Development of data driven models to accurately estimate density of fatty acid ethyl esters

Fatty acid ethyl esters (FAEEs) are widely used in biofuels, pharmaceuticals, and lubricants, offering an eco-friendly alternative due to their biodegradability and renewable nature, contributing to environmental sustainability. The objective of this study is to construct advanced predictive algorithms using various machine learning methods, including AdaBoost, Decision Trees, KNN, Random Forests, Ensemble Learning, CNN, and SVR. These models aim to accurately predict the density of FAEEs across different temperature, pressure, molar mass, and elemental composition (oxygen, carbon, and hydrogen content). Experimental data reported in earlier publications were employed to develop the models. Results indicate that the dataset is highly well-suited for developing reliable models based on data. Analysis reveals that temperature exerts a considerable influence on density, with pressure also playing a critical role. The reliability of the dataset, consisting of 1307 experimental datapoints gathered from the literature, was ensured through the application of a Monte Carlo outlier detection algorithm, which validated its suitability for model training and validation. Through extensive statistical evaluations and visualization techniques, SVR emerged as the most accurate model for density prediction. Sensitivity analysis confirms the influence of all input features, with SHAP analysis identifying temperature as the most dominant factor affecting density. The developed framework provides an economical and time-saving substitute for laboratory-based experimentation density measurements, enabling precise density estimation for FAEEs under various conditions.