Feature selection using fish swarm optimization and shuffled frog leaping algorithm for classification in sentiment analysis

Abstract

The analytics of Big Data reflects all challenges faced by data, which may be unstructured, vast and fast-moving for the traditional methods to manage. Data can be generated from the research institutions, other businesses, governments, and other organizations in a routine manner with unprecedented scope and complexity. The gleaning of meaningful information, along with the competitive advantages derived from the massive data, has today become quite important to all the organizations globally. Making an effect of extracting all meaningful insights efficiently from the sources of data fast and easily is quite challenging. There has been a great improvement in the tools that are available for managing the variety, volume, and velocity of Big Data. newlineTo help sort the problem of high dimensional search space of Big Data during analysis, a lightweight selection of feature is required. The feature selection identifies the feature subset that is related mostly to its target class. The objective in feature selection will be to eliminate the redundant and irrelevant data from the big data for increasing the accuracy level and also for the reduction of dimensionality in data. The present methods of feature selection may be grouped into two as (i) The Filter and (ii) The Wrapper in accordance to if the criteria of subset evaluation are dependent on the algorithm and its subsequent classification: newlineThese filter methods use the process of selection of features, and their machine learning method is considered and a selection based on the trait of the dataset is used. There is also a fast operation leading to its poor performance newline