a_rule.py

# -*- coding: utf-8 -*-

import numpy as np

data_set = {
        'transaction_1': [0,1,0,0,1],
        'transaction_2': [1,0,1,1,0],
        'transaction_3': [1,0,1,0,0],
        'transaction_4': [0,1,0,1,0],
        'transaction_5': [0,0,0,1,0]
        }

print("## data_set")
print("transaction: [item1, item2, item3, item4, item5]")
for k, v in data_set.items():
    print(k, v)

# cast to numpy.ndarray
items = np.array(list(data_set.values()))

def support(items, index_a, index_b):
    '''
    Support(A -> B) = o(A U B) / N
    o: num of B transactions that contain A
    n: num of all transactions

    items: [[item1, ..., itemM],
            ...
            [item1, ..., itemM]]
    index_[a, b]: index of the item
    '''
    item_a = items[:, index_a]
    item_b = items[:, index_b]
    o = (item_a * item_b > 0).sum()
    n = items.shape[0]
    return o / n

def confidence(items, index_a, index_b):
    '''
    Confidence(A -> B) = o(A U B) / n(A)
    o: num of B transactions that contain A
    n: num of transactions that contain A

    items: [[item1, ..., itemM],
            ...
            [item1, ..., itemM]]
    index_[a, b]: index of the item
    '''
    item_a = items[:, index_a]
    item_b = items[:, index_b]
    o = (item_a * item_b > 0).sum()
    n = (item_a > 0).sum()
    return o / n

def lift(items, index_a, index_b):
    '''
    Lift(A -> B) = c(Confidence(A -> B)) / n(B)
    c: value of Confidence(A -> B)
    n: num of transactions that contain B

    items: [[item1, ..., itemM],
            ...
            [item1, ..., itemM]]
    index_[a, b]: index of the item
    '''
    c = confidence(items, index_a, index_b)
    n = (items[:, index_b] > 0).sum()
    return c / n

support_0_3 = support(items, 0, 3)
confidence_0_3 = confidence(items, 0, 3)
lift_0_3 = lift(items, 0, 3)

print("\n=== Association rules ===")
print("support: ", support_0_3)
print("confidence: ", confidence_0_3)
print("lift: ", lift_0_3)