A Method for Predicting the Film Thickness of IC Deposited Films Based on FCBF-CATBOOST

This article presents a novel method for predicting film thickness of IC deposited films. The study uses the category boosting (CatBoost) to predict film deposition thickness, employs the Fast Correlation-Based Filter (FCBF) for data cleaning of the relevant process parameters, feature extraction, and uses grid search and cross-validation for hyperparameter optimization. Through experiments using regression metrics such as R-Squared and Mean Square Error, FCBF-CatBoost demonstrates higher prediction accuracy compared to algorithms like support vector machine (SVM), XGBoost and random forest (RF).