Anonymous Transactional Dataset

Description

The dataset is comprised of real-world historical sales transactions and product metadata collected from a local Food and Beverage (F&B) Micro-Small-Medium Enterprise (MSME) operating multiple outlets in Indonesia. This data can be used by researchers, data scientists, and industry professionals using various techniques in recommender system and machine learning.

Steps to reproduce

The Anonymous Transactional Dataset consisted of two data. i.e. Transactions data and Product metadata. The Transactions data comprises historical sales records collected by the business's internal Point-of-Sale (POS) system between January 2025 and September 2025. This data was extracted from raw spreadsheet files and then structured, filtered, and cleaned using Python scripts. The initial transaction data comprised of 22 features, but investigation revealed that 9 of these features were entirely empty: Customer, Customer Phone, Refund, Reasons of Refunds, Discount, Gratuaty, Other Note, Event, and Served By. Closer inspection showed that the Total Collected – Total Amount and Gross Sales – Net Sales feature contained values identical every record. This constitutes a direct data duplication, which must be removed to prevent data redundancy and avoid introducing bias. These eleven features will be dropped resulting in 11 features in the Transaction data. Meanwhile, Product metadata comprised of 6 features: Product ID, Product Name, Variant, Category, Price, Description.

Institutions

• Bina Nusantara University

  DKI Jakarta, West Jakarta
• Universitas Ciputra Surabaya

  Surabaya

Categories

Licence