Ferroelectric materials are of great interest because of their intriguing physical properties, such as their bi-stable polarization states and fast switching speed. Since reporting fluorite ferroelectrics in 2011, extensive studies have been performed for exploring and improving the ferroelectricity of these materials, for example, HfO2, Hf0.8Zr0.2O2, Al: HfO2, and ZrO2, because of their existing ferroelectricity even in less than 10？nm thin films. The issues with sensitivity and efficiency in traditional PFM methods and resonance-enhanced PFM techniques cannot be ignored. For example, if the piezoresponse is weak because of the ultra-thin material, it can be difficult to accurately analyze the resonance because of a low SNR. In the challenging noisy environments, the fitting results of the methods before can still be poor, eventually causing a misinterpretation. So far, the rapid fitting method capable of handling noise has remained challenging for extracting of resonance information. Therefore, it is highly desirable to explore an accurate and efficient approach to address the functional fitting issue. In this study, Panithan Sriboriboon et al from the School of Materials Science and Engineering, Sungkyunkwan University, demonstrated an approach based on deep neural network (DNN) hybrid with deep denoising autoencoder (DDA) and principal component analysis (PCA) for enhanced FM sensitivity. The DNN combines denoising elements without the need for further optimization by the least-squares (LS) method. The DDA and PCA were assigned for the noise reduction task, and then the DNN was applied to the denoised dataset. These noise-reducing elements directly addressed noisy outliers to improve the SNR, resulting in increased PFM sensitivity. These implemented workflows were validated using ferroelectric model samples of periodically poled lithium niobate (PPLN) and 10？nm-thick Hf0.5Zr0.5O2 (HZO) thin films. Compared to the traditional LS method and a previously reported DNN-LS method, the proposed workflows of DDA-DNN and PCA-DNN remarkably improved the SNR and PFM sensitivity. In particular, the results for HZO thin films presented the feasibility of exploring very weak piezoresponse. The PCA-DNN approach has successfully extracted the low piezoresponse, allowing switching events of the HZO thin films to be observed using a low excitation voltage. Besides, the approach reduced the time necessary for the evaluation of resonance information, providing a faster and more accurate resonance analysis, which especially important when fast feedback is required to the instrument to enable automated and autonomous experiments.