PENGEMBANGAN MODEL ESTIMASI BIAYA PADA DESAIN BERBANTUAN KOMPUTER MENGGUNAKAN PEMBELAJARAN MESIN

• Main Author: Usman, Marzuki
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/72119
• Institution: Institut Teknologi Bandung

CV CSM (Cipta Sinergi Manufacturing) is a company focused on machining products manufacturing such as mechanical spare part, moulding, dies, and other manufactured products. The production used Make-to-Order (MTO) system to accommodate custom production and low volumes. In designing product, the company estimates the design costs using the percentage of the cost of goods sold (COGS) of the product so that the results obtained are considered inconsistent and less representative than the design complexity. Therefore, the purpose of this study is to design a model that can be used to estimate costs based on the level of complexity. This research uses machine learning which refers to the CRISP-DM methodology for estimating design costs. The three algorithms used are random forest, multiple linier regression, dan extreme gradient boosting. The best algorithm is selected by implementing each model with the available dataset. In this case, the model that has the best performance in estimating costs is a random forest algorithm with an average R score of 0.90 in the test set and 0.89 in the training set. It also has MAPE (Maximum Absolute Percentage Error) performance of 35% in the test set and 25% in the training set. Based on the best machine learning model, then a design cost estimation software is developed using the Python programming language that can be run using Google Collaborator. By using the software, a validation test was carried out by estimating costs using data from the CV CSM and obtained the average error value of 24.83%. This value indicates that the model produced by the perangkat lunak prototype considered valid and can be implemented practically in the industry.