Abstract:

As a typical space-time big data, trajectory data has high research and application value. However, the existing trajectory data mining mainly focuses on the spatial features of the trajectory, and less on the depth analysis of the trajectory data semantics. This paper is oriented to the demand of smart travel service, aiming at the semantic annotation of the track stop point, focusing on the automatic labeling problem of the track stay point semantics of the travel track. Firstly, aiming at the characteristics of short text of POI, this paper proposes a method of semantic extension of short text based on "synonym word forest" to extend the feature of POI short text. At the same time, it takes into account the feature set of POI short text and the distribution of category words. In this paper, the POI automatic classification method for improving TF-IDF is proposed. Secondly, based on the POI classification, the Native Bayes method is used to semantically mark the track stay points. The results show that the automatic classification of POI based on the improved TF-IDF method can achieve about 83% accuracy, which can better realize the classification of POI. On the basis of automatic classification of POI, the trajectory semantic annotation based on Native Bayes can reach 74%. The accuracy of the target is automatically achieved by the automatic semantic annotation of the travel track stay point.

Key words: semantic annotation, travel track, stay point, POI classification, feature extension