Research On Shared Accommodations Recommendation In Sparse Datasets

With the quick development of shared accommodations platform,the number of users and accommodations on the platform are rapidly growing,which has caused serious information overload problems.Personalized recommendation algorithm can help users quickly choose the room that suits them,which is an effective way to solve the problem of information overload.However,in shared accommodations platform,the serious sparsity of the users' behavior matrix has caused certain obstacles to accommodations recommendation.In view of the sparsity of data,combined with the characteristics of shared accommodations,this paper proposes a meta-path-based recommendation method that considers the user trust.Firstly,based on the research on the influencing factors of users' willingness to purchase shared accommodations,this paper fully considers the user's preference for the relevant attributes of the accommodation,takes into account the trust of users to host,and constructs a shared accommodations model from house attributes and host attributes.Secondly,in order to alleviate the data sparsity problem during the accommodation recommendation process,this paper proposes a meta-path-based recommendation method that considers user preferences,the method first uses project attribute information and user transaction list to calculate the preference value of the user's selected accommodations,and uses this as the user's implicit score to construct the users-accommodations interaction matrix.Then,this paper uses the objects and relationships in the platform to constructs a heterogeneous weighted network,introduces the concept of meta path to describe the relationship between users and recommendations from different perspectives,and measures the correlation between users and accommodations nodes under each meta path,on this basis constructs user feature vectors and use the supervised learning method to predict the probability of the user selecting other recommendations and generates a Top-N recommendation list.Finally,this paper uses real data from Airbnb to experiment.In the experiment,the results showed the introduction of host attributes can effectively improve the recommendation effect,verified the effectiveness of considers the user's trust preference for the host.At the same time,compared with other traditional recommendation methods,the meta-path-based recommendation method that considers user preferences showed a better recommendation effect in the case of sparse data.This study combines the shared accommodations selection problem with the recommendation algorithm,expands the application range of the recommendation algorithm.The impacts of the host attributes are considered into recommendation process,which enriches the research perspective of shared accommodations recommendation.The proposed recommendation method can describe the relationship between the user and the accommodation from multiple perspectives,expand the user's effective data,and can effectively alleviate the problem of data sparseness in the accommodation recommendation process.