Understanding Transactional Data Patterns from Micro Business using Machine Learning Algorithms

This study aims to predict sales trends using the Auto Regression (AR) method, supported by the Market Basket Analysis (MBA) in predicting hidden patterns in sales transaction data. The subject of this study is Logic Store, a retail store selling computer and cellphone accessories located in Pontianak, Indonesia. This quantitative research utilizes primary data from the Logic Store sales database. The research data for the AR method consists of 158 data points, representing sales revenue over 3 years, from January 2021 to February 2024, with a weekly time interval. The research data for the MBA method includes 2391 transactions, which were then filtered down to 145 transactions. The data used involves transactions containing more than one type of product. The results show that sales trends generally increased from 2021 to 2023. Sales trend predictions indicate that sales will rise at the beginning of each month, especially during the transition between September, October, and November. The results also suggest potential bundling promotions, such as earphone-earphone bag, True Wireless Stereo (TWS)-earphone, mouse-mousepad, and charger adapter-cable. Future research should consider combining other methods, such as decision trees, moving average, and so on with the methods used in this paper.