Analysis Of Airbnb's Short-term Housing Portrait Based On Theme Model

Airbnb,as the world's largest online booking platform for C2 C shor-term rooms,has a large number of rooms and a huge consumer group.Its huge internal flow also represents a high commercial value,is becoming an important position for all kinds of personnel and investors of idle resource sharing to share resources and create income.How to use the housing data to provide the basis of management decision for the online accommodation platform has become the focus of the enterprise.In this paper,short-term housing data on Airbnb platform are used to construct housing portrait,which can be mined and analyzed.The housing is expressed as a multi-dimensional label system to help enterprises accurately locate the housing characteristics.It is the basis of improving housing supply products and housing supply recommendation,and it is of great reference significance to the accommodation booking platform and housing supply side.In this context,this paper studies the construction and analysis of Airbnb short housing portrait based on the theme model,with the purpose of accurately describing the housing characteristics,and carries out category segmentation and description analysis.The Airbnb room source data used includes the basic property data of the room source and the dynamic room source review data.Since the basic property data are filled by the room supplier,the processing of the data focuses on the room source review.The main work of this paper is as follows:(1)Preprocess the acquired review data.All housing was divided into economic,quality and high-end housing according to the per capita price.In the pre-processing of housing review text,meaningless words were removed,used Jieba word segmentation with good effect,and part of speech screening and word stop processing were carried out.(2)use the theme model to extract the features of the housing review.The classical LDA model and the time-sensitive DTM model were selected for the feature engineering to extract the housing theme,and quantitative evaluation indexes were selected for the evaluation.The evaluation results showed that the DTM model had a good effect on the property theme feature extraction.(3)The Word2 vec model considering semantic information is combined with the DTM model to construct the DTM2 vec model to better describe the characteristics of housing under the condition that the review dimension of housing remains unchanged.(4)Canopy and k-meansclustering algorithms were adopted to avoid the impact of noise data on the clustering effect,and appropriate evaluation indexes were used to compare the feature extraction effect of DTM2 vec model and DTM model.(5)the parameter setting of the optimal clustering result is adopted to generate the clustering of house source portrait.Economic housing is divided into 5 categories,quality housing into 4 categories,high-end housing into 5 categories,and summarized.The conclusions of this paper are as follows :(1)for the housing review data,the DTM model considering the time factor has a better theme extraction effect than the LDA model.(2)the DTM2 vec model incorporating contextual semantic information has a better descriptive effect on housing source characteristics,and the clustering effect is better than that of the DTM model.(3)the Canopy fusion k-means clustering algorithm was able to identify price noise housing sources and prevent them from interfering with the clustering effect,resulting in a good clustering effect,which made the clustering results descriptive,and could summarize and identify different types of housing portraits,providing certain reference for the management and development of housing sources.