In order to more accurately realize the purpose of quantitative diagnosis of hydraulic pump faults, we investigated the methods of degradation feature extraction and fault degree diagnosis. We proposed spatial information entropy based on the study of permutation entropy. Then, we analyzed the influence of values variation of partition numbers, embedding dimension and delay time, which were used in the calculation of spatial information entropy, and the conclusion was meaningful for the selection of them. The results obtained by adopting the spatial information entropy algorithm to the simulation signal affirmed the availability and superiority of using spatial information entropy as the degradation feature of pump faults. We extracted the degradation feature set with the spatial information entropy algorithm, considering the nonlinear relationship between the degradation feature and fault degree. We optimized the parameters of the support vector regression machine with a fruit fly optimization algorithm, then used it to realize the quantitative diagnosis of hydraulic pump faults. The analysis results of the practical vibration signal verified that spatial information entropy performed better than did permutation entropy on the degradation state reflection. The results of the diagnosis model demonstrated the improved adaptability of spatial information entropy and the favorable performance of the proposed diagnosis model. The results also demonstrated the better optimizing ability of the fruit fly optimization algorithm compared with traditional algorithms.