Cloud-Enabled Intelligent Testing of Robotic Process Automation Systems Using Dynamic BERT-LSTM and FCNN Fusion Models

Robotic Process Automation (RPA) tools have the potential to impact business operations significantly in the fields of finance, healthcare, logistics and client support services. Moreover, there is special value in the healthcare, finance and logistics sectors, since healthcare is a complex sector, as there is an elaborate array of tasks that need to be completed for automation. Furthermore, the chances of errors decrease significantly and efficiency improves. Similarly, in healthcare, RPA tools can manage patient data as well as claim management and in finance, RPA can improve reconciliation and reporting. Considering all these benefits, RPA is compulsory to the effective functioning of these industries. As RPA systems grow in complexity, intelligent, scalable and context-aware testing solutions become a necessity of the hour. Static embedding-based, keyword frequency-based or manual rule-based testing approaches all falter when it comes to semantic understanding and dynamic test environments, resulting in inaccurate priority assignment, poor duplicate detection and poor scalability. A cloud-based intelligent testing paradigm based on the integration of Bidirectional Encoder Representations from Transformers (BERT), Long Short-Term Memory networks and Fully Connected Neural Networks is proposed. The methodology proposed consists of preprocessing semi-structured RPA test-case datasets, retrieving contextual semantics through BERT, temporal pattern modelling through LSTM and FCNN for final classification and anomaly detection. This combined system is run on a scalable cloud platform (e.g., AWS, Azure) to facilitate enterprise-scale continuous integration and parallel testing. Performance was measured on a Kaggle test-case dataset in comparison to standard methods. Findings show higher performance on several testing utility metrics of 95% improvement in test coverage, 85% decrease in test running time, 82% redundancy reduction, 90% gain in automation coverage and 98% defect detection effectiveness. These results validate that the BERT-LSTM-FCNN fusion model not only improves RPA test quality and operational reliability but also greatly lowers human effort to maintain large-scale automation systems.