Penentuan Kuantitas Supply Kemasan Menggunakan Pendekatan Data Mining pada Store ABC melalui Aplikasi XYZ

Abstract

The development of e-commerce in Indonesia, particularly through companies like PT. XYZ, poses challenges in determining the number of packaging supplies at Store ABC. These challenges can be addressed by utilizing historical purchasing data and data mining techniques such as K-Means Clustering and Random Forest prediction to minimize excess or shortage of packaging. This study aims to design and analyze a data mining model for predicting the packaging supply requirements at Store ABC. The clustering method with the K-Means algorithm is used to group transaction data, while prediction methods with Linear Regression and Random Forest algorithms are used to forecast future packaging needs. The processed sales transaction data from January to December 2023 consists of 34,578 records with relevant attributes. Clustering results show the division of transaction data into 3 clusters representing different types of packaging. The packaging needs prediction using Linear Regression and Random Forest is conducted for the next 12 months. Validation tests are performed using evaluation metrics such as Mean Absolute Percentage Error (MAPE) and Pearson Correlation Coefficient (R^2). The analysis results indicate that the Random Forest method outperforms Linear Regression, with an average MAPE of 13% and an average R^2 of 0.83. Additionally, the operational packaging supply cost, initially Rp 20,012,306, was reduced by 4% with Linear Regression, amounting to Rp 19,218,177, and by 5% with Random Forest, amounting to Rp 19,060,719. Thus, it can be concluded that Random Forest also effectively minimizes operational packaging supply costs compared to Linear Regression.