To solve the complexity and blindness of the tuning process in the manufacture of microwave cavity filters, this study proposes a self-learning tuning method based on the poles and residues of the admittance function. First, the improved Cauchy's method based on the differential evolution algorithm is used to extract the poles and residues of the admittance parameters (Y-parameters) in a non-ideal environment, and the effect of different port phase shifts and cavity losses on the accuracy of parameter extraction is overcome. Second, a parametric model based on the experience and data fusion via the fuzzy neural network method is established according to the collected non-linear relation data. Furthermore, problems such as poor data reliability, low modelling accuracy and weak generalisation ability are solved. On this basis, an adaptive optimisation tuning of microwave cavity filters using an implicit space-mapping algorithm is proposed and problems such as convergence difficulty and dependence on the initial value are solved. The results of the online simulation show that the proposed method has a high tuning accuracy and fast tuning ability.This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.