Machine learning (ML) approach to modeling\udand predicting real-world dynamic system behaviours has\udreceived widespread research interest. While ML capability in\udapproximating any nonlinear or complex system is promising,\udit is often a black-box approach, which lacks the physical\udmeanings of the actual system structure and its parameters, as\udwell as their impacts on the system. This paper establishes a\udmodel to provide explanation on how system parameters affect\udits output(s), as such knowledge would lead to potential useful,\udinteresting and novel information. The paper builds on our\udprevious work in machine learning, and also combines an\udevolutionary artificial neural networks with sensitivity analysis\udto extract and validate key factors affecting the cloud data\udcenter energy performance. This provides an opportunity for\udsoftware analyst to design and develop energy-aware\udapplications and for Hadoop administrator to optimize the\udHadoop infrastructure by having Big Data partitioned in\udbigger chunks and shortening the time to complete MapReduce\udjobs.
展开▼
