app.py

import streamlit as st
import pandas as pd
import numpy as np
from sklearn.cluster import KMeans, DBSCAN, AgglomerativeClustering
from sklearn.metrics import silhouette_score
from sklearn.preprocessing import StandardScaler
from statsmodels.tsa.arima.model import ARIMA
import plotly.express as px
import plotly.graph_objects as go

# Streamlit app title
st.title('Clustering and Time Series Analysis')

# Step 1: Upload CSV file
uploaded_file = st.file_uploader("Upload a CSV file", type=["csv"])

if uploaded_file is not None:
    data = pd.read_csv(uploaded_file)
    st.write("Dataset Preview:", data.head())

    # Step 2: Data Preprocessing
    # Selecting only numerical columns for clustering
    numerical_cols = data.select_dtypes(include=[np.number]).columns.tolist()
    st.write("Numerical columns for clustering:", numerical_cols)

    # Step 2.1: Data Standardization using StandardScaler (always applied)
    scaler = StandardScaler()
    data_scaled = scaler.fit_transform(data[numerical_cols])
    st.write("Data has been standardized using StandardScaler.")

    # Step 3: Clustering Algorithm Selection
    clustering_method = st.selectbox("Choose a clustering method", ["K-Means", "Hierarchical Clustering", "DBSCAN"])

    if clustering_method == "K-Means":
        k_range = st.slider("Select number of clusters for K-Means", min_value=2, max_value=7, value=3)
        kmeans = KMeans(n_clusters=k_range, random_state=42)
        cluster_labels = kmeans.fit_predict(data_scaled)
        silhouette_avg = silhouette_score(data_scaled, cluster_labels)
        st.write(f"K-Means Silhouette Score for {k_range} clusters: {silhouette_avg}")

    elif clustering_method == "Hierarchical Clustering":
        k_range = st.slider("Select number of clusters for Hierarchical Clustering", min_value=2, max_value=7, value=3)
        hierarchical = AgglomerativeClustering(n_clusters=k_range)
        cluster_labels = hierarchical.fit_predict(data_scaled)
        silhouette_avg = silhouette_score(data_scaled, cluster_labels)
        st.write(f"Hierarchical Clustering Silhouette Score for {k_range} clusters: {silhouette_avg}")

    elif clustering_method == "DBSCAN":
        eps_value = st.slider("Select eps value for DBSCAN", min_value=0.1, max_value=2.0, value=0.5)
        min_samples_value = st.slider("Select minimum samples for DBSCAN", min_value=1, max_value=10, value=5)
        dbscan = DBSCAN(eps=eps_value, min_samples=min_samples_value)
        cluster_labels = dbscan.fit_predict(data_scaled)
        
        # Check if DBSCAN found valid clusters
        if len(set(cluster_labels)) > 1:
            silhouette_avg = silhouette_score(data_scaled, cluster_labels)
            st.write(f"DBSCAN Silhouette Score: {silhouette_avg}")
        else:
            st.write("DBSCAN did not form valid clusters. Try adjusting eps or min_samples.")

    # Step 4: Visualize the clusters using Plotly
    if len(set(cluster_labels)) > 1:
        st.write("Cluster Labels:", np.unique(cluster_labels))
        
        # Create Plotly scatter plot
        fig = px.scatter(x=data_scaled[:, 0], y=data_scaled[:, 1], color=cluster_labels, title="Clustering Results", 
                         labels={'x': numerical_cols[0], 'y': numerical_cols[1]})

        # Update y-axis range to be from -1 to 1 with 0.2 intervals
        fig.update_layout(
            yaxis=dict(range=[-1, 1], dtick=0.2),
            xaxis_title=numerical_cols[0],
            yaxis_title=numerical_cols[1]
        )

        st.plotly_chart(fig)

    # Step 5: ARIMA Time Series Analysis
    # Checking if there are any time-related columns
    time_series_col = None
    for col in data.columns:
        if pd.api.types.is_datetime64_any_dtype(data[col]):
            time_series_col = col
            break

    if time_series_col:
        st.write("Time Series Analysis (ARIMA) on column:", time_series_col)
        time_series_data = data[time_series_col].dropna()
        
        # ARIMA model order
        p = st.number_input("ARIMA p value", min_value=0, max_value=5, value=1)
        d = st.number_input("ARIMA d value", min_value=0, max_value=2, value=1)
        q = st.number_input("ARIMA q value", min_value=0, max_value=5, value=1)
        
        arima_model = ARIMA(time_series_data, order=(p, d, q))
        arima_result = arima_model.fit()
        
        # Display ARIMA result summary
        st.write(arima_result.summary())

        # Plotting the ARIMA results
        fig = go.Figure()
        arima_result.plot_predict(dynamic=False, ax=fig.add_subplot(1, 1, 1))
        st.plotly_chart(fig)

    # Step 6: Create Silhouette Score Table for K-Means and Hierarchical Clustering
    st.write("### Silhouette Score Table for 2-7 Clusters")
    silhouette_scores = {'Number of Clusters': [], 'K-Means Silhouette Score': [], 'Hierarchical Silhouette Score': []}

    for n_clusters in range(2, 8):
        # K-Means
        kmeans = KMeans(n_clusters=n_clusters, random_state=42)
        kmeans_labels = kmeans.fit_predict(data_scaled)
        kmeans_silhouette = silhouette_score(data_scaled, kmeans_labels)

        # Hierarchical
        hierarchical = AgglomerativeClustering(n_clusters=n_clusters)
        hierarchical_labels = hierarchical.fit_predict(data_scaled)
        hierarchical_silhouette = silhouette_score(data_scaled, hierarchical_labels)

        silhouette_scores['Number of Clusters'].append(n_clusters)
        silhouette_scores['K-Means Silhouette Score'].append(kmeans_silhouette)
        silhouette_scores['Hierarchical Silhouette Score'].append(hierarchical_silhouette)

    silhouette_df = pd.DataFrame(silhouette_scores)

    # Plot the Silhouette Score Table using Plotly
    fig = go.Figure()

    # Plot K-Means Silhouette Scores
    fig.add_trace(go.Scatter(x=silhouette_df['Number of Clusters'], y=silhouette_df['K-Means Silhouette Score'],
                             mode='lines+markers', name='K-Means Silhouette Score'))

    # Plot Hierarchical Silhouette Scores
    fig.add_trace(go.Scatter(x=silhouette_df['Number of Clusters'], y=silhouette_df['Hierarchical Silhouette Score'],
                             mode='lines+markers', name='Hierarchical Silhouette Score'))

    # Set the y-axis range from -1 to 1 with intervals of 0.2
    fig.update_layout(
        title="Silhouette Scores for K-Means and Hierarchical Clustering",
        xaxis_title="Number of Clusters",
        yaxis_title="Silhouette Score",
        yaxis=dict(range=[-1, 1], dtick=0.2)
    )

    st.plotly_chart(fig)