In this section, some existing research works on HMM prediction and on check in data mining will be reviewed as below.
As for HMM strategies, (Blasiak & Rangwala, 2011) applied HMM to the classification, which completed the sequence classification by combining Baum-Welch, Gibbs sampling and change function together. (Antwarg, & Rokach 2012) used multiple HMM to distinguish user from different devices, ages and gender. Thus, it could provide a more accurate prediction of user's next behavior. With the growing popularity of mobile terminals, HMM also began to be applied to the field of path analysis. (Hamada & Kubo, 2013) used a modified BP-AR-HMM algorithm to predict user's driving behavior under multi-time series. (Wesley 2012) converted earth into a plurality of triangular surface, and completed the prediction of user's next location through putting labeled triangle into HMM learning model.