Machine Learning Approach for Big Data Classification

Nowadays, the database size increases rapidly. So the traditional techniques such as data mining cannot be directly used for solving any problem based on a big database. This paper proposed a new approach depending on applying big data to two different big data sets which are Covtype, Covtype-2, Poker, and Poker-2. This approach consists of four phases namely; a preprocessing phase which uses fitting algorithms, the second phase uses map reduce algorithm, the third one based on two traditional techniques which are KNN and FKNN and the last phase is reduced phase which uses ANN to produce the total accuracy rate. The experimental result of this approach shows that the accuracy rate is 71% compared with the traditional techniques. The traditional algorithms is applied on big data. The K-nearest neighbor, Fuzzy K-Nearest Neighbor and the neural network algorithms are introduced using the map-reduce paradigm to find ways of optimizing and enhancing the efficiency of data mining techniques for classification process on big data.