There are different algorithms of static data grouping proposed by scientific community. Nevertheless, because of the massive generation of time series data, the algorithms must be adapted to be applied to this kind of data. One of the algorithms that have a better performance than K-means and fuzzy K-means is the one of gravitational clustering, based on randomized interactions of data points RAIN, reason why it is proposed the extension of the use of that algorithm for univariate time series. This work uses the technique of time series grouping, based on models, making a pre-processing of the time series; then, it is made a modeling of the technique using regressive modeling methods and, finally, an algorithm of gravitational clustering is applied to the model created. An analysis of a series of models is made in order to recommend the using of one of them to the case of time series.