data(usage_series)
     model = Prophet(
         yearly_seasonality=False,
         weekly_seasonality=True,
         daily_seasonality=True,
-        changepoint_prior_scale=0.01
+        changepoint_prior_scale=0.01,   # 👈 Reduced sensitivity to recent changes
+        changepoint_range=0.6           # 👈 Only look at early 60% of data for trend changes
     )
     model.fit(prophet_df)
     return model
@@ -32,9 +45,7 @@ def train_model_with_usage(usage_series):
 def make_forecast(model, periods):
     future = model.make_future_dataframe(periods=periods)
     forecast = model.predict(future)
-    daily_forecasts = forecast['yhat'].tail(periods).tolist()
-    print(f"Daily forecasts for {periods} days:", [round(y) for y in daily_forecasts])  # Debug
-    return round(sum(max(0, y) for y in daily_forecasts))  # Clip negative values
+    return round(forecast['yhat'].tail(periods).sum())
 
 # Function to validate input
 def validate_usage_series(usage_str):
@@ -51,9 +62,8 @@ def validate_usage_series(usage_str):
 # Main Streamlit app
 def main():
     st.title("SmartLab Consumables Forecast")
-    if st.button("Clear Cache"):
-        st.cache_resource.clear()
-        st.write("Cache cleared!")
+    
+    models = load_model()
     
     st.header("Input Parameters")
     consumable_type = st.selectbox("Consumable Type", ['Filters', 'Reagents', 'Vials'])
@@ -65,7 +75,6 @@ def main():
         if error:
             st.error(error)
             return
-        st.write("Debug: Input usage series:", usage_list)  # Debug
         
         try:
             model = train_model_with_usage(usage_list)
@@ -85,25 +94,67 @@ def main():
         st.header("Threshold Alerts")
         if current_stock < forecast_7:
             st.warning(f"Alert: Current stock ({current_stock}) is below 7-day forecast ({forecast_7}). 🚩")
-        else:
-            st.write("No alert for 7-day forecast.")
         if current_stock < forecast_14:
             st.warning(f"Alert: Current stock ({current_stock}) is below 14-day forecast ({forecast_14}). 🚩")
-        else:
-            st.write("No alert for 14-day forecast.")
         if current_stock < forecast_30:
             st.warning(f"Alert: Current stock ({current_stock}) is below 30-day forecast ({forecast_30}). 🚩")
-        else:
-            st.write("No alert for 30-day forecast.")
         
         st.header("Order Suggestions")
-        order_7 = max(0, round(forecast_7 - current_stock))
-        order_14 = max(0, round(forecast_14 - current_stock))
-        order_30 = max(0, round(forecast_30 - current_stock))
-        
+        order_7 = max(0, forecast_7 - current_stock)
+        order_14 = max(0, forecast_14 - current_stock)
+        order_30 = max(0, forecast_30 - current_stock)
         st.write(f"**For 7 Days**: Order {order_7} additional units.")
         st.write(f"**For 14 Days**: Order {order_14} additional units.")
         st.write(f"**For 30 Days**: Order {order_30} additional units.")
+        
+        st.header("Forecast Visualization")
+        forecast_data = pd.DataFrame({
+            'Period': ['7 Days', '14 Days', '30 Days'],
+            'Units': [forecast_7, forecast_14, forecast_30]
+        })
+        fig_forecast = go.Figure()
+        fig_forecast.add_trace(go.Scatter(
+            x=forecast_data['Period'],
+            y=forecast_data['Units'],
+            mode='lines+markers',
+            name='Forecasted Units',
+            line=dict(color='blue'),
+            marker=dict(size=10)
+        ))
+        fig_forecast.update_layout(
+            title='Consumable Usage Forecast',
+            xaxis_title='Time Period',
+            yaxis_title='Units',
+            template='plotly_white'
+        )
+        st.plotly_chart(fig_forecast)
+        
+        st.header("Threshold Alerts Visualization")
+        alert_data = pd.DataFrame({
+            'Category': ['Current Stock', '7-Day Forecast', '14-Day Forecast', '30-Day Forecast'],
+            'Units': [current_stock, forecast_7, forecast_14, forecast_30],
+            'Alert': [
+                False,
+                current_stock < forecast_7,
+                current_stock < forecast_14,
+                current_stock < forecast_30
+            ]
+        })
+        fig_alerts = go.Figure()
+        fig_alerts.add_trace(go.Bar(
+            x=alert_data['Category'],
+            y=alert_data['Units'],
+            marker_color=['green'] + ['red' if alert else 'blue' for alert in alert_data['Alert'][1:]],
+            text=[f"🚩" if alert else "" for alert in alert_data['Alert']],
+            textposition='auto'
+        ))
+        fig_alerts.update_layout(
+            title='Stock vs Forecast with Alerts (🚩 indicates low stock)',
+            xaxis_title='Category',
+            yaxis_title='Units',
+            template='plotly_white'
+        )
+        st.plotly_chart(fig_alerts)
 
 if __name__ == "__main__":
-    main()
+    main()