Design choice and machine learning model performances

An increasing number of publications present the joint application design experiments (DOE) and machine learning (ML) as a methodology to collect analyze data on specific industrial phenomenon. However, literature shows that choice for collection model analysis is often not driven by statistical or algorithmic advantages, thus there lack studies which provide guidelines what designs ML models jointly use analysis. This article discusses in relation performances. A study conducted considers 12 experimental designs, seven families predictive models, test functions emulate physical processes, eight noise settings, both homoscedastic heteroscedastic. The results research can have an immediate impact work practitioners, providing practical applications DOE ML.