Research And Development Of Personalized Itinerary Recommendation Based On Uncertain Datasets

Therearetwomainchallengeswhileplanningitineraries. First,it’sdifculttoplanitineraries in perfect accordance with people’s preferences from massive information.Second, planning an itinerary involves selecting locations as well as determining thetime and order to visit each location, which is quite complicated and time-consuming.Naturally, efective and reasonable itinerary recommendation is the answer to all theseproblems. However, existing studies on itinerary recommendation have focused onlimited scenarios. Besides, the low-level automatic and personalized recommendationonly leads to user dissatisfaction.In this paper, we have analyzed existing studies on itinerary planning and rec-ommendation. Considering the characteristics of uncertain location data, we proposea transfer-based model which solves the problem of data uncertainty with collectivewisdom. The model describes the correlation between locations from the perspectiveof reachability and sequence. It also reduces the search space. On the basis of themodel, we introduce a personalized itinerary recommendation method. Firstly, we uselatent factor model to identify and predict user preference from historical data. Sec-ondly, we generate weighted density distribution of both visit time and transfer timeand determine the proper time context with KL divergence in order to achieve con-text personalization. Finally, we propose an algorithm to generate itineraries based onboth user preferences and context information. We also design and implement a per-sonalized itinerary recommendation system. Server-side ofine computing identifes and predicts user preference and online computing is responsible for processing userrequests. Generated itineraries are displayed by system client. Last but not least, wevalidate the method on real-world dataset.Firstly,weintroducebackgroundoftheresearchandreviewliteraturesaboutitineraryplanning and recommendation which presents issues in existing studies. Secondly, wepropose the transfer based model after analyzing characteristics of location data. Withthis model, we introduce a personalized itinerary recommendation method and realizepersonalization from the perspective of transfer and related context. Finally, we de-scribe the implementation of personalized itinerary recommendation system. We alsoconduct experiments on real-world dataset to validate the method.