The thermal boundary conductance at the nanoscale solid-liquid interface exhibits several challenges in modeling and prediction. A large number of variables, mutually correlated or independent, influence the transmission of interfacial heat carriers at nanoscale interfaces. Conventional approaches hold limitations to study the phenomenon with the complexity involved in handling the simultaneous influence of multiple parameters. Most empirical correlations developed so far are limited to a narrow parametric range and lacked applicability in a universal setting. In this context, we propose investigating the contribution of these variables in high-dimensional feature space using machine learning and developing a predictive model with simulation data. An exhaustive list of structural and thermodynamic variables is selected as predictors for different nanochannels. We primarily investigate 2D semiconductors like graphene, boron nitride, and transition metal dichalcogenides forming interfaces with organic and ionic liquids. Molecular dynamics is an effective tool to parametrize the predictors and generate data. Secondly, understanding the contribution of different variables is vital to explain the physics of carrier transmission. With the impurity reduction-based feature-significance techniques, we investigate the governing predictors in most high-performing algorithms. The density depletion, work of adhesion, and system geometry are historically known to affect the value of thermal boundary resistance. We will quantify their contributions to gain a deeper insight into the fundamental physics of the phenomenon. The machine learning algorithm is expected to significantly enhance prediction accuracy compared to conventional empirical correlations derived from experimental and simulation data.