Features Extraction based on TF-IDF and Text Classification of the LDA Model

The ever-growing textual content has imposed the need for faster, more objective methods of text classification, while traditional manual processing remains limited. The overreliance on subjective judgment often entails inconsistencies like "different classifications for similar texts," thus daunting the need for the adoption of big data technology to aid institutional efficiency and fairness in text classification. It tries to enhance the accuracy of text classification with the aid of big data technology and popular deep learning methods. The research starts with preprocessing a large dataset by using jieba segmentation and word frequency statistics for managing the texts that have been collected, and TF-IDF for constructing a corpus at one's own discretion. The establishment of an LDA model offers part of the theoretical foundations for performing text classification. It can be argued based on the text argumentation that the Word2vec model is used to train text vectorization, which is then embedded into the CNN for text classification and prediction tasks. Results have shown that the model can guarantee a classification accuracy of 0.9, hence finishing the text classification task with high precision.