The strength prediction of adhesive-bonded single lap joints (SLJs) is essential for reliable joint design. However, due to the non-linear interactions between geometrical parameters and material properties, accurate prediction remains challenging. In this study, a hybrid modeling approach combining Taguchi experimental design, multiple linear regression (MLR) and artificial neural networks (ANN) is proposed to predict the failure load of SLJs made from glass fiber-reinforced polyester composite adherends. By using a Taguchi L9 orthogonal array experimental matrix, the effect of overlap length, adhesive thickness and material property on joint strength was investigated. From the experimental results, it is observed that overlap length is the most influential factor, contributing 46.9% to the total variation. The experimental failure load data obtained from the tests were used to develop predictive models. Initially, MLR model was developed, yielding a coefficient of determination (R2) = 89.37%. To address the limitation of a small dataset, for ANN training, the MLR model was used to generate additional data, increasing the dataset size from 9 to 27 samples. The Levenberg-Marquardt algorithm was used to train a feedforward neural network. The ANN model demonstrated superior accuracy with the mean absolute percent error (MAPE) of 2.14% as compared to 5.44% for the MLR model. The improved performance is due to the ANN's ability to capture non-linear relationships that the linear MLR model could not fully represent. This study offers an efficient methodology for improving the design and reliability of adhesive-bonded composite joints.