PENGEMBANGAN MODEL PENENTUAN TITIK PEMESANAN ULANG SUKU CADANG BERBASIS MACHINE LEARNING DI PT X

• Main Author: Nur Alya, Raihana
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/68802
• Institution: Institut Teknologi Bandung

PT X is one of many companies that engaged in manufacturing cars and their components, which acts as a manufacturer and exporter of products and spare parts. In 2019, the realization of PT X's exports was able to achieve export performance of 208,500 units to more than 80 countries. One of the reasons for the good reception in PT X's export destinations is because PT X always maintains the quality of its services in delivering products to providing after-sales service activities. One form of after-sales service activity offered by PT X is periodic service which includes inspection and maintenance of its spare parts. In managing spare parts inventory for PT X, it is necessary to carry out inventory management, one of which is in determining the value of the reorder point (ROP). Optimal determination in determining ROP will provide benefits in managing safety stock, reducing inventory costs, increasing productivity, and increasing revenue. Calculation of determining ROP using traditional methods is considered to require a lot of time for the company because considering the large number of spare parts that have to be calculated. Therefore, an Artificial Intelligence (AI) based approach is important to be able to provide efficiency in determining the ROP value. The AI method used in predicting the ROP value is the machine learning method. In this study, the machine learning algorithms used to predict ROP are Artificial Neural Network (ANN), Support Vector Regression (SVR), Linear Regression, Decision Tree, and Random Forest. The conclusion of the study is that the ANN model provides the best performance with the error generated through an average MAPE value of 8.36% and an average R2 value of 0.9682 or 96.82%. The machine learning method also provides a computation time efficiency of 98.51% compared to the computation time from the use of traditional algorithm models.