Perfomant API for Bayesian Networks in Java

This study presents the implementation of a REST API for predicting emergency situations at cellular network base stations using Bayesian networks. Unlike most studies limited to theoretical models, this work offers an engineering solution implemented in Java using the jSMILE library and the Spring Boot framework. The choice of Java is moti­vated by its mature and extensive ecosystem, which emphasizes security, reduced development time, and support for microservice architecture. This approach eliminates a common bottleneck in model implementation—significant resource consumption during Bayesian network initialization—which negatively affects request processing time. To address this issue, the Object Pool design pattern was introduced, significantly improving system performance. The application of the Object Pool pattern to a Bayesian network is done for the first time, representing the scientific novelty of this research. Load testing conducted using Locust allowed for the evaluation of system resilience under load, response time measurements, and monitoring the number of pool objects used at different load levels.