app.py

import streamlit as st
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.cluster import KMeans
from sklearn.preprocessing import StandardScaler, OneHotEncoder, LabelEncoder
from sklearn.pipeline import Pipeline
from sklearn.metrics import silhouette_score, davies_bouldin_score, calinski_harabasz_score

st.title("Sales Trend Prediction using KMeans Clustering")

def load_data():
    return pd.read_csv("shopping_trends.csv")

df = load_data()

# Select relevant features for clustering
features = ['Gender', 'Item Purchased', 'Previous Purchases', 'Frequency of Purchases', 'Purchase Amount (USD)']
df_filtered = df[features].copy()

# Convert Frequency of Purchases to string
df_filtered['Frequency of Purchases'] = df_filtered['Frequency of Purchases'].astype(str)

# One-hot encode categorical features
categorical_features = ['Gender', 'Item Purchased', 'Frequency of Purchases']
numerical_features = ['Previous Purchases', 'Purchase Amount (USD)']

ohe = OneHotEncoder(drop='first', sparse_output=False)
encoded_cats = ohe.fit_transform(df_filtered[categorical_features])
categorical_df = pd.DataFrame(encoded_cats, columns=ohe.get_feature_names_out(categorical_features), index=df.index)

df_processed = pd.concat([df_filtered[numerical_features], categorical_df], axis=1)

# Standardizing the data
scaler = StandardScaler()
X_scaled = scaler.fit_transform(df_processed)

# KMeans Clustering
n_clusters = 3  # Set the number of clusters
kmeans = KMeans(n_clusters=n_clusters, random_state=42)
df['Cluster'] = kmeans.fit_predict(X_scaled)

# Compute Clustering Metrics
silhouette = silhouette_score(X_scaled, df['Cluster'])
davies_bouldin = davies_bouldin_score(X_scaled, df['Cluster'])
calinski_harabasz = calinski_harabasz_score(X_scaled, df['Cluster'])

def predict_cluster(user_input):
    """Predicts the cluster for a new user input."""
    user_df = pd.DataFrame([user_input])
    user_df['Frequency of Purchases'] = user_df['Frequency of Purchases'].astype(str)
    user_cats = ohe.transform(user_df[categorical_features])
    user_processed = pd.concat([user_df[numerical_features], pd.DataFrame(user_cats, columns=ohe.get_feature_names_out(categorical_features))], axis=1)
    user_scaled = scaler.transform(user_processed)
    return kmeans.predict(user_scaled)[0]

# Create Tabs
tab1, tab2, tab3 = st.tabs(["Dataset & Metrics", "Visualization", "Sales Trend Prediction"])

with tab1:
    st.subheader("Dataset Preview")
    st.write(df.head())
    
    st.subheader("Clustering Metrics")
    st.write(f"Number of Clusters: {n_clusters}")
    st.write(f"Silhouette Score: {silhouette:.4f}")
    st.write(f"Davies-Bouldin Score: {davies_bouldin:.4f}")
    st.write(f"Calinski-Harabasz Score: {calinski_harabasz:.4f}")

with tab2:
    st.subheader("Data Visualizations")
    
    # Elbow Method Visualization
    distortions = []
    K_range = range(2, 11)
    for k in K_range:
        kmeans_tmp = KMeans(n_clusters=k, random_state=42)
        kmeans_tmp.fit(X_scaled)
        distortions.append(kmeans_tmp.inertia_)
    
    fig, ax = plt.subplots()
    ax.plot(K_range, distortions, marker='o')
    ax.set_xlabel('Number of Clusters')
    ax.set_ylabel('Distortion')
    ax.set_title('Elbow Method for Optimal K')
    st.pyplot(fig)
    
    # Cluster Distribution Plot
    fig, ax = plt.subplots()
    sns.countplot(x=df['Cluster'], palette='viridis', ax=ax)
    ax.set_xlabel('Cluster')
    ax.set_ylabel('Count')
    ax.set_title('Cluster Distribution')
    st.pyplot(fig)
    
    # Visualizations for Item Purchased distribution
    fig, ax = plt.subplots(figsize=(10, 5))
    sns.countplot(y=df['Item Purchased'], order=df['Item Purchased'].value_counts().index, hue=df['Item Purchased'], palette='viridis', ax=ax)
    ax.set_title("Overall Item Purchase Distribution")
    ax.set_xlabel("Count")
    ax.set_ylabel("Item Purchased")
    st.pyplot(fig)
    
    # Separate Correlation Matrices for Each Label
    labels = ['Gender', 'Frequency of Purchases', 'Item Purchased']
    for label in labels:
        df[f'{label}_Numeric'] = LabelEncoder().fit_transform(df[label])
        correlation_matrix = df[[f'{label}_Numeric', 'Purchase Amount (USD)', 'Previous Purchases']].corr()
        fig, ax = plt.subplots(figsize=(10, 6))
        sns.heatmap(correlation_matrix, annot=True, cmap='coolwarm', fmt='.2f', ax=ax)
        st.subheader(f"Feature Correlation Matrix for {label}")
        st.pyplot(fig)

with tab3:
    st.subheader("Enter Customer Details")
    gender = st.selectbox("Gender", df['Gender'].unique())
    item_purchased = st.selectbox("Item Purchased", df['Item Purchased'].unique())
    previous_purchases = st.number_input("Previous Purchases", min_value=0, max_value=100, value=10)
    frequency_of_purchases = st.selectbox("Frequency of Purchases", df['Frequency of Purchases'].unique().astype(str))
    purchase_amount = st.number_input("Purchase Amount (USD)", min_value=1, max_value=500, value=50)

    if st.button("Predict Sales Trend"):
        user_input = {
            'Gender': gender,
            'Item Purchased': item_purchased,
            'Previous Purchases': previous_purchases,
            'Frequency of Purchases': frequency_of_purchases,
            'Purchase Amount (USD)': purchase_amount
        }
        
        predicted_cluster = predict_cluster(user_input)
        st.write(f"Predicted Cluster: {predicted_cluster}")
        
        st.subheader(f"Sales Trend Analysis for Cluster {predicted_cluster}")
        cluster_data = df[df['Cluster'] == predicted_cluster]
        
        # Visualization of top-selling items in the cluster
        fig, ax = plt.subplots(figsize=(10, 5))
        sns.countplot(y=cluster_data['Item Purchased'], order=cluster_data['Item Purchased'].value_counts().index, palette='viridis', ax=ax)
        ax.set_title("Top Selling Items in This Cluster")
        ax.set_xlabel("Count")
        ax.set_ylabel("Item Purchased")
        st.pyplot(fig)
        
        # Display average purchase amount trend
        avg_purchase_amount = cluster_data.groupby('Item Purchased')['Purchase Amount (USD)'].mean()
        st.write("### Average Purchase Amount for Items in This Cluster:")
        st.write(avg_purchase_amount)