Exploration of Text Classification Algorithms Based on Word Vector Techniques

This article employs Python web scraping techniques to extract text data from videos on Bilibili. It then explores the impact of different text vectorization methods and classification algorithms on the classification results. Initially, various methods such as LDA, Word2vec, Bert, and a fusion of Word2vec and Bert are used to vectorize the text. Subsequently, text classification is carried out using methods like KNN, SVM, and BP, followed by integrated processing of the classification results. For text vectorization, the fused model is used to vectorize the text to get better classification results. In the integrated processing, combining multiple "weak" classifiers to form a "strong" classifier achieves a globally optimal F1 score of 0.9686. In real-world classification problems, it is important to employ the concepts of fusion and integration, leveraging the latest algorithms or tools to enhance classification effectiveness and better handle text categorization tasks.