Paper Title
Effect of Feature Selection on Cloud Metaheuristic Machine Learning

Abstract
Data mining approaches and their associated optimization algorithms have made big data more attractive because they can be used to extract global information that supports the construction of reliable prediction models. We developed a novel and interactive cloud-computing tool, called the Cloud Metaheuristic Machine Learning System, which enables universal access for mining data through the internet without the need of any computer software. The web-based computing system can be trained to solve many kinds of regression problem, such as the prediction of the compressive strength of concretes. However, to increase predictive accuracy and computing efficiency for engineering purposes, the most relevant information should be selected. Irrelevant variables or features must be eliminated to yield the most useful result. Therefore, this work seeks to determine significant features that can be used to predict the outputs and to improve the accuracy of the previous full-mode prediction model without feature selection. In the preprocessing stage, stepwise regression is used to delete redundant or insignificant input variables. The result of the case study herein concerning the compressive strength of selected multi-country concretes after a standard curing of 28-days demonstrates that the current model outperforms the full-mode model with respect to accuracy and computing time. Index Terms - Civil Engineering Application, Construction Material, Feature Selection, Machine Learning, Web-based Analytics.