The uninterrupted incremental increase in the electricity demand sparked the need for diversified energy resources. The electricity sector must be supplied with smooth and stable energy sources. This necessitates the employment of energy storage systems (ESSs) to meet the surplus demand. However, meeting these demands requires maintaining a relatively high round-trip efficiency of the ESS. To enhance the system's efficiency, machine learning algorithms are integrated into the ESS. Machine learning algorithms have widely been incorporated with mature ESSs, such as battery storage systems. Nevertheless, they have not been extensively applied in compressed air energy storage (CAES) systems. In this work, machine learning algorithms are utilized to maximize the efficiency of the CAES experimental system. First, regression-supervised machine learning algorithms are used to validate the reliability of the data and generate accurate estimations of the system's efficiency. The developed linear regression model is found to outperform other regression models in terms of the determination coefficient (R2) value. Upon validating the reliability of the data, unsupervised machine learning algorithms, association rules, can be utilized to extract patterns from the data. These patterns show the parameter configuration that can be used to optimize the system's efficiency. These patterns are extracted to achieve the best modeling of the system. As a result, this work paves the way for developing an efficient predictive model that is able to accurately predict and optimize the CAES system's efficiency.