Services composition algorithm based on Neural Network in smarts environments

Abstract

The smart environment is an ecosystem where users constantly interact with smart objects (sensors, smartphones, devices, etc.) to improve their daily lives. This type of intelligence requires the cooperation of several smart devices offering different services to satisfy the increasing demand of the user, however, the user's requirements often exceed a single service which explains the great demand for the composition. With the rapid growth in the number of functionally equivalent services, Quality of Service (QoS) has become an important factor in distinguishing between functionally equivalent services. However, the challenge of selecting the best set of services to compose taking into account the global QoS constraints imposed by the user has become an NP-hard problem. In this thesis, we propose the Services Composition algorithm based on Neural Network (SC2N) to solve the QoS-aware services composition problem in the context of smart environments. The comparison results obtained in the simulation scenarios demonstrate that the SC2N algorithm is efficient in terms of composition time and utility thanks to its unique structure and its operators; the Bias phase ensures the diversity of the compositions from one generation to another, whereas the TF operator phase brings the compositions closer to the best composition in terms of the utility value.