Feature Selection Based on Binary Tree Growth Algorithm Using Opposition-Based Learning

Data science and data mining were the sufferers owing to complicated data dimensionality problem and a rapid cumulating of data volume. The researchers of data-heavy areas are often demotivated by the high memory costs, poor precision performance, and high computational costs. How much of a computer’s capacity and memory is required for machine learning developments is greatly dependent upon the amount of training data that is involved. Therefore, reducing the number of factors which are important for learning as well as dimensionality of original data can be done with the help of feature selection. Feature selection is the key step in many intelligent and expert systems like in the identification of network breaks and prediction of disease. The main component was supplemented in the primal Tree Growth Algorithm (TGA) aiming to overcome its weakness however TGA could be applied to feature selection challenge. In the beginning period of BTGA, in contrast to the preceding method, OBL is being employed to further broaden the population variance such that all the best possible solutions will not be able to fit the specific conditions (OBL). In order to validate the effectiveness of the suggested artificial combinatorial technique, it was used to operate on three datasets observed at the UCI repository of machine learning datasets. The primary point of difference between the proposed algorithm OBL-BTGA and the initial algorithm BTGA is that the proposed algorithm provides better accuracy rate and needs fewer distinctive model features.