Statistical optimization of forecast data from Adaptive Boosting and Support Vector Machine Algorithms

Abstract

This study aimed to calculate the optimum values of estimation data based on adaptive boosting (AB) and support vector machine (SVM) algorithms from machine learning (ML) models with statistical optimization models. Three independent and two dependent variables were used in this study. It was arranged for ML algorithms using 750 data of each variable. The training and testing phases in ML algorithms were set at a rate of 90% and 10%, respectively. The RMSE, MSE, and MAE values, which are the error rates, and the coefficient of Determination R2 values, were compared to verify the validity of the ML algorithms. The estimation results of the independent variables were analyzed with a nonlinear optimization model. The results obtained were validated with a high degree of desirability and the validity of the optimization model. AB algorithm provided the best performance for y1 and y2 dependent variables. The desirability degree of the optimization model of the variables y1 and y2 was calculated as 0.945. Based on the AB algorithm, the optimum value of the y1 and y2 variables were computed at 6.89 and 0.6169, respectively. The optimum values of the x1, x2, and x3 independent variables for both optimization models were calculated as 3.729, 0.509, and 13.814, respectively. As a result, the desirability values of the optimum values of ML models were calculated, and the validity of the optimum values of the optimum and actual data was verified in this study.