Research on Chinese Spam SMS Classification Based on ECA-TextCNN Model

Abstract

In response to the issues of traditional Chinese spam SMS classification models, which fail to comprehensively extract text features and obtain representative text features, we propose a text classification model based on the ECA-TextCNN model, combining an efficient channel attention mechanism and TextCNN classification model. Firstly, the model utilizes a skip-gram pre-training model for text representation. Secondly, it employs convolutional kernels of different sizes for feature extraction. Furthermore, ECA-Net is used to assign corresponding weights to each channel feature. Finally, the feature vectors are inputted into the softmax layer to obtain the classification results. The experiments are conducted on the publicly available Message80W Chinese SMS dataset. The results show that compared to the baseline model, the text classification model based on ECA-TextCNN exhibits varying degrees of improvement in classification accuracy, precision, recall, and F1 value.