Market basket analysis is an analysis method that associates two different products and mines their relations by analyzing user purchasing data.
You can calculate lift according to this document. There is a high correlation between the two products with lift more than 1, indicating that the combination of the two products is effective and they can be sold together.
Note:The join in this document will result in a Cartesian product, causing data to expand N times. You are advised to use a small portion of data for market basket analysis (avoid excessive data expansion).
Identify customers' purchasing patterns.
Will the purchase of product A by users have any impact on product B?
Do different users have different purchasing patterns?
Which products should be bundled together for sales?
Association analysis is measured by the following indicators.
Definition
Probability
Example
Support
The probability that both product A and product B are purchased at the same time, or the proportion of a product combination's purchasing times to two products' total purchasing times
Support indicates how representative the association rule is in all things. The higher the support is, the more important the association rule is.
Support of product A for product B
P(A ∩ B)
There are a total of 10 orders, among which milk and bread are purchased six times. So the support of the milk and bread group is 6 / 10 = 60%.
Confidence
The conditional probability of purchasing product B after purchasing product A (purchase product B due to purchasing product A)
Confidence of product B for product A
P(A | B)
There are a total of 10 orders, among which product A is purchased eight times and product A with product B is purchased six times. The confidence is 6 / 8 = 75%.
Lift
The lift of purchasing product A for purchasing product B. Lift is used to determine whether the combination of products has actual value (more purchasing times of the combination products than those of the single product). Combination is effective with lift more than 1.
L= P(A ∩ B) / [P (A) * P (B)]
There are a total of 10 orders, among which product A is purchased eight times, product B is purchased six times, and product A with product B is purchased six times. So the lift is 0.6 / (0.8*0.6) = 1.25 (> 1), indicating that the combination of product A and product B is effective.
Calculation formulas:
Support: The order number of purchasing product A and product B simultaneously / total order number of purchasing
Confidence: The order number of purchasing product A and product B simultaneously / the order number of purchasing product A
Lift: support / ((number of purchasing product A / total number of purchasing orders) * (number of purchasing product B / total number of purchasing orders))
Set the field Order Number as the join basis to obtain all product combinations in the order.
1. Create an analysis subject called Market Basket Analysis. Add data, click Field Settings, and tick the fields Order Number and Product Name.
2. Add data again, click Join with the fields Order Number and Product Name in the table Market Basket Analysis 1 as the join datasets.
3. Click Left Join and set Order Number as the join basis.
4. Click Field Settings. Change Product Name into Product A and Product Name 1 into Product B.
Indicators needed include order number of purchasing product A and product B simultaneously, order number of purchasing product A, order number of purchasing product B, and order number of the total purchasing.
Add a column, group it according to the field Product A, and count unique values to obtain the number of purchasing product A.
Add a column, group it according to the field Product B, and count unique values to obtain the number of purchasing product B.
Add a column, group it according to the fields Product A and Product B, and count unique values to obtain the order umber of purchasing product A and product B simultaneously.
Add a column and count unique values for all order numbers to obtain the total number of orders.
Next, you need to calculate the support, confidence, and lift.
Support = order number of purchasing product A and product B simultaneously / order number of the total purchasing. Click Formula Column.
Confidence = order number of purchasing product A and product B simultaneously / order number of purchasing product A. Click Formula Column.
Lift = support / ((order number of purchasing product A / order number of the total purchasing) * (order number of purchasing product B / order number of the total purchasing)). Click Formula Column.
It is meaningless to calculate the support, confidence, and lift for the same product (same fields of Product A as that of Product B). You can filter the combination of same product A and product B.
Save and update the analysis subject.
1. Use Market Basket Analysis to create a component. Select Custom Chart and drag Product A into Horizontal Axis and Product B into Vertical Axis.
Select Graphic Property, set the chart to rectangular block, and drag the field Support into Color. Due to the addition for the values of multiple orders, you need to set Summary Mode to Average.
2. Create another component to display the detailed data of support, confidence, and lift.
Set Summary Mode to Average due to the merging summation of multiple orders. Set Numeric Format to Percentage.
For details, see the section "Preview".
滑鼠選中內容，快速回饋問題
滑鼠選中存在疑惑的內容，即可快速回饋問題，我們將會跟進處理。
不再提示
10s後關閉
Submitted successfully
Network busy