Fuzzy Associative Classification Driven MapReduce Computing Solution for Effective Learning from Uncertain and Dynamic Big Data
Raghuram Bhukya, Jayadev Gyani · International Journal of Database Theory and Application · 2018
Handling uncertainty and dynamic changes in data sets supposed for analysis is always been challenging task for data analytics community.The same challenges even inherited to the embryonic big data analytics which are generally mentioned as veracity and velocity properties.Indeed, in case of big data, handling uncertainty and dynamism data could be more typical because of the scalability factor which is a result of data storage in distributed file system structure.In order to overcome difficulties of handling uncertainty and dynamic changes in big data analytics and considering efficiency provided by fuzzy associative classification techniques in handling uncertainty of data, we propose a dynamically scalable fuzzy associative classifier extraction model for Mapreduce framework.The important contributions of the paper is that it proposes a data driven fuzzy generalization approach for handling uncertainty, Tid-list based classification approach for easy scalable computation among multiple nodes and data chunks driven updating of classification model in case of dynamic changes to dataset with respect Map-reduce framework.The experimental evaluation results shows that the proposed Map-reduce model for fuzzy associative classification rule extraction can efficiently handle data uncertainty and dynamic changes to data stored in distributed file system, along with satisfying scalability factor.