Implementation of Random Forest Regression with Hyperparameter Tuning and Shap Analysis to Predict Pull Test Result at Soft Laminating Machine

Abstract

Efficiency is the key in the manufacturing industry. In this industry, manufacturers are competing to be the most efficient company and the one that spends the least. PT. XYZ is a company that operates in the audio industry, manufacturing speakers and audio peripherals. In the audio industry, the quality of speaker housing is exceptional. It will prove the quality of the product over a long period. The housing is usually made from MDF coated with PVC. This study tries to collect the historical data of adhesive strength produced in the manufacturing company and make a systematic predictive method that can approximate the relative strength of the product made. The result from the study shows that the method has an error rate of about 0.21 from real adhesive strength, suggesting an accurate prediction method for the company. The result can also be analyzed using SHAP analysis, which is also included as the output of the method.