【Python×Apriori】アソシエーション・バスケット分析実践法｜マーケティングレコメンド手法への応用支援

pip install mlxtend

import matplotlib.pyplot as plt
import pandas as pd

# データセット
data = pd.read_csv('Grocery Products Purchase.csv.xls')

# レコード数
record_len = len(data)
# カラム数
column_len = len(data.columns)

# トランザクション形式に加工
transactions = []
for i in range(record_len):
    # データをリスト型に変更
    values = [str(data.values[i,j]) for j in range(column_len)]
    values_notnull = []
    for check in values:
        if check != 'nan':
            values_notnull.append(check)
    transactions.append(values_notnull)

# 結果出力
print(transactions)

# 出力イメージ
# [['citrus fruit', 'semi-finished bread', 'margarine', 'ready soups'],
#  ['tropical fruit', 'yogurt', 'coffee'],
#  ['whole milk'],
#  ['pip fruit', 'yogurt', 'cream cheese', 'meat spreads'],
#  ['other vegetables','whole milk','condensed milk','long life bakery product'],
#  ['whole milk', 'butter', 'yogurt', 'rice', 'abrasive cleaner'],
#  ['rolls/buns'],
#  ['other vegetables','UHT-milk','rolls/buns','bottled beer','liquor (appetizer)'],
#  ['potted plants'],
#  ['whole milk', 'cereals'],
#  ['tropical fruit','other vegetables','white bread','bottled water','chocolate'],]

from mlxtend.preprocessing import TransactionEncoder

# データをテーブル形式に加工
te       = TransactionEncoder()
te_array = te.fit(transactions).transform(transactions)
df       = pd.DataFrame(te_array, columns=te.columns_)

from mlxtend.frequent_patterns import apriori

freq_items = apriori(df,                     # データフレーム
                     min_support  = 0.01,    # 支持度(support)の最小値
                     use_colnames = True,    # 出力値のカラムに購入商品名を表示
                     max_len      = None,    # 生成されるitemsetsの個数
                     verbose = 0,            # low_memory=Trueの場合のイテレーション数
                     low_memory = False,     # メモリ制限あり＆大規模なデータセット利用時に有効
                    )

# 結果出力
freq_items = freq_items.sort_values("support", ascending = False).reset_index(drop=True)
print(freq_items)

# 出力イメージ
#      support            itemsets
# 0   0.255516        (whole milk)
# 1   0.193493  (other vegetables)
# 2   0.183935        (rolls/buns)
# 3   0.174377              (soda)
# 4   0.139502            (yogurt)
# 5   0.110524     (bottled water)
# 6   0.108998   (root vegetables)
# 7   0.104931    (tropical fruit)
# 8   0.098526     (shopping bags)
# 9   0.093950           (sausage)
# 10  0.088968            (pastry)
# 11  0.082766      (citrus fruit)
# 12  0.080529      (bottled beer)
# 13  0.079817        (newspapers)
# 14  0.077682       (canned beer)

from mlxtend.frequent_patterns import association_rules

# アソシエーション・ルール抽出
df_rules = association_rules(freq_items,             # supportとitemsetsを持つデータフレーム
                             metric = "confidence",  # アソシエーション・ルールの評価指標
                             min_threshold = 0.1,    # metricsの閾値
                            )

results = df_rules[(df_rules['confidence'] > 0.2) &  # 信頼度
          (df_rules['lift'] > 3.0)]                  # リフト値

# 結果出力
print(results.loc[:,["antecedents","consequents","confidence","lift"]])

# 出力イメージ
#                             antecedents           consequents  confidence  lift
# 162                              (beef)     (root vegetables)    0.331395  3.040367   
# 301  (other vegetables, tropical fruit)     (root vegetables)    0.342776  3.144780 
# 302   (root vegetables, tropical fruit)    (other vegetables)    0.584541  3.020999  
# 414    (citrus fruit, other vegetables)     (root vegetables)    0.359155  3.295045 
# 415     (citrus fruit, root vegetables)    (other vegetables)    0.586207  3.029608 
# 443          (yogurt, other vegetables)  (whipped/sour cream)    0.234192  3.267062