Contribution of machine learning in continuous improvement processes

Purpose The aim of this study is to predict one of the key performance indicators used to improve continually production systems using machine learning techniques known by the ability to teach the machine to perform complex things as opposed to simple statistical methods by giving this machine the historical dataset, according to the kind of machine learning the authors will use, the machine will be able to predict a new output data from the input data given by the user. Design/methodology/approach This work is divided into six sections: In the first section, the state of art for OEE, machine learning, and regression models. In the second section, the methodology, followed by an experimental study conducted in an automotive company specialised in the manufacturing of manual transmissions. Findings The three models show a very high accuracy (higher than 99%), a comparison between these three models was done using three indicators, namely mean absolute error (MAE) mean square error (mean squared error and mean absolute percentage error which shows that the best model is the least angle followed by Bayesian Ridge and automatic relevance determination regression. Originality/value As the authors can see many works were done in the different production systems for prediction, the most relevant works were done to predict a parameter in the production system such as The prediction of part thickness in aluminium hot stamping process with partition temperature control the prediction of CO 2 trapping performance the prediction of crop yield the prediction of lean manufacturing in automotive parts industry the contribution of the work will be to use the machine learning techniques to predict the key performance indicator “used to measure manufacturing efficiency” which is the overall equipment effectiveness used in the authors’ case to measure the improvement of the production system.