Insurance Claims Disbursement Optimization via Federated Learning in ERP Modules
Main Article Content
Abstract
The ERP modules that this paper will be dealing with are Federated Learning (FL) to improve insurance claim disbursement. The information on insurance, particularly, is often personal, skewed, and available to various businesses, and thus, the conventional centrally oriented machine learning cannot occur with ease. Concurrent modeling capability ensures the privacy, as well as security guaranteed by federated Learning; the insurers do not receive access to raw information on customers directly. The review shows that FL is able to enhance claims loss modeling, detecting of zest falsification and model of expense in a better fashion. The study has demonstrated that the federated models prove superior to the oldfashioned frameworks in which the accuracy gains in federated models were noted to be approximately 10-15 percent in addition to reduction in the threat of privacy. Other optimization measures such as asynchronous updates, its good aggregation as well as lesser cost of communication also boost the extent of scalability. Fl can be used in detecting journeys to identify claims of fraud in a period that is longer, project larger malfunctions and model vehicles with forms of paying off concerning the ERP. The findings tell that federated learning can be a plausible solution that will enable the insurance companies to enhance the efficiency and safety of the completed claims settlements and make them more transparent, in response to the privacy requirements.