An evaluation of regression algorithms performance for the chemical process of naphthalene sublimation

Abstract

Different regression algorithms are applied for predicting the sublimation rate of naphthalene in various working conditions: time, temperature, trainer rate and shape of the sample. The original Large Margin Nearest Neighbor Regression (LMNNR) algorithm is applied and its performance is compared to other well-established regression algorithms, such as support vector regression, multilayer perceptron neural networks, classical k-nearest neighbor, random forest, and others. The experimental results obtained show that the LMNNR algorithm provides better results than the other regression algorithms.