The goal of this research is to study the periodic and stochastic models of data series of the weeklycumulative rainfall. The study was done by using daily rainfall series with long data 25 years(1977-2001) of Purajaya station. Based on the daily rainfall series, a weekly cumulative rainfallseries was calculated. The series of rainfall is assumed to be free of trend. Periodicities ofperiodic rainfall model of the weekly rainfall data series were presented by using 509 periodiccomponents. Stochastic components of the weekly rainfall data are assumed as residues betweenthe rainfall data with the periodic rainfall model. The stochastic components were calculatedusing the approach of autoregressive model. Stochastic Model presented in this research is usingthe thirth order autoregressive model. Validation between rainfall data and the periodic-stochastic rainfall model is done by calculating the correlation coefficient. For this study, thecorrelation coefficient between the data and the model of the weekly cumulative rainfall is 0,9999.From the results of this study can be inferred that the model of the weekly rainfall from Purajayastation gives very highly accurate approach.