Hyper-Parameter Selection of Machine Learning Models for Predicting the State of a Technical Object

To decrease emergency response time in the operation process of a technical object, a timely response to its functional failure is necessary. Hence, there is a problem of predicting the state of a technical object and possible violations during its operation based on the results of studying the parameter set for object operation. Machine learning methods can be used to solve this problem. The article also considers random forest and support vector machine as machine learning models. However, a machine learning model depends on hyper-parameters that must be tuned before the training process, for the model to function correctly and maintain high accuracy in predicting the object state. The paper discusses three methods for hyper-parameter selection: grid search, random search and Bayesian optimization. The authors evaluated a hydraulic valve to study the effectiveness of methods for selecting hyper-parameters of machine learning models for predicting the state of a technical object.