ABSTRACT
Micellization phenomenon occurs in natural and technical processes, necessitating the need to develop predictive models capable of predicting self-assembly behavior of surfactants. A least squares support vector machine (LSSVM) based quantitative structure property relationships (QSPR) model is developed in order to predict critical micelle concentration (CMC) for sugar-based surfactants. Model development is based on training and validating a predictive LSSVM strategy using a comprehensive data base consisting of 83 sugar-based surfactants. Model's reliability and robustness has been evaluated using different visual and statistical parameters, revealing its great predictive capabilities. Results are also compared to previously reported best multi-linear regression (BMLR) based QSPR and group contribution based models, showing better performance of the proposed LSSVM-based QSPR model regarding lower RMSE value of 0.023 compared to the group contribution based and the best results from BMLR-based QSPR.
