High Power Pulse Magnetron Sputtering (HPPMS) processes are characterized by high peak powers and peak voltages. This results in a high fraction of ionized metal ions within the coating plasma. In order to analyze the correlations between parameters of the coating unit and the intensities of the excited and ionized plasma species, optical emission spectroscopy (OES) can be used. In those experiments, several process parameters are varied in a single coating process. Currently, the prediction of plasma parameters based on coating unit data follows deterministic models which cannot describe the complexity in total. Therefore, not all correlations can be fully understood. Artificial neural networks (ANN) can be used to identify correlations between process parameters and plasma species. This enables prediction of OES data based on data of the coating machine. In the present study different coating processes containing the elements Al, Cr, Ti, N and O were investigated. Current and voltages of the cathodes, substrate bias, chamber pressure, gas flows and the target compositions were used as input parameter of the ANN. Time resolved OES data of metal and gas species were used as output data. To determine the most appropriate training algorithm for the current predictions, multiple algorithms were employed. A good prediction accuracy of OES intensity ratios based on coating unit data for industrial scale coating processes for TiAlN was obtained. For CrAlON the prediction of the intensity ratios of the gas species showed good results. Nowadays a high amount of coating machine parameter variations in PVD processes is needed in order to achieve tailored coating parameters. By using plasma diagnostics, such as OES, cost intensive coating deposition processes can be reduced significantly. A further shortening of process development time is possible by using ANN, since plasma compositions can be determined based on coating unit data.