A college entrance exam score prediction model based on improved TF-IDF and LSTM and its application in an AI admission platform

To address the increased difficulty of college application decisions under the new college entrance examination reform, this paper developed a personalized AI-based mini-program for recommending college majors to candidates. By combining data scraping, TF-IDF (Term Frequency-Inverse Document Frequency), and LSTM (Long Short-Term Memory) neural networks, the mini-program analyzes candidates' exam scores, selected subjects, and attended high schools. Compared with traditional LSTM models, this paper innovatively incorporates scraped news data, processed through an improved TF-IDF method, as additional features for model training. The improvements include sentiment analysis and a time-decay factor, which enhance the accuracy of predictions. The mini-program provides candidates with more accurate and personalized major recommendations based on their individual information, thereby improving the accuracy and satisfaction of their application decisions.