Liquid Membrane-based Extraction of Arsenic: Part 2-Optimization through Statistical and Machine Learning Approach

This paper is a continuation of the work presented in Part 1 of this series. The aim of this Part 2 is to find optimal operating condition of the process through two comparative approaches, viz. statistical approach and machine learning based model. Three important aspects of the statistical analysis, viz. descriptive, correlational and inferential statistical analyses are performed in comprehensive manner. Associated tests such as Shapiro-Wilk test and Kolmogorov-Smirnov test are conducted to check for normality. The homogeneity of variances of the dependent variable with respect to the independent variable are checked through Levene’s test. The correlational analysis are studied using Spearmann’s test and Pearson’s correlational analysis along with standard ANOVA and/or MANOVA. Based on the main effects from the test of between-subjects obtained for the variables, post hoc (Tukey HSD) analyses are computed to understand the effect of individual and combined independent variables on the dependent variables. Artificial Neural Network has been adopted in machine learning model along with Genetic Algorithm based optimization tool to compare their performances with the obtained experimental and statistical data. The data points have been divided to train, test and validate the ANN based on maximum extraction\% and recovery\% with minimum mean squared error.