app.py

import streamlit as st
import pandas as pd
import matplotlib.pyplot as plt
import joblib
from sklearn.decomposition import PCA
from sklearn.preprocessing import StandardScaler
from sklearn.cluster import KMeans, DBSCAN
from scipy.cluster.hierarchy import fcluster, linkage

# 讀取保存的模型
scaler = joblib.load('scaler.sav')
pca = joblib.load('pca_model.sav')
kmeans = joblib.load('kmeans_model.sav')
linked = joblib.load('hierarchical_model.sav')
dbscan = joblib.load('dbscan_model.sav')

# 標題和簡介
st.title("聚類分析 - KMeans, Hierarchical Clustering 和 DBSCAN")
st.write("上傳 CSV 文件並查看聚類結果")

# 上傳文件
uploaded_file = st.file_uploader("上傳 CSV 文件", type=["csv"])

if uploaded_file is not None:
    # 讀取上傳的 CSV 文件
    data = pd.read_csv(uploaded_file)
    
    # 移除 'Time' 欄位
    numerical_data = data.drop(columns=['Time'])
    
    # 標準化數據
    scaled_data = scaler.transform(numerical_data)
    
    # 使用 PCA 進行降維
    pca_data = pca.transform(scaled_data)
    
    # 創建包含主成分的 DataFrame
    pca_df = pd.DataFrame(pca_data, columns=['PC1', 'PC2'])
    
    # 使用保存的 K-means 模型進行聚類
    kmeans_labels = kmeans.predict(pca_df)
    
    # 使用保存的Hierarchical Clustering 階層式聚類結果
    hclust_labels = fcluster(linked, 3, criterion='maxclust')
    
    # 使用保存的 DBSCAN 模型進行聚類
    dbscan_labels = dbscan.fit_predict(pca_df)
    
    # ================== 圖表選擇 ==================
    chart_option = st.selectbox(
        "選擇要顯示的聚類結果圖表",
        ("K-means", "Hierarchical Clustering", "DBSCAN")
    )
    
    # ================== 根據選擇顯示對應的圖表 ==================
    if chart_option == "K-means":
        st.subheader("K-means_聚類結果")
        fig_kmeans, ax_kmeans = plt.subplots()
        ax_kmeans.scatter(pca_df['PC1'], pca_df['PC2'], c=kmeans_labels, cmap='viridis')
        ax_kmeans.set_title('K-means Clustering')
        ax_kmeans.set_xlabel('PC1')
        ax_kmeans.set_ylabel('PC2')
        st.pyplot(fig_kmeans)
    
    elif chart_option == "Hierarchical Clustering":
        st.subheader("Hierarchical Clustering_階層式聚類結果")
        fig_hclust, ax_hclust = plt.subplots()
        ax_hclust.scatter(pca_df['PC1'], pca_df['PC2'], c=hclust_labels, cmap='viridis')
        ax_hclust.set_title('Hierarchical Clustering')
        ax_hclust.set_xlabel('PC1')
        ax_hclust.set_ylabel('PC2')
        st.pyplot(fig_hclust)
    
    elif chart_option == "DBSCAN":
        st.subheader("DBSCAN_聚類結果")
        fig_dbscan, ax_dbscan = plt.subplots()
        ax_dbscan.scatter(pca_df['PC1'], pca_df['PC2'], c=dbscan_labels, cmap='viridis')
        ax_dbscan.set_title('DBSCAN Clustering')
        ax_dbscan.set_xlabel('PC1')
        ax_dbscan.set_ylabel('PC2')
        st.pyplot(fig_dbscan)