Implementasi Metode Random Forest Untuk Memprediksi Jumlah Penjualan Gorden Berdasarkan Data Historis

Abstract

The rapid development of information technology has encouraged companies, including Tova Gorden, a small business engaged in curtain sales, to adopt technology to improve operational efficiency and competitiveness. Tova Gorden often faces obstacles in fulfilling orders, especially when demand suddenly increases, which is exacerbated by limited stock, raw material difficulties (such as smokers), fabric pre-order systems, and time-consuming production processes. Determining stock that is still based on employee estimates often leads to inefficiencies in the form of shortages or excesses of goods. This condition highlights the urgent need for an accurate prediction system to optimize inventory management. This study aims to implement and test the performance of the Random Forest algorithm, which is an ensemble learning method, to predict the number of curtain sales based on historical sales data. The collected data includes historical information related to curtain sales, including sales weeks, curtain motifs, and sales volumes. Unlike previous studies that generally use Linear Regression and focus on products with stable sales patterns, this study applies Random Forest to address more fluctuating curtain demand patterns. This research method includes several stages, namely Data Collection, Exploratory Data Analysis (EDA), Data Preprocessing, Data Splitting (70% training, 15% validation, 15% testing), Modeling with Random Forest, Evaluation, and Deployment. The evaluation results show that the model has excellent performance, with a coefficient of determination (R²) value of 97.83% on training data, 93.72% on validation data, and 96.64% on test data. Furthermore, the model is integrated into a web-based system using the Flask framework. This system is equipped with data upload features, prediction processes and curtain category grouping, and presentation of model evaluation results.