Abstract：Using Machine Learning to Predict Chat Difficulty This study aims to evaluate the effectiveness and potential utility of using machine learning and natural language processing techniques to develop models that can reliably predict the relative difficulty of incoming chat reference questionsUsing a relatively large sample size of chat transcriptsN15690an empirical experimental design was used to test and evaluate 640 unique modelsResults showed the predictive power of observed modeling processes to be highly statistically significantThese findings have implications for how library service managers may seek to develop and refine reference services using advanced analytical methodsConsulting ServiceThis research mainly focuses on the difficulty of using AI prediction chat reference issueswhich is part of the consulting serviceThrough the difficulty of predicting problemslibraries can more effectively allocate resources and provide better reference servicesNLPNatural Language TreatmentThe purpose of the research is to evaluate the relative difficulty of using machine learning and natural language processing technology to predict the reference problem of chat referenceMachine learningThis study uses a large number of chat records to test and evaluate the model with empirical experimental design to determine the predictive ability of the model