Research on Learning Outcomes Prediction in Mechanical Engineering Courses based on Machine Learning

The prediction of learning outcomes is becoming an increasingly crucial aspect of educational quality assurance and the provision of targeted student support. Conventional assessment techniques frequently prove inadequate for identifying students who are at risk of failing to learn in a timely manner. This makes it challenging to intervene in a timely manner. This work proposes an innovative predictive model based on the combination of machine learning and deep learning, which aims to accurately predict the learning outcomes in mechanical engineering courses by analyzing students' multi-dimensional data. Initially, classical principal component analysis (PCA) is employed for data preprocessing purposes. This enables the extraction of key features, reduction of data dimensions, and identification of the core factors influencing academic performance. Subsequently, the preprocessed features are fed into a deep learning model based on a long short-term memory network (LSTM) architecture. LSTM is designed to process time series data, thereby enabling the capture of dynamic changes in students' learning behaviors, including performance trends over time and the time-dependence of learning activities. Furthermore, the model incorporates an attention mechanism to dynamically assign weights to different features, thereby facilitating more accurate identification of the factors with the greatest impact on learning outcomes. The experimental results demonstrate that the hybrid model exhibits superior performance in terms of accuracy and generalization ability when compared to a single machine learning algorithm and traditional statistical methods.