The fabrication of YBa₂Cu₃O_7-δ (YBCO) high-temperature superconducting thin films is governed by multiple process parameters. However, the accumulation of large-scale experimental data in actual thin-film growth processes remains challenging. This study utilizes a small-sample experimental dataset to examine the guiding role of a neural network model in optimizing the pulsed laser deposition (PLD) processes for YBCO thin films. The dataset encompasses ten key process parameters, including substrate type, substrate size, laser focal length, target-substrate distance, output voltage, laser frequency, oxygen pressure, growth time, substrate temperature, and annealing time, which serve as the model inputs. The YBCO (005) peak position and the full width at half maximum (FWHM) of the rocking curve, obtained from X-ray diffraction (XRD), are the structural parameters to be predicted and constitute the model outputs. A multilayer perceptron (MLP) neural network model was constructed, incorporating bootstrap resampling and early stopping mechanisms to mitigate overfitting under small-sample conditions. The model yields reasonable predictions of the peak position and the FWHM, and demonstrates favorable generalization performance on an independent test set, thereby validating the feasibility of data-driven approaches for modeling complex thin-film processes. Notably, the root-mean-square error (RMSE) for the (005) peak position prediction is as low as 0.01°, approaching the instrumental resolution of XRD measurements (±0.005°). On this basis, Pearson correlation analysis is further conducted to perform a preliminary exploration and visual assessment of the complex interrelations between the process parameters and the structural features. Specifically, the peak position, which reflects lattice constant variations, is primarily regulated by oxygen content and macroscopic stress, exhibiting a relatively clear relationship. In contrast, the FWHM, which reflects crystalline quality, is governed by complex growth kinetics and exhibits a more sensitive and highly nonlinear dependence on the process parameters. This work establishes a scalable and transferable analytical framework for superconducting thin-film experiments under small-sample constraints. It paves a feasible pathway for data-driven process optimization in resource-limited circumstances, and facilitates comprehensive in-depth interpretation of process-structure-property correlations.