Abstract: Taking the paper book loan data in Jiangnan University Library from 2002-2017 as example, this paper analyzed the change rules of book loan time series data, and established multiple seasonal ARIMA with Autoregressive conditional heteroscedasticity model in view of the complicated seasonal effect and long-term trend of the sequence caused by its randomness and human factors. According to Akaike Information Criterion (AIC) and the minimum relative error criterion, it selected the optimal ARIMA-ARCH model to fit the data and predict the prospect effectively. According to the forecast, the borrowing capacity would continue to decrease slowly in the future with yearly period, so as to provide valuable information to library managers.