Now, we study another interesting movement pattern called Mining Semantics-Rich Movement Patterns. We'll first see what is Frequent Movement Pattern? Frequent Movement Pattern is a movement sequence, frequently appears in the input trajectory database, but we care more on the semantics here. For example, from home to Starbucks to office. So maybe people, the home maybe different or office maybe different location, but this pattern maybe interesting, even we can consider the home and office could be just the same semantic units in different locations. So comparing to Frequent Movement Pattern versus Frequent Sequential Pattern, we can say, they both are at finding frequent subsequences from input sequence database. However, for mining frequent movement patterns, we would like to think of similar places may need to be grouped together collectively to form frequent subsequences from the semantic point of view. Now, we look at Mining Semantics-Rich Movement Patterns. The semantics-rich patterns means in addition to know how people move from one region to another, we want to also understand the functions of the regions. That means we want them see people moving from office to restaurant or from home to gym. So, we promote a two-step top-down mining approach. The first step is try to find a set of a coarse patterns that reflects people semantic level transitions. That means as long as a reaching carry the similar semantic functions, we may one to consider they are the same function or same entity. Then in the step two, we split each coarse pattern into several fine-grained ones by grouping similar movement snippets together. This time, we'll consider spatial, temporal, besides the semantics. We're going to consider spatial and temporal features more seriously. A recent study published in VLDB 2014 called Splitter, Mining Fine-Grained Sequential Patterns in Semantic Trajectories is an interesting study in this direction. [MUSIC]