Aims: When cardiac tissue is paced at very fast periods, it can develop a period doubling bifurcation called alternans where the propagating electrical waves are not constant anymore and instead, its wavelength varies (alternates) from short to long values as it propagates, which can facilitate arrhythmia induction. This well-known arrhythmic mechanism can be studied with cardiac cell models. However, the number of degrees of freedom required to accurately investigate this phenomena in space with these mathematical models can be very large and thus complex to understand. Selecting specific observables and using symmetry reduction methods it is possible to greatly reduce the complexity of the system to just a few key degrees of freedom. Methods: we illustrate the reduction of two generic models for cardiac tissue exhibiting various levels of alternans. These model reductions are based on data-driven proper orthogonal decomposition (POD) and sparse linear regression methods. Results: We simulate the models in one dimension and calculate the maximum amount of action potential duration for a large range of pacing periods. By going to the moving frame, we calculate a number of basis vectors by POD that characterizes the system and describes the global behavior wile reducing its dimensionality. This way we reduce the large system of partial differnetial equations to just a pair of ordinary differential equations (ODEs) that can reproduce/predict the APD alternans for all pacing periods as shown in the figure. Conclusions: We show that it is possible to reduce the dimensionality of cardiac models when investigating alternans from about two thousand degrees of freedom to just two. The resulting system of two ODEs have enough information to closely reproduce the dynamics and bifurcations to alternans obtained from the full simulations given by the whole PDE system.