Data Reduction Using Feature Reduction Technique and Particle Swarm Optimization

Abstract

Feature reduction, commonly referred to as dimensionality reduction, is the process of minimizing the number of features involved in a computation-heavy task without sacrificing crucial information. By decreasing the number of features, the computational load is reduced, enabling more efficient and faster processing. Dimensionality reduction techniques are categorized into two primary types: feature selection and feature extraction. Feature selection focuses on identifying the most relevant features, while feature extraction transforms existing features into a new, reduced set that still retains the essential information. Feature selection - it is where naturally or physically chose which contribute a large portion of the reduction variable or output. In other word, it is a process in which a superior set of features as the best subset is selected. There are three benefits such as reduces over- fitting, improving accuracy, reduces training time. Feature selection is implemented using three techniques such as Wrapper, Filter and Embedded. Wrappers evaluate a specific model sequentially using different potential subsets of features to get the subset that best works in the end. They are profoundly expensive and have a high possibility of over-fitting. Channels techniques are quicker elective that don’t test a specific calculation, however rank the first highlights as per their relationship with the issue and simply select the highest point of them. It is a statistical test used to assess the independence of variables, determining whether or not there is a significant dependency between them. This method helps identify relationships or associations between variables by evaluating if the observed distribution of data deviates from what would be expected under the assumption of independence. Few techniques in this category includes Correlation coefficients: removes duplicate features, Information gain or mutual information. A detail discussion on advantages and disadvantages of different filters and wrapper approach for feature reduction is going to be highlighted. Feature selection is a significant information pre-handling procedure, yet it’s anything but a troublesome issue due basically to the large search space. Particle swarm optimization (PSO) is a highly effective evolutionary computation technique. Be that as it may, the conventional individual best and global best refreshing component in PSO limits its presentation for highlight choice and the capability of PSO for feature selection has not been completely explored. Evolutionary computation (EC) strategies are notable for their global accessibility. Particle Swarm Optimization (PSO) is a relatively recent evolutionary computation (EC) method, known for being more computationally affordable compared to many other EC algorithms. In this manner, PSO has been utilized as a powerful method in highlight choice.