Add correlation charts

page_attitudes.py

@@ -1,8 +1,12 @@
 import streamlit as st
 import matplotlib.pyplot as plt
 import seaborn as sns
+import networkx as nx
+import pandas as pd
+import numpy as np
 from matplotlib.font_manager import FontProperties
 from fields.likert_fields import likert_fields
+from fields.likert_flat_fields import likert_flat_fields
 from fields.field_translation_mapping import field_translation_mapping
 from fields.translation_mapping import translation_mapping
 
@@ -15,6 +19,14 @@ def show(df):
     # Chinese font
     chinese_font = FontProperties(fname='mingliu.ttf')
 
+    #generate_correlation_chart(df, chinese_font)
+    create_likert_charts(df, chinese_font)
+
+    st.title("Correlations Between Fields")
+    create_correlation_network(df, 0.4, chinese_font)
+    
+
+def create_likert_charts(df, chinese_font):
     if df is not None:
 
         # Rename the columns in the DataFrame for visualization
@@ -61,3 +73,79 @@ def show(df):
 
             # Show the plot
             st.pyplot(fig)
+
+def create_correlation_network(df, threshold, chinese_font):
+    
+    filtered_df = df[likert_flat_fields]
+    filtered_df = filtered_df.apply(pd.to_numeric, errors='coerce')
+
+    # Now you can calculate the correlation matrix and create the network
+    corr_matrix = filtered_df.corr()
+
+    # Create a graph
+    graph = nx.Graph()
+
+    # Iterate over the correlation matrix and add edges
+    for i in range(len(corr_matrix.columns)):
+        for j in range(i):
+            if abs(corr_matrix.iloc[i, j]) > threshold:  # only consider strong correlations
+                graph.add_edge(corr_matrix.columns[i], corr_matrix.columns[j], weight=corr_matrix.iloc[i, j])
+
+    # Draw the network
+    pos = nx.spring_layout(graph, k=0.1, iterations=20)
+    edges = graph.edges()
+    weights = [graph[u][v]['weight'] for u, v in edges]  # Use the weights for edge width
+
+    plt.figure(figsize=(10, 10))
+    nx.draw_networkx_nodes(graph, pos, node_size=500, node_color='lightblue', edgecolors='black')
+    nx.draw_networkx_edges(graph, pos, edgelist=edges, width=weights, alpha=0.5, edge_color='gray')
+
+    # Set Chinese font
+    for label in graph.nodes():
+        x, y = pos[label]
+        plt.text(x, y, label, fontsize=9, fontproperties=chinese_font, ha='center', va='center')
+
+    plt.title('Correlation Network', fontproperties=chinese_font)
+    plt.axis('off')  # Turn off the axis
+
+    # Use Streamlit to render the plot
+    st.pyplot(plt)
+
+def generate_correlation_chart(df, chinese_font):
+
+    boolean_fields = [
+        '你/妳覺得目前有任何投資嗎？'
+    ]
+
+    # Encode boolean fields
+    for field in boolean_fields:
+        df[field + '_encoded'] = df[field].map({'有': 1, '沒有': 0})
+
+    # Combine all fields for correlation
+    all_fields = likert_flat_fields + [f"{field}_encoded" for field in boolean_fields]
+    
+    # Calculate the correlation matrix
+    correlation_data = df[all_fields].corr()
+    
+    # Define a threshold for strong correlations
+    threshold = 0.5
+    
+    # Find all fields that have at least one strong correlation
+    strong_fields = correlation_data.columns[np.abs(correlation_data).max() > threshold]
+
+    # Filter the correlation matrix to only include these fields
+    filtered_correlation_data = correlation_data.loc[strong_fields, strong_fields]
+    
+    # Plot the correlation matrix
+    plt.figure(figsize=(10, 8))
+    ax = sns.heatmap(filtered_correlation_data, annot=True, fmt=".2f", cmap="coolwarm")
+    
+    # Set the labels with the Chinese font
+    ax.set_xticklabels(ax.get_xticklabels(), fontproperties=chinese_font, rotation=45, ha='right')
+    ax.set_yticklabels(ax.get_yticklabels(), fontproperties=chinese_font, rotation=0)
+    
+    # Set the title with the Chinese font
+    plt.title("強相關分析", fontproperties=chinese_font)
+
+    # Show the plot in Streamlit
+    st.pyplot(plt)