Service-focused wireless computer applications rely on the connecting infrastructure to assist different end-user applications. Seamless application support relies on infrastructure availability and swift discovery of the services. In this article, service discovery based on predictive availability (SD-PA) is introduced. This proposed technique relies on the operating and available occurrences of the infrastructure to provide reliable service discovery. A predictive machine learning technique is used in this service discovery process to mitigate failures in random infrastructures. The response and discovery of applications and services in the IoT environment are balanced using the predictive discoveries of finite infrastructures. The performance of the proposed technique is assessed using the metrics of discovery time, service availability, outage probability, and failures.