Adaptive youth self-leadership modelling using optimized graph neural networks

Abstract

Background The complex interaction of socio-economic, educational, and demographic factors makes it challenging to accurately predict youth self-leadership confidence.Objective This study aims to develop an effective deep learning framework for predicting self-leadership confidence among young individuals by identifying relevant influencing factors and modeling their complex relationships.Methodology A dataset of 303 young individuals was collected using age, gender, socio-economic status, parental and personal education, support system, leadership position, and self-leadership confidence measured on a 1–5 scale. Missing values were handled and the data were normalized using a hybrid Fuzzy Rough Sets and Fuzzy Min-Max Neural Network (FRS-FMNN). Relevant features were selected using the Orchard Greater Cane Ratio Algorithm (OGCRA).Proposed Model A Feedback Deformable Graph Convolutional Network (FDGCN) was developed to model complex feature interactions, while the Wombat Optimization Algorithm (WOA) optimized its hyperparameters.Results: The proposed FDGCN-WOA achieved 99.95% accuracy, 99.60% precision, 99.45% F1-score, and 99.40% specificity, with RMSE of 4.85% and MAE of 3.12.Conclusion The results demonstrate the effectiveness of FDGCN-WOA for accurately predicting youth self-leadership confidence.