Research on Parallelization of Logsf Feature Selection Algorithm

With the advent of the large data age, it is becoming more and more important to reduce the dimension of data. As a common means of data reduction preprocessing, feature selection has attracted more and more attention in recent years. At present, it is mainly based on the traditional feature selection algorithm, but the traditional method has been unable to meet the requirements of massive high-dimensional data processing. So, this paper studies the Logsf feature selection algorithm based on local learning, and improves the shortcomings of the existing algorithm. Then, by using the Mapreduce framework in the Hadoop platform, the improved Logsf feature selection algorithm is designed and implemented in parallel. Finally, the experimental results show that the improved Logsf feature selection algorithm for parallel processing is superior to the existing Logsf feature selection algorithm in the accuracy and the speedup. It fully shows that the characteristics of the parallel Logsf algorithm has a good reliability in this paper.