A three-phase decision making approach for self-adaptive systems using web services

A self-adaptive system adapts itself to changes in a dynamic environment. The core in self-adaptive systems is making an adaptation decision. The current practice focuses on a single layer of decision making using either a local knowledge base or a shared knowledge base shared by multiple units through a network. While the use of a local knowledge base is efficient, it suffers from its limited maturity. A shared knowledge base can address the maturity problem, but it is inefficient in adaptation due to communication overheads. In this work, we present a three-phase decision making approach for self-adaptive systems to improve precision while being competitive for efficiency. The approach consists of three phases for making a decision. The first phase uses the local knowledge base of the self-adaptive unit to identify an object. If the object cannot be identified locally, the unit sends a request to shared knowledge bases through web services in the second and third phases. The approach makes use of B-kNN for object identification and web services for accessing shared knowledge bases. We conducted quantitative validation in terms of accuracy, precision, recall, and F-measure using a set of scenarios. The results show that 99.58% of accuracy, 94.01% of recall, 94.04% of precision, and 94.01% of F-measure can be achieved. We also conducted comparative analysis by comparing the presented approach with the traditional approach and the cloud-based approach. The results show that the presented approach improves 45% in object identification with an increase of 0.66 s over the traditional approach and the same performance in object identification with a decrease of 0.95 s over the cloud-based approach.