Update app.py

app.py

@@ -4,8 +4,7 @@ import numpy as np
 import os
 import matplotlib.pyplot as plt
 import seaborn as sns
-from sklearn.cluster import KMeans
-from statsmodels.tsa.holtwinters import ExponentialSmoothing
+from pycaret.regression import setup, compare_models, predict_model, pull
 from langchain_openai import ChatOpenAI
 from dotenv import load_dotenv
 
@@ -38,7 +37,7 @@ for mall, file in competitor_files.items():
     competitor_data[mall] = pd.read_csv(file, encoding="ISO-8859-1")
 
 # ✅ Streamlit Page Configuration
-st.set_page_config(page_title="AI Store Replacement, Demand Prediction & Chatbot", layout="wide")
+st.set_page_config(page_title="AI Store Replacement with PyCaret KPI Forecasting", layout="wide")
 
 # ✅ Title
 st.title("🔄 AI-Powered Store Replacement, Demand Prediction & Chatbot")
@@ -61,7 +60,7 @@ if selected_mall:
     store_to_replace = st.selectbox("🚪 Select the store that is closing:", mall_data["Store_Brand"].unique())
 
     if store_to_replace:
-        llm = ChatOpenAI(model="gpt-3.5-turbo", api_key=openai_api_key)
+        llm = ChatOpenAI(model="gpt-4", api_key=openai_api_key)
 
         # ✅ Identify Store Category
         category_prompt = f"Given that {store_to_replace} is a retail brand, what type of store is it? Categorize it broadly (e.g., Fashion, Electronics, Footwear, Food, Home Decor, etc.)."
@@ -80,55 +79,45 @@ if selected_mall:
             for competitor in competitor_list:
                 st.write(f"- {competitor}")
 
-        # ✅ AI-Based Store Clustering (Finding Best Replacements)
-        store_features = df[["Annual_Sales", "Footfall", "Gross_Margin"]].dropna()
-        kmeans = KMeans(n_clusters=5, random_state=42)  # 5 Clusters
-        df["Cluster"] = kmeans.fit_predict(store_features)
+        # ✅ Train PyCaret Model for KPI Prediction
+        st.subheader("📈 Training PyCaret Model for KPI Forecasting")
 
-        # ✅ Find replacement store from `kpi.xlsx` (Not in this mall)
-        available_stores = list(set(df["Store_Brand"].unique()) - set(mall_data["Store_Brand"].unique()))
+        kpi_data = df.drop(columns=["Store_Brand", "Mall", "Location", "Country", "City"])  # Exclude categorical features
+        reg = setup(kpi_data, target="Annual_Sales", silent=True)
+        best_model = compare_models()
+        st.write("🏆 Best Model Selected:", best_model)
 
-        # ✅ Debugging: Check available replacements
-        st.write(f"🔍 **Available Replacement Stores:** {available_stores}")
+        # ✅ Find a replacement store in the same category (From kpi.xlsx, Not in this mall)
+        available_stores = list(set(df["Store_Brand"].unique()) - set(mall_data["Store_Brand"].unique()))
 
         if available_stores:
             replacement_prompt = f"Suggest 3 replacement stores from the list below that are not already in {selected_mall}. The stores should match the mall’s customer base and the category: {store_category}. Store List: {available_stores}"
             replacement_response = llm.invoke(replacement_prompt).content
-
-            # ✅ Debugging: Print AI Response
-            st.write(f"🧠 **AI Raw Response:** {replacement_response}")
-
-            # ✅ Ensure AI Response is Valid
             replacement_stores = [store.strip() for store in replacement_response.split("\n") if store.strip()]
             replacement_stores = [store for store in replacement_stores if store in available_stores]
 
             if replacement_stores:
-                st.success(f"🔄 **Recommended Replacement Stores:**")
+                st.success(f"🔄 **Recommended Replacement Stores with Forecasted KPIs:**")
                 for replacement_store in replacement_stores:
                     store_data = df[df["Store_Brand"] == replacement_store]
                     
                     if not store_data.empty:
-                        sales_history = store_data["Annual_Sales"]
-                        if len(sales_history) > 3:
-                            # ✅ Time-Series Forecasting for Demand Prediction
-                            model = ExponentialSmoothing(sales_history, trend="add", seasonal="add", seasonal_periods=3)
-                            fit = model.fit()
-                            forecast_sales = fit.forecast(1).values[0]  # Predict next year's sales
-
-                            expected_footfall = store_data["Footfall"].mean()
-                            expected_conversion = store_data["Conversion_Rate"].mean()
-                            expected_margin = store_data["Gross_Margin"].mean()
-
-                            st.write(f"**{replacement_store}**")
-                            st.write(f"📊 **Projected Sales:** AED {forecast_sales:,.2f}")
-                            st.write(f"👣 **Projected Footfall:** {expected_footfall:,.0f} visitors")
-                            st.write(f"💰 **Projected Margin:** AED {expected_margin:,.2f}")
-                            st.write(f"🔄 **Projected Conversion Rate:** {expected_conversion:.2f}%")
-
-                            # ✅ Justification
-                            justification_prompt = f"Explain why {replacement_store} is a good replacement for {store_to_replace} in {selected_mall}. Include numeric KPIs: expected sales, footfall, conversion rate, and profit margins."
-                            justification = llm.invoke(justification_prompt).content
-                            st.write(f"🤖 **AI Justification:** {justification}")
+                        forecast_kpis = predict_model(best_model, data=store_data)
+                        forecasted_sales = forecast_kpis["Label"].values[0]
+                        expected_footfall = store_data["Footfall"].mean()
+                        expected_conversion = store_data["Conversion_Rate"].mean()
+                        expected_margin = store_data["Gross_Margin"].mean()
+
+                        st.write(f"**{replacement_store}**")
+                        st.write(f"📊 **Projected Sales:** AED {forecasted_sales:,.2f}")
+                        st.write(f"👣 **Projected Footfall:** {expected_footfall:,.0f} visitors")
+                        st.write(f"💰 **Projected Margin:** AED {expected_margin:,.2f}")
+                        st.write(f"🔄 **Projected Conversion Rate:** {expected_conversion:.2f}%")
+
+                        # ✅ Justification
+                        justification_prompt = f"Explain why {replacement_store} is a good replacement for {store_to_replace} in {selected_mall}. Include numeric KPIs: expected sales, footfall, conversion rate, and profit margins."
+                        justification = llm.invoke(justification_prompt).content
+                        st.write(f"🤖 **AI Justification:** {justification}")
 
 # ✅ AI Chatbot for Dynamic Insights
 st.subheader("💬 AI-Powered Chatbot")