Hyperparameters and Tuning Methods for Random Forest Using Python Sklearn Package Relevant to Psychology Studies

Abstract

Machine learning (ML) is used to create well-functioning prediction models for predicting the prognosis of psychiatric disease or to distinguish the disease from other psychiatric diseases such as distinguishing schizophrenia from methamphetamine addiction. Parameter tuning is necessary to create optimum machine learning (ML) models that successfully produce solutions for classification or regression problems. ML methods such as random forest (RF) and support vector machine (SVM) are commonly used in prediction studies in both psychology and psychiatry literature for solving various complex problems. However, studies are not consistent in terms of ML methods since they may adopt different hyperparameter tuning strategies, or they may not report their use of the ML method. For example, some researchers may use autotuning ML methods while others may prefer designing the code by themselves without using default values of automatically designed ML methods. Thereby, it is important to identify and explain the methodologic aspects of the ML method to have a reproducible output.