Aiming at the problems of complex data features and difficult model parameter tuning in predicting the inspection cycle of rod pumping wells, this study proposes a hyperparameter optimization method for the Blending fusion model based on the Multi-strategy Improved Sparrow Search Algorithm (MISSA).Traditional fusion models often face challenges in fully optimizing parameters under complex data characteristics, leading to limited prediction accuracy. In contrast, MISSA enhances population diversity through Sobol sequence initialization and strengthens global search capability by integrating the sine–cosine strategy and Lévy flight mechanism. When applied to tune the hyperparameters of the Blending model (composed of Support Vector Machine, RBF-XGBoost, and BP neural network), this method effectively addresses the parameter tuning dilemma in complex scenarios and improves the model’s adaptability to nonlinear data. Using 2000 sets of production data from a Daqing oilfield as an example, results show the MISSA-optimized model reduces RMSE and MAE by 32.36% and 28.60% compared to traditional fusion models, with R2 reaching 94.15%. This study provides an efficient hyperparameter tuning framework for complex industrial prediction tasks, supporting intelligent decision-making in oilfield production processes.