Taking the loosening and conditioning process, feeding moisture returning process, hot air moistening process and the whole silk making process as the research objects, the artificial neural network and multiple regression modeling method were used to investigate the influence of different modeling methods on the moisture prediction accuracy of each process. moreover, the predicted results were tested experimentally. The experimental results showed that the mean of the absolute value of the prediction error was 0.24% for the loose moisture returning process with multiple regression modeling method. The mean of the absolute value of the prediction error was 0.20% for feeding moisture returning process with artificial neural network method. The mean of the absolute value of the prediction error was 0.10% for hot air moistening process with artificial neural network method. The mean of the absolute value of the prediction error was 0.05% for all the silk making process with artificial neural network method. The model computing system was developed based on C# language, with the use of SQLSERVER database for data storage. The developed model operation system had strong ability of data analysis and production prediction, which could be used to predict the moisture content of each key process in cigarette silk making process.