Massive Spatio-temporal data often implies valuable topic information within a certain Spatio-temporal range.Recommending these topics to users can help them quickly understand current fashion trends or hot topics.However,most of the existing topic recommendation algorithms have not considered the user’s topic preference.The un-consideration may not achieve personalized topic recommendations for different users.With the expansion of the Spatio-temporal range,their efficiency of topic recommendation may also not be high enough.To remove the above issue for the practical application of personalized topic recommendation,this thesis proposes effective algorithms by studying the characteristics and existing problems of the recommendation under Spatio-temporal topic data.Firstly,this thesis proposes a topic mining algorithm for Spatio-temporal topic data.The algorithm considers the overall importance of the topic in the entire topic dataset and the attention of different users to different topic objects.It calculates the topic frequency via the above consideration.It then uses the topic frequency to determine the user’s topic preference.It considers the user’s preference for the Spatio-temporal topic mining.Secondly,according to the user’s preferred Spatio-temporal topic,this thesis proposes a personalized topic recommendation algorithm.The algorithm contains two sub-algorithms-the personalized topic relevance sub-algorithm and the personalized topic index sub-algorithm.The first sub-algorithm measures the relevance of the user’s preferred topic and the topic in the entire topic dataset.When the Spatio-temporal range becomes larger,the second sub-algorithm further reduces the aggregation size between the topic lists in different regions.The algorithm achieves the rapid recommendation of personalized topics for users by the above two sub-algorithms.Finally,this thesis applies the proposed personalized topic recommendation algorithm to the intelligent traffic trajectory recommendation system.The application process mainly includes the following three steps.The first step uses the Spatio-temporal topic mining algorithm to mine the user preference trajectory.The second step uses the personalized topic relevance sub-algorithm to calculate the correlation between the user preference trajectory and all the candidate trajectories.The calculation is based on the driver-vehicle-road relationship between the user preference trajectory and all candidate trajectories.The third step uses the personalized topic index sub-algorithm to quickly recommend personalized traffic trajectories.Meanwhile,this thesis experimentally compares the proposed algorithms and state-of-theart methods with the real dataset.The comparison verifies the effectiveness of the proposed algorithms.The advantages and disadvantages of the proposed algorithms are summarized.At last,the future research directions are also discussed. |