Quiz tổng hợp

1# Bước 1: Xác định cohort month (tháng đầu tiên mua hàng)
2df['order_month'] = df['order_date'].dt.to_period('M')
3cohort = df.groupby('customer_id')['order_month'].min().rename('cohort_month')
4df = df.merge(cohort, on='customer_id')
5
6# Bước 2: Tính cohort index (khoảng cách tháng)
7df['cohort_index'] = (df['order_month'] - df['cohort_month']).apply(lambda x: x.n)
8
9# Bước 3: Đếm số unique customers mỗi cohort × cohort_index
10cohort_table = df.groupby(['cohort_month', 'cohort_index'])['customer_id'].nunique().unstack()
11
12# Bước 4: Tính retention rate (%)
13cohort_size = cohort_table.iloc[:, 0]
14retention = cohort_table.divide(cohort_size, axis=0) * 100

1from scipy import stats
2import numpy as np
3
4# Conversion rates
5cr_a = 200 / 5000  # 4.0%
6cr_b = 260 / 5000  # 5.2%
7
8# Z-test for proportions
9p_pool = (200 + 260) / (5000 + 5000)  # pooled proportion
10se = np.sqrt(p_pool * (1 - p_pool) * (1/5000 + 1/5000))
11z_stat = (cr_b - cr_a) / se
12p_value = 2 * (1 - stats.norm.cdf(abs(z_stat)))
13# p_value ≈ 0.007 < 0.05 → Bác bỏ H₀
14# Kết luận: Variant B có conversion rate cao hơn có ý nghĩa thống kê
15# Relative lift: (5.2% - 4.0%) / 4.0% = +30%

1import pandas as pd
2from datetime import datetime
3
4# Tính R, F, M cho mỗi khách hàng
5snapshot_date = df['order_date'].max() + pd.Timedelta(days=1)
6rfm = df.groupby('customer_id').agg(
7    Recency=('order_date', lambda x: (snapshot_date - x.max()).days),
8    Frequency=('order_id', 'nunique'),
9    Monetary=('revenue', 'sum')
10)
11
12# Gán điểm 1-5 bằng qcut
13rfm['R_score'] = pd.qcut(rfm['Recency'], 5, labels=[5,4,3,2,1])
14rfm['F_score'] = pd.qcut(rfm['Frequency'].rank(method='first'), 5, labels=[1,2,3,4,5])
15rfm['M_score'] = pd.qcut(rfm['Monetary'], 5, labels=[1,2,3,4,5])
16
17# Gán segment
18def segment(row):
19    if row['R_score'] >= 4 and row['F_score'] >= 4:
20        return 'Champions'
21    elif row['R_score'] >= 3 and row['F_score'] >= 3:
22        return 'Loyal'
23    elif row['R_score'] >= 4 and row['F_score'] <= 2:
24        return 'New Customers'
25    elif row['R_score'] <= 2 and row['F_score'] >= 3:
26        return 'At Risk'
27    else:
28        return 'Need Attention'
29
30rfm['Segment'] = rfm.apply(segment, axis=1)