Forecasting KPIs of Production Systems Using LSTM Networks

Key Performance indicators (KPIs) are the bases of management, decision support and forecasting tools of production systems. They are monitored by companies to analyze and control their manufacturing processes. KPIs allow measuring how successful is an organisation towards a set of objectives to be achieved. Moreover, accurate forecasting of KPIs allows decisions to be redirected to ensure performance optimization while reducing costs and effort, and to anticipate potential disruptions in the production system. This paper aims at applying deep learning based on Long Short-Term Memory Networks (LSTM) to predict KPIs of a production system. Several configurations of a simulated production system are used to provide a large comparison allowing to judge the effectiveness of the proposed approach. The first results obtained in terms of prediction accuracy are promising in comparison with classical approaches.