More and more, people are looking to vocational institutions to help raise the bar on education quality. The level of vocational schools’ quality assurance is strongly correlated to the quality of their courses, as courses are the fundamental building blocks of vocational college instruction. As a consequence of their rapid growth, vocational institutions now provide lower-quality courses with inadequate teaching resources. Reforms to higher education have always centred on one central goal: raising standards. The most pressing social issue is the development of criteria for objectively assessing the quality of scientific curricula. Consequently, there is some practical need for enhancing the quality of the curriculum to conduct a thorough study of the elements influencing its quality and to build a system and process for evaluating its quality. This study suggests a network for evaluating the quality of vocational education courses by integrating deep learning models with course evaluations. Afterwards, corresponding improvement measures will be implemented based on the evaluation results. In order to study the link between the quality of different vocational education courses explicitly, this paper suggests using attention network. It employs parallel networks with separate convolution kernels to combine scale features at the same geographical location in order to better extract information on the relationship between the quality components of vocational education courses. Compared to other machine learning algorithms, Multi-scale Attention Convolutional Neural Network (MSACNN) achieves better results in terms of accuracy and recall.