ERP-based Quality Prediction using Long Short-Term Memory Networks Approach for Bio industry

The biotech industry is relying on ERP (Enterprise Resource Planning) systems to regulate production, resources, and quality information. However, it is very difficult to predict the quality of products because of the time-related and complexity of ERP records. The conventional types of regressions fail to capture the temporal relationship and thus achieve lower accuracies (predictive accuracies). To overcome this, the proposed framework uses a Long Short-Term Memory (LSTM)-based model to predict the quality. The data used in this article is the ERP-MTD (Enterprise Resource Planning - Manufacturing and Test Data) dataset, which combines ERP modules such as batch process logs, sensor readings, and quality inspection reports to form a comprehensive and structured record of bio-production data. Preprocessing will be performed, which involves Min-Max scaling, where the values of features are normalized to a specified range, leading to consistency in training and faster convergence. A sliding window application is used to obtain a temporal perspective and sequential context, focusing on temporal tendencies that are crucial for assessing bio-production quality. The classification and prediction services utilise the LSTM network to capture long-term dependencies across the variables of the process and the quality outputs. The predictive ability of the model is compared based on Accuracy of 96.47%, Precision of 91.8%, Recall of 92.6%, F1-Score of 92.2%, which yields a better result than the baseline techniques.