This study proposes a comparison of state-of-charge estimation utilizing machine-learning classification to address the current limitations in battery management systems for zinc-air batteries. This objective is pursued by means of an analysis of features including impedance spectra. The examination covers four machine learning models, namely Naive Bayes, distance-weighted k-Nearest Neighbors, Decision Tree, and Support Vector Classification with different kernels. The performance of these algorithms is evaluated in comparison to a baseline scenario. The input features utilized by these algorithms include measurements of voltage, current, temperature, and complex impedance across various frequencies, along with additional extracted features that were evaluated.