| dc.description | We have a great discussion in the 20 minutes presentation, there are four questions and comments:
1. It seems the bar tour recommendation could be used in other application domains, e.g. sightseeing, museum visiting, shopping, and cultural and historical spots learning. In deed, the recommendation method proposed by this research is based on time-series behavior analysis, it could be used in variety application domains as long as the behavior data is sequential one.
2. How the proposed method deals with distractive behaviors? The proposed method currently only uses time gap and minimal range in distance to deal with noise data. We could consider Wavelet and other DSP (Digital Signal Processing) methodologies to filter the noise.
3. How large data set the proposed method needs? Of course, the method welcomes large data set very much, however, due to the method actually uses the length of the association rule’s LHS (Left-Hand-Side) to estimate the strength of the rule, the size of data set has less important.
4. How to decide the support threshold? Just like other association rule mechanism for Very Large DataBase (VLDB), the support threshold will be vary depends on application domains and data set sizes.
My further research will then focus on dealing with the application domain and distractive behavior filter issues. Due to I also work with Dr. Kinshuk in his iCORE project on personalization and adaptivity in informatics, I will try to immigrate the recommendation method to mobile cultural and history learning. | en |
