FUZZY ANALYTICAL HIERARCHY PROCESS (FAHP) USING GEOMETRIC MEAN METHOD TO SELECT BEST PROCESSING FRAMEWORK ADEQUATE TO BIG DATA

摘要

Big data is considered a hotspot, as all organizations realize the importance of their big data to gain insights that help organizations develop and understand consumer requirements. Big data needs large storage capacity and strong processing frameworks in order to clean, process, and analyze it. Fortunately, cloud computing offers many services and processing frameworks that facilitate the storage and processing of big data. But the issue here is how to choose the best suited processing framework for big data of financial services. The best processing framework is chosen based on big data criteria and financial services requirements. We used MCDM methods to solve this decision problem and evaluated five big data processing frameworks (Spark, Hadoop, Flink, Storm, and Samza) based on twelve criteria. These criteria were collected from previous researches. Analytical hierarchy process (AHP) is a powerful and simple method of MCDM methods, but many of researchers believe that it has some weakness due to some uncertainty issues. Many researchers have preferred to combine fuzzy set theory with AHP to solve the uncertainty problem. This work introduces fuzzy AHP using geometric mean method in cloud service selection based-problem. The results show that Hadoop framework has the highest level of security, availability, scalability,and compatability, but has the heighest cost. Spark has the highest storage capacity , speed, and best processing mode at lower cost. Flink has the best usability, and processing. Storm has the best performance, sustainability, and is the cheapest. The validity of our results and the robustness of our hybird proposal were aproved by applying sensitivity analysis.