High-frequency applications including magnetic write heads, inductors, transformers and sensors etc, demand thin films with tremendous soft magnetic properties. In this regard, the resonance frequency of the soft magnetic thin films is of crucial importance which has a direct relation with material properties. Meanwhile, machine learning (ML) has recently evolved as an efficient tool for material designing tasks including material-property relationship. For an ML algorithm to efficiently perform such tasks, feature engineering is implemented, in which a set of descriptors is designed that can best represent the data. Here, we employ the ML technique, combined with data augmentation and feature engineering, for the estimation of the resonance frequency (f r) value of thin films for high-frequency applications from different sets of theoretical and experimental features. Among the selected experimental features, when anisotropy field is combined with saturation magnetization, the ML model’s performance is improved, which is in line with the Kittel’s expression for f r estimation. Finally, it is shown that the data augmentation can improve the training score, and reduce the mean squared error as well as bias of the models. The outcomes of this work are of key importance in paving the way for ML implementation in material-property analysis and time and cost-effective high-frequency device designing.