error rate optimization

The article outlines the development of neural network frameworks to counter future cyberattacks in Enterprise Resource Planning (ERP) systems. A lack of security can lead to major risks. Hence, an ERP system is more prone to cyberattacks. Many researchers have suggested the integration of big data analytics to counter cyberattacks. We have designed a neural network framework for binary classification to predict the attack classes in real-time. We have compared various types of neural networks to check which neural network is more effective and has less error in predicting cyberattacks. The findings revealed that out of the proposed methodologies, the ensemble of the Recurrent Neural Network as an autoencoder is the most effective design. We proposed three designs, and we have found that the ensemble of deep learning provides a 97.5% error rate. This design is most effective for big data architecture in the real-time pipeline of ERP. The trends of deep learning work on big data but face some practical issues in implementing the models. The study is beneficial for organizations using ERP, which is the largest ERP vendor and the most costly product widely used worldwide. Hence, it provides research in the field of cybersecurity and contributes to the latest technology approach redesign for practitioners. Deep learning methodologies face practical challenges in the real-time representation of any event. The relevance of the computing approach of the novel and deep learning model in practice is identified, and it is performed by the researchers.