A neural network weight determination model designed uniquely for small data set learning

Environment characteristics are dynamic and changeable. In customized or flexible manufacturing systems, the collected data used for analysis is often small. There are many studies on small data set problems. However, most papers attack the problem by developing data pre-treatment methods which normally require abstruse mathematical knowledge, deterring engineers from applying the methods in practice. This paper develops a unique neural network to accurately predict small data sets. This neural network is developed based on the concept of the data central location tracking method (CLTM) to determine net weights as the learning rules. It not only makes accurate forecasts using small data sets but it also facilitates knowledge learning for engineers.