This data was used in the machine learning analysis of the Cosmic Microwave Background data in: https://github.com/IndiraOcampo/CMB_ML_based_model_selection.git and https://dx.doi.org/10.1088/1475-7516/2025/02/004 The objective is to train a neural network architecture on the different polarization modes (TT, TE, EE and joint) to perform model selection between the standard cosmological model, ΛCDM and a model that introduces a Feature Template (FT) in the primordial power spectrum - related to the early Universe physics. The first row corresponds to the multipole moment "\ell" and the remaining ones correspond to the different components of the Cl's angular power spectrum, for the different values of A_lin (the feature oscilation parameter). While A_0 = 10^-2 is a reasonable value that still agrees with observations, A_0 = 0 corresponds to the ΛCDM model. Finally, our aim is to apply SHAP to perform feature importance (interpretability) in our results.