This project analyses 3,000 receipts of a fictional convenience store to find combinations of items bought together. The output is a set of association rules with support, confidence and lift that feed directly into shelf layout and promo bundles.
Business Goal
Find items often bought together to arrange shelves and design promo bundles. Application area: Retail. Project level: intermediate.
RapidMiner Tools and Operators
The process is built with these operators:
- Read CSV
- Numerical to Binominal
- FP-Growth
- Create Association Rules
- Frequent Item Sets to Data
- Write CSV
What You Will Learn
- Understand support, confidence and lift with a real retail example
- Turn association rules into shelf layout and promo bundle decisions
- See why lift is more honest than confidence for popular items
Working Steps
- Read the receipt data — Read CSV loads transaksi_ritel.csv: one row per receipt and one column per item (1 = bought, 0 = not). The id_transaksi column gets the id role.
- Convert 0/1 numbers to binominal — FP-Growth only accepts binominal columns. Numerical to Binominal turns 0 into false and anything else into true.
- Find frequent itemsets — FP-Growth with minimum support 0.03 (at least 3% of receipts) and at most 4 items per itemset. A low support keeps niche combinations visible.
- Create association rules — Create Association Rules uses a minimum confidence of 0.4. Each rule reads "if A is bought, then B is bought" with its support, confidence and lift.
- Save itemsets and rules — Frequent Item Sets to Data turns itemsets into a table and Write CSV saves it. The association rules go to a result port so you can see them in the RapidMiner Results panel and export them from there.
- Filter the meaningful rules — In the results panel, sort the rules by lift. A lift above 1 means the items are bought together more often than by chance; ignore lift below 1.
Practice Dataset
3,000 fictional receipts with 24 products such as bread, butter, instant noodles, eggs, coffee, sugar, soap and detergent. A value of 1 means the item is on that receipt. (3,000 rows, 25 columns).



