<p>This study aims to develop a hybrid modeling approach that combines Artificial Neural Networks and Multivariate Linear Regression using a Decision Tree to model the behaviors of a compliant robotic gripper effectively. The focus is on enhancing the accuracy of predictions regarding critical performance metrics such as stroke and frequency. Notably, this is one of the first studies to propose a hybrid learning method that combines Artificial Neural Networks and Multiple Linear Regression based on a Decision Tree. This approach leverages the advantages of both Artificial Neural Networks and Multiple Linear Regression to more accurately model the behavior of the compliant gripper. A finite element analysis is employed to collect data on the gripper’s performance. The proposed method utilizes Artificial Neural Networks and Multivariate Linear Regression, trained separately, with a Decision Tree determining the model used for different design scenarios. The proposed model addresses the drawbacks of existing ensemble learning models and improves prediction accuracy simultaneously. Specifically, the proposed model achieves a Root Mean Square Error of 0.377 Hz for frequency and 0.002 mm for stroke. Validation through experimental tests confirms the model’s reliability, showing minimal error margins. This research highlights the potential of integrating Artificial Neural Networks and Multivariate Linear Regression to improve modeling techniques for soft actuators, addressing significant gaps in the existing literature. The findings indicate that the proposed method is effective and adaptable for future applications in robotic manipulation.</p>