Abstract
Educational attainment plays a critical role in societal outcomes. It affects employment prospects, economic development, and social mobility. However, the prediction of educational achieve- ment remains difficult due to complex socioeconomic and regional factors. Many previous studies fo- cus mainly on academic variables. The combined effects of demographic, economic, and geographical factors are often ignored. In addition, many studies rely on single learning models with limited gen- eralizability. To address these gaps, a novel meta–model blending approach is proposed in this study. The approach combines the strengths of a Deep Neural Network (DNN) and Gradient Boosting (GB). The aim is to improve the accuracy and robustness of educational attainment prediction. A prepro- cessing pipeline is applied before model development. Noise and missing values are handled. Feature refinement is performed. Min–Max normalization is also applied. After these steps, a hybrid ensemble model is constructed. The model is evaluated with a dataset of 1100 samples. The dataset includes 22 socioeconomic and regional features from towns in England. The proposed approach presents high performance with R2 = 0.9975, Mean Absolute Error (MAE) = 0.1582, and Root Mean Squared Error (RMSE) = 0.2058, which exceeds that of single DNN, single GB, and simple blending methods.