The invention discloses a method for predicting service performance of a video website. The method comprises: step 101) through simulating a user request, measuring video website performance parameters at regular time, calculating a correlation coefficient of the performance parameters; if the absolute value of the correlation coefficient exceeds a preset threshold value, storing values of all performance parameters to a performance prediction data set of a corresponding website, if not, just storing download velocity to a performance prediction data set of the corresponding website, wherein the performance parameters include connection time and download velocity; step 102) performing parameter nondimensionalization and normalization preprocessing on the prediction data set, based on the data and a time sequence model or a machine learning model obtained by the preprocessing, predicting service performance of the video website at next moment, performing reverse normalization on the normalization prediction result output by the time sequence model or the machine learning model, to obtain a real service performance prediction value which can be obtained by accessing the video website by the user at next moment.