Improved Genetic Algorithm and Transformer-based Framework for Adaptive Online English Course Teaching

Advancements in online English course teaching have enabled flexible and accessible learning. However, current machine-learning-based systems struggle to efficiently optimize assessments and accurately predict student outcomes. Using the University of California, Irvine (UCI) student performance dataset, our proposed Improved Genetic Algorithm (IGA)Trans framework begins with robust preprocessing, applying an Interquartile Range (IQR) for outlier detection and winsorization to mitigate extreme values without data loss. Next, was performed using an IGA that incorporates adaptive mutation, elitism, and hybrid crossover to generate balanced and ill-comprehensive assessment patterns. Finally, student performance was predicted using a transformer-based model that processed chronological learning sequences by combining numerical academic features with textual assignment data. The multihead attention mechanism of the model captures long-term dependencies, enabling precise performance forecasts. Integration of Advanced Optimization with Deep Contextual modeling. The proposed IGA-Trans model obtained high results when compared with the existing Fuzzy Comprehensive (FC) context with an accuracy of 99.9%, precision of 99.6%, recall of 98.3%, and f1-score of 99.2%.