In the field of seismic retrofitting, a recent innovation involves incorporating Cross-Laminated Timber (CLT) panels as infills within Reinforced Concrete (RC) building structures. These panels, connected using metal dowel-type fasteners, aim to enhance the in-plane response of RC frames. Despite emerging scholarly interest, comprehensive quantification of the mechanical behavior of this technique and the development of simplified macro-models for structural analyses remain underexplored. This study introduces a novel framework for establishing the stress-strain relationship for equivalent strut models of CLT panels infilled in moment-resisting RC frames. High-fidelity models of RC frames with CLT infills are developed using a parametric Finite Element (FE) approach in OpenSeesPy. The RC frame is represented with fiber-section discretization in displacement beam-column elements, and the CLT panels are modeled as orthotropic elastic shells with metal dowel-type fasteners represented as zero-length elements. A comprehensive parametric study using Latin Hypercube Sampling was conducted to generate 8,000 different frame configurations, varying parameters describing both the frame and the infill. Calibration of the stress-strain curves was conducted using a genetic algorithm-based optimization process, aimed at minimizing the mean squared error (MSE) between the global force-displacement curves of the high-fidelity model and the equivalent strut model. Optimal parameters found by the genetic algorithm for the material properties assigned to the strut were used to derive a predictive model. Symbolic regression identified mathematical expressions for the elastic modulus of the struts, yielding satisfactory performance in the predictions. Building on this, the XGBoost machine learning algorithm was employed to develop a more comprehensive predictive model based on 16 input variables, enhancing the accuracy of the parameter estimation for the stress-strain curve of the struts. Additionally, Shapley analysis was used to interpret the contributions of each input variable, providing insights into their relative importance in predicting the properties of the stress-strain curve.