Python中如何分析用户行为？漏斗模型实现

import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

# 模拟用户事件数据
# 实际工作中，这些数据可能来自数据库查询、日志文件等
data = {
    'user_id': [1, 1, 1, 1, 2, 2, 2, 3, 3, 4, 4, 4, 4, 5, 5, 5],
    'event_time': pd.to_datetime([
        '2023-01-01 10:00:00', '2023-01-01 10:05:00', '2023-01-01 10:10:00', '2023-01-01 10:15:00',
        '2023-01-01 11:00:00', '2023-01-01 11:05:00', '2023-01-01 11:10:00',
        '2023-01-01 12:00:00', '2023-01-01 12:05:00',
        '2023-01-01 13:00:00', '2023-01-01 13:05:00', '2023-01-01 13:10:00', '2023-01-01 13:15:00',
        '2023-01-01 14:00:00', '2023-01-01 14:05:00', '2023-01-01 14:10:00'
    ]),
    'event_name': [
        'view_product', 'add_to_cart', 'checkout', 'purchase_complete',
        'view_product', 'add_to_cart', 'checkout',
        'view_product', 'add_to_cart',
        'view_product', 'add_to_cart', 'checkout', 'purchase_complete',
        'view_product', 'add_to_cart', 'checkout'
    ]
}
df_events = pd.DataFrame(data)

# 定义漏斗阶段及其顺序
funnel_stages = [
    'view_product',
    'add_to_cart',
    'checkout',
    'purchase_complete'
]

# 初始化一个DataFrame来存储每个用户在每个阶段的首次完成时间
user_funnel_progress = pd.DataFrame({'user_id': df_events['user_id'].unique()})

# 遍历漏斗阶段，找到每个用户首次完成该阶段的时间
for stage in funnel_stages:
    # 筛选出当前阶段的事件
    stage_events = df_events[df_events['event_name'] == stage].copy()
    # 找到每个用户首次完成该阶段的时间
    first_occurrence = stage_events.groupby('user_id')['event_time'].min().reset_index()
    first_occurrence.rename(columns={'event_time': f'{stage}_time'}, inplace=True)

    # 将结果合并到user_funnel_progress中
    user_funnel_progress = pd.merge(user_funnel_progress, first_occurrence, on='user_id', how='left')

# 确保用户按照漏斗阶段的顺序进行
# 如果用户跳过了某个阶段，或者阶段发生顺序错误，则认为该用户未通过该阶段
# 这里我们检查每个阶段的时间是否晚于或等于前一个阶段的时间
for i in range(1, len(funnel_stages)):
    prev_stage_time_col = f'{funnel_stages[i-1]}_time'
    curr_stage_time_col = f'{funnel_stages[i]}_time'

    # 如果当前阶段的时间为空，或者早于前一个阶段的时间，则将当前阶段时间设为NaN
    user_funnel_progress.loc[
        (user_funnel_progress[curr_stage_time_col].isnull()) | 
        (user_funnel_progress[curr_stage_time_col] < user_funnel_progress[prev_stage_time_col]), 
        curr_stage_time_col
    ] = pd.NaT # pd.NaT for datetime NaNs

# 计算每个阶段的用户数
stage_counts = {}
for i, stage in enumerate(funnel_stages):
    # 对于第一个阶段，计算所有有该事件的用户数
    if i == 0:
        stage_counts[stage] = user_funnel_progress[f'{stage}_time'].notna().sum()
    else:
        # 对于后续阶段，计算那些成功通过前一阶段且通过当前阶段的用户数
        prev_stage_col = f'{funnel_stages[i-1]}_time'
        curr_stage_col = f'{funnel_stages[i]}_time'
        stage_counts[stage] = user_funnel_progress[
            user_funnel_progress[prev_stage_col].notna() & 
            user_funnel_progress[curr_stage_col].notna()
        ].shape[0]

# 将结果转换为DataFrame以便计算和可视化
funnel_df = pd.DataFrame(list(stage_counts.items()), columns=['Stage', 'Users'])
funnel_df['Users_Prev_Stage'] = funnel_df['Users'].shift(1).fillna(funnel_df['Users'].iloc[0])
funnel_df['Conversion_Rate'] = (funnel_df['Users'] / funnel_df['Users_Prev_Stage'] * 100).round(2)
funnel_df['Cumulative_Conversion'] = (funnel_df['Users'] / funnel_df['Users'].iloc[0] * 100).round(2)

print("漏斗分析结果:")
print(funnel_df)

# 可视化漏斗
plt.figure(figsize=(10, 6))
sns.barplot(x='Stage', y='Users', data=funnel_df, palette='viridis')
plt.title('用户行为漏斗模型')
plt.xlabel('漏斗阶段')
plt.ylabel('用户数')

# 在柱状图上添加转化率标签
for index, row in funnel_df.iterrows():
    plt.text(index, row['Users'], f"{row['Users']} ({row['Cumulative_Conversion']}%)", color='black', ha="center", va='bottom')
    if index > 0:
        plt.text(index, row['Users'] / 2, f"{row['Conversion_Rate']}%", color='red', ha="center", va='center', weight='bold') # 阶段间转化率
plt.grid(axis='y', linestyle='--', alpha=0.7)
plt.show()