Any successful model of coronal heating in non-flaring active region core loops must be able to re-produce the full range of observational signatures. These signatures, or observables, include, but are not limited to, the slope of the emission measure distribution below the peak temperature in log-log space, the time delay which maximizes the cross-correlation between narrow-band intensities, and the presence of "very hot" (>8 MK) plasma. Quantitatively assessing agreement between models and observations in the context of these observables is critical to constraining the parameter space of possible coronal heating mechanisms. In this talk, I will discuss the importance of forward modeling and machine learning in making meaningful comparisons between observations and simulations. In particular, I will highlight the results of a recent investigation into the frequency of energy deposition in active region NOAA 11158 and discuss how machine learning, specifically random forest classification, is used to assess agreement between observations and forward models of low-, intermediate-,and high-frequency heating.