Penerapan Data Mining Untuk Memprediksi Penjualan Pakaian Menggunakan Metode K-Nearest Neighbor (KNN) (Studi kasus : Trifting Second 3).

Abstrak

The trend of secondhand clothing, or thrifting, has seen significant growth, especially among young people. One of the businesses operating in this field is Thrifting Second 3, which markets its products online. However, this business faces challenges in forecasting sales, leading to excess stock of less popular items and stockouts of high-demand products. This study aims to utilize the K-Nearest Neighbor (KNN) algorithm to predict sales of thrift products based on historical sales data. KNN was chosen due to its capability to handle non-linear data and its proven accuracy in previous studies. The system development follows the Rational Unified Process (RUP) approach, which includes the stages of inception, elaboration, construction, and transition. The sales prediction system was tested using the black box method and proved to function properly, with all main features such as admin login, product and sales data management, sales prediction, prediction reports, and product display performing as expected. Based on the User Acceptance Test (UAT), the system achieved a total score of 87.84%, categorized as “Excellent,” indicating it meets user requirements. Sales prediction using the K-Nearest Neighbor (KNN) algorithm showed that with K = 3, MAE = 3.28 and MAPE = 4.52%, while with K = 5, MAE increased to 3.77 and MAPE to 5.52%, indicating slightly lower accuracy. The system remains effective and reliable as a decision support tool for sales planning.