Multi-sensor information fusion aids different services to meet the application requirements through independent and joint data assimilation. The role of multiple sensors in smart connected applications helps to improve their efficiency regardless of the users. However, the assimilation of different information is subject to resource and time constraints at the time of application response. This results in partial fulfillment of the application services, and hence, this article introduces a service selective information fusion processing (SSIFP) scheme. The proposed scheme identifies service-specific sensor information for satisfying the application service demands. The identification process is eased with deep recurrent learning in determining the level of sensor information fusion. This level identification reduces the unavailability of services (resource constraint) and delays in application services (time constraint). Through this identification, the applications' precise demands are detected, and selective fusion is performed to mitigate the issues above. The proposed system's performance is verified using the metrics delay, fusion rate, service loss, and backlogs.