An Improved Optimization Techniques For High Dimensions Data Analysis

Abstract

Modern data analytics systems have many potential real-time applications, such as managing student grades, reading tax returns, processing zip codes and checks, and analyzing medical data. In general, multivariate sequence overlap, separation, concatenation, feature selection, and feature extraction for data with multivariate sequence length have been analyzed by various researchers. Previous methods are planned to predict unconstrained numbers that include all four preprocessing phases’ beforehand classification. The most difficult task of this predictive model is feature extraction and classification. These phases invented to offer improved classification accuracy. Since the number and strings of data are usually unknown, determining the optimal boundary between them is very complicated. Some existing feature extraction methods rely on heuristic models inspired by nature to generate latent features for classification. Therefore, the optimal choice of features is difficult to predict due to the variability of features. Some common methods in feature extraction improved the classification results and produced the best results. The method consists of three basic modules. Dataset acquisition, classification module as Convolutional Neural Network (CNN) and optimization module (squirrel search optimization). Likewise, performance metrics such as precision, recall, precision and execution time are evaluated and compared to various existing methods.