Time series data modeling can be done by modeling each object one by one. Monthly rainfall data is an example of time series data. The purpose of time series analysis is to find patterns of past data and then forecast the future characteristics of data. The data used in this study is the Banten Province rainfall data which contained 19 rainfall stations. So it will require 19 models to forecast the rainfall data. The pattern of time series data in Banten Province monthly rainfall data in several locations has similarities. So that the similarity of this pattern can be considered in the clusters. In time series clustering, the idea is to investigate the similarity of time series in a cluster. The accuracy of distance similarity size measurements is performed on the generation data generated from 3 models, namely AR (1), AR (2), and AR (3). The piccolo method has an average accuracy of 0.62. While the maharaj method has an average accuracy of 0.41. This means that the Ward hierarchical clustering method using the Piccolo distance approach has a greater accuracy value than the Maharaj distance approach. Furthermore, the Piccolo method can be used as an alternative to the excellent distance method for grouping time series data in case data. The Banten Province rainfall station has 3 optimal clusters. Modeling individual level and cluster level has accuracy values that are not much different.