Visualize shopping data

page_shopping.py

@@ -4,7 +4,6 @@ import matplotlib.pyplot as plt
 import seaborn as sns
 import pandas as pd
 import numpy as np
-import networkx as nx
 from fields.likert_flat_fields import likert_flat_fields
 #from fields.boolean_fields import boolean_fields
 
@@ -15,11 +14,19 @@ def show(df):
     st.title("Shopping")
     st.markdown(
                 f"<h2 style='text-align: center;'>Boycott Count (Overall)</h2>", unsafe_allow_html=True)
-    show_boycott_count(df, font_prop=chinese_font)
-    #generate_correlation_chart(df, chinese_font)
-    create_correlation_network(df, 0.4, chinese_font)
+    show_boycott_count(df, chinese_font)
+    st.markdown(
+                f"<h2 style='text-align: center;'>Why Boycott</h2>", unsafe_allow_html=True)
+    summarize_why_boycott(df, chinese_font)
+    st.markdown(
+                f"<h2 style='text-align: center;'>Trusted Brands</h2>", unsafe_allow_html=True)
+    summarize_trusted_brands(df, chinese_font)
+    st.markdown(
+                f"<h2 style='text-align: center;'>Choice Experiments</h2>", unsafe_allow_html=True)
+    visualize_shopping_data(df, chinese_font)
 
-def show_boycott_count(df, font_prop):
+
+def show_boycott_count(df, chinese_font):
     # Count the number of people who have invested and who have not
     boycott_count = df["你/妳有沒有抵制過某公司？"].value_counts().reset_index()
     boycott_count.columns = ['Boycott', 'Count']
@@ -28,92 +35,87 @@ def show_boycott_count(df, font_prop):
     plt.figure(figsize=(10, 6))
     barplot = sns.barplot(x='Boycott', y='Count', data=boycott_count, palette='viridis')
     ax = plt.gca()  # Get the current Axes instance on the current figure matching the given keyword args, or create one.
-    ax.set_xticklabels(ax.get_xticklabels(), fontproperties=font_prop)
+    ax.set_xticklabels(ax.get_xticklabels(), fontproperties=chinese_font)
 
     # Add labels and title
-    plt.xlabel('Have you ever boycotted a company?', fontsize=12, fontproperties=font_prop)
-    plt.ylabel('Count', fontsize=12, fontproperties=font_prop)
-    plt.title("Number of People Who Have/Haven't Boycotted a Company", fontsize=16, fontproperties=font_prop)
+    plt.xlabel('Have you ever boycotted a company?', fontsize=12, fontproperties=chinese_font)
+    plt.ylabel('Count', fontsize=12, fontproperties=chinese_font)
+    plt.title("Number of People Who Have/Haven't Boycotted a Company", fontsize=16, fontproperties=chinese_font)
 
     # Display values on the bars
     for index, value in enumerate(boycott_count['Count']):
-        plt.text(index, value, str(value), ha='center', va='bottom', fontproperties=font_prop)
+        plt.text(index, value, str(value), ha='center', va='bottom', fontproperties=chinese_font)
 
     # Display the chart in Streamlit
     st.pyplot(plt)
 
-def create_correlation_network(df, threshold, chinese_font):
-    
-    filtered_df = df[likert_flat_fields]
-    filtered_df = filtered_df.apply(pd.to_numeric, errors='coerce')
+def summarize_why_boycott(df, chinese_font):
 
-    # Now you can calculate the correlation matrix and create the network
-    corr_matrix = filtered_df.corr()
+    boycott_reasons = df["為什麼抵制？"].value_counts()
+    summary = boycott_reasons.sort_values(ascending=False)
 
-    # Create a graph
-    graph = nx.Graph()
+    st.write("Summary of Why Boycott:")
+    st.table(summary)
 
-    # 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
+def summarize_trusted_brands(df, chinese_font):
+    # Get the count of responses in the "你/妳有信任的品牌嗎？" field
+    trusted_brands = df["你/妳有信任的品牌嗎？"].value_counts()
 
-    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')
+    # List of responses to combine as 'no trusted brand'
+    no_brand_responses = ["無", "沒有", "沒有特別", "🈚️", "目前沒有", "No", "沒", "沒有特別關注", "沒有特別信任的", "不知道", "無特別選擇", "目前沒有完全信任的", "沒有特定的", "沒有特定", "沒有特別研究", "目前沒有特別關注的品牌","N", "none", "無特別", "目前無", "沒有特別想到", "沒有固定的", "x", "沒在買", "nope", "一時想不到…", "沒有特別注意", "無特別的品牌", "無絕對信任的品牌", "不確定你說的範圍", "還沒有"]
 
-    # 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')
+    # Calculate the combined 'no trusted brand' count
+    no_brand_count = trusted_brands[no_brand_responses].sum()
 
-    plt.title('Correlation Network', fontproperties=chinese_font)
-    plt.axis('off')  # Turn off the axis
+    # Remove the individual 'no brand' responses and add the combined count
+    trusted_brands_combined = trusted_brands.drop(no_brand_responses)
+    trusted_brands_combined.loc['No trusted brand'] = no_brand_count
 
-    # Use Streamlit to render the plot
-    st.pyplot(plt)
+    # Combine non-specified brands
+    have_but_not_specified = ["有", "有", "Yes", "應該有"]
 
-def generate_correlation_chart(df, chinese_font):
+    # Calculate the combined 'not specified brand' count
+    have_but_not_specified_count = trusted_brands_combined[have_but_not_specified].sum()
 
-    boolean_fields = [
-        '你/妳覺得目前有任何投資嗎？'
-    ]
+    # Remove the individual 'not specified brand' responses and add the combined count
+    trusted_brands_combined = trusted_brands_combined.drop(have_but_not_specified)
+    trusted_brands_combined.loc['Have but not specified'] = have_but_not_specified_count
 
-    # Encode boolean fields
-    for field in boolean_fields:
-        df[field + '_encoded'] = df[field].map({'有': 1, '沒有': 0})
+    summary = trusted_brands_combined.sort_values(ascending=False)
 
-    # 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]
+    # Return the sorted series with combined 'no brand' count
+    st.write("Summary of Trusted Brands:")
+    st.table(summary)
 
-    # Filter the correlation matrix to only include these fields
-    filtered_correlation_data = correlation_data.loc[strong_fields, strong_fields]
+
+def visualize_shopping_data(df, chinese_font):
+    # Shopping fields with their corresponding titles
+    shopping_fields = {
+        "你/妳會買哪一種番茄？": "Which Type of Tomatoes Would You Buy?",
+        "你/妳買哪種牛奶？": "Which Type of Milk Would You Buy?",
+        "你/妳會買哪種雞蛋？": "Which Type of Eggs Would You Buy?"
+    }
     
-    # Plot the correlation matrix
-    plt.figure(figsize=(10, 8))
-    ax = sns.heatmap(filtered_correlation_data, annot=True, fmt=".2f", cmap="coolwarm")
+    # Create a figure and a set of subplots
+    fig, axes = plt.subplots(len(shopping_fields), 1, figsize=(10, 6 * len(shopping_fields)))
     
-    # 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)
+    # If there's only one field to plot, axes will not be an array, so we wrap it in a list
+    if not isinstance(axes, np.ndarray):
+        axes = [axes]
     
-    # 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)
+    for ax, (column_name, title) in zip(axes, shopping_fields.items()):
+        # Summarize the data
+        data = df[column_name].value_counts().head(20)  # Adjust the number as needed
+        
+        # Plot the data
+        data.plot(kind='bar', color='skyblue', ax=ax, fontsize=12)
+        ax.set_title(title, fontproperties=chinese_font)
+        ax.set_xlabel('Options', fontproperties=chinese_font)
+        ax.set_ylabel('Count', fontproperties=chinese_font)
+
+        # Set the properties for the x-tick labels
+        ax.set_xticklabels(data.index, rotation=45, ha='right', fontproperties=chinese_font)
+    
+    plt.tight_layout()
+    st.pyplot(fig)