<p>Industry 4.0 aims to integrate intelligent systems in the machineries of manufacturing and service processes. This in turn helps efficient monitoring of production processes as well as maintenance of the machines. Use of different sensors enables substantial production data collection which in-turn can be used for different prediction mechanisms that can help towards early assessment of failure. A potential prediction system can be immensely helpful to avoid sudden breakdowns, timely machine maintenance and maximization of profit by minimizing the down time. Nevertheless the accuracy of the prediction is of prime importance and corresponding risk factors can guarantee for the correct maintenance decisions. At the same point of time the recent trends of internet of things (IoT) and entrepreneur resource planning (ERP) systems enable to integrate production data into management information systems (MIS). Hence proper prediction can reflect the real time press condition remotely to the management and even proper corrective actions can be initiated remotely. This paper presents a machine learning based approach for early prediction of failure in print production house. The experiments have been performed on real-time actual newspaper publishing house where the cost of press failure has immense effect on deadline. The potential of deep neural network (DNN) based prediction mechanism has as well been explored along with some of the conventional machine learning models. An attempt to integrate the proposed system with the printing workflow management system (WMS) has as well been presented. The results show that the proposed model can be a promising development towards failure prediction of printing house in terms of accuracy and efficiency.</p>