Article Preview
TopIntroduction
With the rapid development of location-based social networks (LBSNs), more and more users share their location information on online social platforms and comment on specific locations such as Foursquare (Gao H J & Liu H, 2014) and Gowalla (McKenzie G, et al., 2013). By analyzing these shared data, users' personal preferences, point of interest (POI) and routes can be mined to provide route recommendation and search services for people's travel (Bao J, et al., 2014).
Billions of users share their thoughts by posting photos and texts on social networking services. They are interested in various topics and usually have different emotional tendencies and posting activities. A model was proposed to characterize the posting activities of social network users and to predict the user's interest. A typical model of user postings is built to represent the user's posting behavior as the probability distribution of the posting mode. The user's behavioral characteristics is extracted from the results of the posting mode, the language features are extracted from the homepage, an interest prediction model is built (Hu C & Cui X H. 2020). In order to effectively capture the spatial characteristics of rich check-in data and social multi-dimensional context information and to deeply dig into the nonlinear interaction between users and POI, a POI recommendation algorithm with enhanced spectral embedding is proposed, and a spectral clustering algorithm is designed with enhanced preference and spectrum embedding enhanced neural network (Liu Z, et al., 2020). In view of the structural approximation clustering algorithm cannot solve the problem of asymmetric network clustering, according to the characteristics of social networks, a directed social network clustering algorithm is proposed based on structural approximation. By abstracting the social network into a graph structure, the network clustering is regarded as the problem of subgraph division in graph theory, the accurate clustering of social networks is achieved (Wang Y Y, et al., 2020). How to integrate the node's own attributes and network structure information, social network node classification is achieved, a social network node classification algorithm is proposed based on graph coding network (Hao Z F, et al., 2020).
Most existing route recommendations are aimed at individual user needs. Route recommendations are based on location and route popularity (McKenzie G, et al., 2013; Hsieh H P & Li C T, 2012; Yoon H, et al. 2012), these recommendations are combined with user’s preferred route recommendation and search (Cheng A J, et al., 2011; Bao J & Zheng Y, et al., 2012), and the location, time and weather route search are considered (Hsieh H P, et al., 2012; Majid A, et al., 2012). For the recommendation of group users, existing research usually aggregates the preferences of group users, and then the method of individual route recommendation is adopted to solve (Garcia I, et al., 2012; Masthoff J, 2011; Song X Y, et al., 2015).
In real life, when the user preferences of group travel partners are quite different, it is difficult to find a peer route that meets the preferences of all users. When accessing specific POIs, individual users can choose the POIs around this point and maximize their profits for access according to their own preferences. Revenue is the user's satisfaction with the POI, it is measured by the degree to which the location category in the route meets the user's preference.
In response to this demand, a local decentralized travel route search is proposed for group users. This route search is combined with the query location of group users, individual user preferences, and the range of areas around the POI, users can decentralize to search for a group of users with a locally distributed POI and the group's most profitable access route. In this paper, the local dispersion degree is used to limit the user's access to the surrounding areas of the POI. The specific definition will be given in Chapter 3. Each user's optimal travel route may contain different POIs, but they are all within the POI which is included in the optimal route, the purpose of common transfer and decentralized access are achieved.