Assessment of New Financial English Classroom Learning Based on Multi-Scale Attention – Gated Recurrent Unit

Nowadays, the increasing demand of English teaching in colleges as well as universities leads to improve the teaching process by combining the evaluation systems. To construct the evaluation system of English teaching and analysis the needs of English while mixed of subjection theoretical model so potential lack to predicted the teaching. This research, proposed Multi-Scale Attention – Gated Recurrent Unit (MSA-GRU) is efficiently handle the sequence data and complex patter in the data. The pre-processing using Min-max normalization involve to scale the data from the collected data, then categories the features. The feature extraction using Bidirectional Encoder Representations from Transformers (BERT) it has bidirectional nature improve the understanding of student responses and learning patterns. The MSA-GRU enhance the model ability to focus on various level of information while GRU efficiently process the learning data. It processing the refined feature vector while retaining essential sequential data and dense layer with softmax activation provides the final prediction. The proposed method achieves better accuracy 92.23%, precision 91.45%, recall 91.23% and f1 score 91.53%, respectively.