Implementation of AdaBoost and genetic algorithm machine learning models in prediction of adsorption capacity of nanocomposite materials

Abstract

Simulation of adsorption capacity of a nanocomposite material was performed in this study in order to save time and cost in performing adsorption experiments. By developing the predictive models, one can easily predict the adsorption capacity of materials in wide range of processing conditions. Here, we developed two machine learning techniques for correlation of adsorption parameters including adsorbent dosage and solution pH to the adsorption capacity of the adsorbent. One of the most important issues and concerns in issues like the one we have in this study is ensuring the generality of output models, given that the dataset is a small one. As a result of this issue, we chose simple models to improve the modeling result in this study using AdaBoost and GA optimization models. The data was collected from literature and used in the fitting and simulations. After optimizing the hyper-parameters of these models, the final results showed very high efficiency. Using the RMSE criteria, linear regression, Bayesian ridge regression, and Huber regression received error rates of 2.5, 2.3, and 1.7, respectively.