python实现apriori关联规则算法代码实现

1、apriori关联规则原理

从频繁项集生成关联规则的函数

2、apriori关联规则算法实现

第一步、导入原始数据，统计出现的数据种类

# 导入包
import mlxtend
import numpy as np
import pandas as pd
from mlxtend.frequent_patterns import apriori, fpmax, fpgrowth, association_rules

# 打开数据
data = pd.read_csv('./apriori.txt', header=None, dtype=object)
a, b = data.shape  # 930,7

# 统计出现的数据种类
df = data
all_kind = []
for i in range(a):
    for j in range(b):
        all_kind.append(df.loc[i, j])

from collections import Counter
result = Counter(all_kind)
print('各数据出现的次数:\n--------------------\n', pd.DataFrame([list(result.keys()), list(result.values())],
                                                        index=['数据', '次数']))  # 6510

第二步，将原始数据转为换布尔型

kinds = list(result.keys())
# 创建0矩阵数据框
data_ = pd.DataFrame(np.zeros([a, len(kinds)]))
data_.columns = kinds

# 转布尔型True/False
for i in range(a):
    for j in range(b):
        if df.loc[i, j] in kinds:
            data_.loc[i, df.loc[i, j]] = True
df_ = data_.replace(0, False)
print('转为布尔类型后的数据：\n-------------------\n', df_)

# 验证True个数(可省略)
num_ = []
for i in range(a):
    for j in range(len(kinds)):
        num_.append(df_.loc[i, kinds[j]])
print(Counter(num_))

第三步、计算频繁项集

frequent_itermsets = apriori(df_, min_support=0.06, use_colnames=True) # 最小支持度设为0.06
print('频繁项集：\n-----------------------\n', pd.DataFrame(frequent_itermsets))

第四步、根据置信度筛选关联规则

result = association_rules(frequent_itermsets, metric='confidence', min_threshold=0.75)  # 最小置信度设为0.75,根据置信度筛选
print('筛选之后的关联规则：\n-----------------------\n', result[['antecedents', 'consequents',  'support', 'confidence']])

第五步、输出结果、得出结论

筛选之后的关联规则：
-----------------------
     antecedents consequents   support  confidence
0      (F4, C3)        (H4)  0.075269    0.875000
1      (F4, B2)        (H4)  0.062366    0.794521
2      (A3, F4)        (H4)  0.078495    0.879518
3      (E3, C2)        (D2)  0.092473    0.754386
4  (H4, D2, F3)        (A2)  0.062366    0.753247

相关数据下载链接