Market basket recommendation system using Apriori

market-basket-recommendation/app.py

@@ -0,0 +1,100 @@
+import pandas as pd
+import streamlit as st
+from apyori import apriori
+
+df=pd.read_csv("data.csv")
+
+transactions = []
+for i in range(0, 7218):
+  transactions.append([str(df.values[i,j]) for j in range(0, 20)])
+
+
+rules = apriori(
+    transactions=transactions,
+    min_support=0.005,
+    min_confidence=0.1,
+    min_lift=3,
+    min_length=2,
+    max_length=2
+)
+res=list(rules)
+
+
+# Define inspect function
+def inspect(results):
+    product1         = [tuple(result[2][0][0])[0] for result in results]
+    product2         = [tuple(result[2][0][1])[0] for result in results]
+    supports    = [result[1] for result in results]
+    confidences = [result[2][0][2] for result in results]
+    lifts       = [result[2][0][3] for result in results]
+    return list(zip(product1, product2, supports, confidences, lifts))
+DataFrame_intelligence = pd.DataFrame(inspect(res), columns = ['product1', 'product2', 'Support', 'Confidence', 'Lift'])
+
+
+
+# Define get_recommendations function
+def get_recommendations(user_item, rules_df, confidence_threshold=0.2, lift_threshold=1.0):
+    related_items = []
+    for _, row in rules_df.iterrows():
+        if user_item == row['product1'] and row['Confidence'] >= confidence_threshold and row['Lift'] >= lift_threshold:
+            related_items.append((row['product2'], row['Confidence'], row['Lift']))
+        elif user_item == row['product2'] and row['Confidence'] >= confidence_threshold and row['Lift'] >= lift_threshold:
+            related_items.append((row['product1'], row['Confidence'], row['Lift']))
+
+    if related_items:
+        related_items.sort(key=lambda x: (x[1], x[2]), reverse=True)
+        top_recommendations = related_items[:3]
+        return top_recommendations
+    else:
+        return []
+
+# Define display_recommendations function
+def display_recommendations(user_item):
+    recommendations = get_recommendations(user_item, DataFrame_intelligence)
+
+    if recommendations:
+        st.write("\nYou may also need:")
+        for item, _, _ in recommendations:  # Ignoring confidence and lift values
+            st.write(f"{item} - because customers who bought {user_item} also bought {item}.")
+    else:
+        st.write("No strong recommendations found for the item you entered.")
+
+        new_items = st.text_input(f"What would you like to buy along with {user_item}? (Enter items separated by commas): ",key="new_items").strip().split(',')
+        new_items = [item.strip() for item in new_items]
+        add_new_association(user_item, new_items)
+        # Display recommendations after adding new 
+        display_recommendations(user_item)
+
+
+# Define add_new_association function
+def add_new_association(item1, item2):
+    global DataFrame_intelligence
+    new_row = pd.DataFrame({
+        'product1': [item1],
+        'product2': [item2],
+        'Support': [0.001],  
+        'Confidence': [0.5],  
+        'Lift': [2.0]  
+    })
+    DataFrame_intelligence = pd.concat([DataFrame_intelligence, new_row], ignore_index=True)
+    st.write(f"New association added: {item1} -> {item2}")
+
+# Main function for Streamlit app
+def main():
+    st.title("Recommendation System")
+    st.write("This is a recommendation system.")
+
+    while True:
+        user_item = st.text_input("Enter an item (or type 'exit' to quit): ", key=f"user_item_{st.session_state['iteration_count']}", value='', help='Enter item here')
+        if user_item.lower() == 'exit' or not user_item:
+            break
+        display_recommendations(user_item)
+        st.session_state['iteration_count'] += 1
+
+    st.write("Thank you for using the recommendation system!")
+
+if __name__ == "__main__":
+    st.session_state['iteration_count'] = 0
+    main()
+
+