College curricula and the associated understanding of skills are experiencing significant change in response to the needs of social and economic growth. Big data and data mining were already extensively applied in a variety of fields to extract important knowledge and give decision assistance, both of which have contributed to the demise of traditional teaching methods and the introduction of significant new educational innovations. College students who study subjects other than English play a significant role in the improvement of various languages and developing students' abilities to use those languages in practical settings. As a consequence, the level of a college's English teaching (ET) is a strong indicator of the overall excellence of its academic programs. For this reason, measuring the effectiveness of ET in colleges has emerged as a pressing issue. This study presents a novel artificial bee-tuned back-propagation neural network (AB-BPNN) for assessing the efficacy of ET innovation in colleges. The techniques of feature extraction and feature selection employ Kernel Principle Component Analysis (K-PCA) and information gain (IG), respectively. In terms of accuracy, f1-score, recall, precision, run time, and prediction error, the proposed AB-BPNN approach's performance is assessed. Comparative assessments of the proposed and existing approaches are also presented in these metrics.