彙整

Python 的 Market Basket Analysis

Isaiah Hull

Visiting Associate Professor of Finance, BI Norwegian Business School

探索資料

import pandas as pd

# Load novelty gift data.
gifts = pd.read_csv('datasets/novelty_gifts.csv')

# Preview data with head() method.
print(gifts.head())

  InvoiceNo                          Description
0    562583      IVORY STRING CURTAIN WITH POLE 
1    562583        PINK AND BLACK STRING CURTAIN
2    562583                PSYCHEDELIC TILE HOOK
3    562583                ENAMEL COLANDER CREAM
4    562583  SMALL FOLDING SCISSOR(POINTED EDGE)

探索資料

# Print number of transactions.
print(len(gifts['InvoiceNo'].unique()))

# Print number of items.
print(len(gifts['Description'].unique()))

修剪與彙整

修剪（Pruning） 圖示為一份項目清單，其中部分項目被劃掉，表示已被修剪。

彙整（Aggregation） 圖示為一份項目清單被對應到較小的清單，以表示彙整。

彙整資料

# Load one-hot encoded data
onehot = pd.read_csv('datasets/online_retail_onehot.csv')

# Print preview of DataFrame
print(onehot.head(2))

    50'S CHRISTMAS GIFT BAG LARGE   DOLLY GIRL BEAKER ...  ZINC WILLIE WINKIE  CANDLE STICK
0                           False               False                               False
1                           False               False                               True

彙整資料

# Select the column names for bags and boxes
bag_headers = [i for i in onehot.columns if i.lower().find('bag')>=0]
box_headers = [i for i in onehot.columns if i.lower().find('box')>=0]

# Identify column headers
bags = onehot[bag_headers]
boxes = onehot[box_headers]
print(bags)

       50'S CHRISTMAS GIFT BAG LARGE   RED SPOT GIFT BAG LARGE  
0                              False                     False   
1                              False                     False
...                             ...                      ...

彙整資料

# Sum over columns
bags = (bags.sum(axis=1) > 0.0).values
boxes = (boxes.sum(axis=1) > 0.0).values
print(bags)

[False  True False ... False  True False]

彙整資料

# Add results to DataFrame
aggregated = pd.DataFrame(np.vstack([bags, boxes]).T, columns = ['bags', 'boxes'])

print(aggregated.head())

    bags  boxes
0  False  False
1   True  False
2  False  False
3  False  False
4   True  False

用彙整做市場籃分析

彙整流程：
- Items -> Categories
- 計算指標
- 擬定規則

# Compute support
print(aggregated.mean())

bags     0.130075
boxes    0.071429

一起來練習吧！

Python 的 Market Basket Analysis