| Title | : | Recommending Places and Regions of Interest |
| Speaker | : | Shiladitya Pande (IITM) |
| Details | : | Wed, 1 Mar, 2017 3:00 PM @ BSB 361 |
| Abstract: | : | We have several web services today that provide information about Points of Interest(PoI) like restaurants, parks etc. In our work, we explore how this information can be leveraged to provide recommendations to users in the form of similar PoIs and the best Regions of Interest(RoIs), i.e. a neighborhood of PoIs. To this end, we develop two frameworks- SkyGraph recommends the best neighborhoods based on a user-given set of query keywords in terms size of the neighborhood and coverage of the query keywords. Different users have different preferences with regard to the neighborhoods they would want to visit. For example, if a user is exploring a city on foot, he may be willing to visit a smaller neighborhood while sacrificing the coverage of some his desired PoIs. However, if a user plans to visit the PoIs on a car, he would probably prefer a neighbourhood covering all his desired PoIs. SkyGraph balances this trade-off between the size and coverage to recommends the best neighbourhoods to the user. Our second framework, GeoScop, clusters PoIs in the form of check-ins based on their spatial proximity as well as social similarity. Specifically, two places are considered similar if they are spatially close and visited by people of similar communities. While community detection is typically done on social networks, in our problem, we lack any network data. Rather, two people belong to the same community if they visit similar geo-social clusters. We address this chicken-and-egg problem in our work and mine semantically meaningful clusters that cannot be found by using any of the existing clustering techniques. |
