A Hybrid Deep Learning Model for Comprehensive Higher Education Quality Evaluation

Higher education quality is the driving factor for excellence in academe, reputation of an institution, and progress in society. Conventional appraisal systems must heavily depend on quantitative measures in the form of student performance metrics, faculty credentials, and available resources, while the qualitative dimensions such as sentiment in feedback and learner involvement are left out. Current machine learning methods, such as simple neural networks and statistical models, have achieved moderate levels of success in processing structured data but find it difficult to properly incorporate unstructured textual feedback, resulting in incomplete and even biased appraisals. In response to these shortcomings, this research introduces a hybrid proposed framework that processes both structured institutional indicators and unstructured feedback comments simultaneously. The DNN branch identifies intricate nonlinear interactions in quantitative features, and the Attention-BiLSTM branch derives semantic meaning from qualitative ratings, focusing on contextually relevant words and phrases. Their combination yields an overall representation of educational quality, allowing for more accurate predictions. Experimental findings on a benchmark higher education dataset confirm that the proposed model has an accuracy of 97.4 %, performing significantly better than traditional models. The proposed method provides a robust, scalable, and interpretable framework for full-range higher education quality evaluation