The Recommendation And Visual Analysis Of Shared Accommodation Based On Sentiment Calculating

In the context of the sharing economy,short-term rental of homestays have developed rapidly.Online reviews of homestays contain a wealth of emotional information,reflect-ing the attitudes and emotions of consumers.Through the comments,we can discover the advantages and problems of homestays,which can provide consumers with useful in-sights and promote the promotion of service providers.Therefore,sentiment analysis and mining of homestay reviews are critical.At the same time,the recommendation of home-stay listings is also an important research direction in the industry.Good recommendation results can save users' time and increase user loyalty and promote the healthy develop-ment of the homestay industry.However,the recommendation of a homestay is different from the recommendation of a traditional hotel,and there are many factors that need to be considered.Therefore,this article focuses on the above-mentioned research,focusing on the three aspects of homestay online reviews,homestay listing recommendation and how to perform visual analysis,and proposes the Bert-BiL STM-EIC sentiment analysis network,the listing recommendation algorithm based on the topic model and emotional characteristics and design schemes of visual analysis system for shared homestays.Firstly,in the sentiment analysis problem of homestay reviews,this article combines the advantages of sentiment dictionaries and deep learning and proposes a network model Bert-BiL STM-EIC based on Bert and Bi LSTM network models for the large and complex online texts.The network model contains two input channels,a semantic channel and an emotional channel,so that the network can better integrate emotional features in semantic feature extraction,thereby improving the accuracy of the model.At the same time,we built a data set for sentiment analysis of homestay reviews through web crawler technol-ogy,and annotated the data set based on the review scores.Our model finally obtained an accuracy rate of 94.42%,and proved the effectiveness of our method through multiple sets of comparative experiments.Secondly,in the recommendation problem of homestay listings,this article focuses on the many details that affect the selection of listings.First,we start with the mining of comment topics.Based on the BTM short text topic model,we designed an experiment to obtain the best model and visualized it.Eight aspects that users care about are obtained.Then we conducted a correlation analysis of factors such as ”landlord trust” and ”risk perception”,and included these characteristic factors when analyzing user preferences.Then we use the well-trained emotional model combined with the topic model to propose a house emotional feature generation algorithm to obtain the hidden emotional feature of the house.Finally,based on the LFM algorithm,we proposed a housing recommenda-tion algorithm that considers emotional characteristics.Experiments have proved that our algorithm has better recommendation effects than traditional recommendation algorithms.Finally,to meet the analysis needs of users,we design a shared homestay recommen-dation visual analysis system named Opinion Manager.We visually integrate multiple sources of data such as reviews,location,and prices of homestays,and design and imple-ment various visual components such as ranking comparison view,theme rectangular tree map,emotional river view,etc.,to facilitate users to better visualize analysis.In addition,this paper also used the system to conduct user evaluation and multiple case studies,and relevant feedback confirmed the effectiveness and practicality of Opinion Manager.