The Collaborative Filtering Recommendation Algorithm Based On Personalized Features And Its Application In The Field Of Tourism

With the improvement of life quality,people want to "go somewhere new" increasingly,the rapid development of tourism means that people’s demand for services began to largen,travel sites are part of tourism services.The main function of the travel sites is to provide a platform to customers to preview the beautiful scener ies without going out and to book their own journeys in advance before traveling.Science and technology is progressing,the quality of services on some major websites is also improving,travel sites do not just stick to providing a simple function of browsing and booking,in order to improve customer volume and transaction amount,the larger websites such as Ctrip,ELong begin to introduce magical function like recommending,the reason why it is magical is that the websites seem to be able to see through the customers’ mind and choose the sceneries that the customers may be interested in from many items,the algorithm used in the recommended function is the focus of this paper--the collaborative filtering recommendation algorithm.The algorithm is based on the scores that the users give the commodities,it trains the model and recommends the commodity items to the customers that they may be interested in.At present,the collaborative filtering recommendation algorithm is the most widely used recommendation algorithm,it is reflected on many commercial websites,the business area s it involves are not single,such as Taobao,Jingdong,Ctrip,Dangdang and so on.Although the collaborative filtering recommendation algorithm is widely used,the traditional collaborative filtering recommendation algorithm has some shortcomings,the accuracy of the recommended result is directly affected by the authenticity of the score data because it recommends just by the scores that the users give the items,in addition,the score data are often sparse,most lack of the data will lead to a large deviation in the forecast score,thereby reducing the recommended accuracy.Aiming at the shortcomings of the traditional collaborative filtering recommendation algorithm,this paper put forward adding the users’ and projects’ personalized features data to the calculation,so that the recommendation process does not rely on the scores that the users give the items entirely,thereby improving the defect caused by the authen-ticity and sparseness of data.There are two parts main work involved in this paper:(1)Based on the traditional user-based collaborative filtering algorithm,adding the users’ and projects’ characteristics,the original dynamic recommendation based on the score data is combined with the feature-based static recommendation to improve the algorithm.Then contrast and verify the classical algorithm and the improved algorithm by the movie scores data provided by the Movie Lens site to prove the correctness of the improved algorithm.Finally,the improved algorithm is applied to the data of the tourism website,recommending the scenic items which the users may be interested in.(2)Achieving the basic and expanded functions of the tourism website,including two modules,front-end site and background management.And according to the website ’s own characteristics,applying recommended results of the improved recommendation to the recommended function of the front-end site,to show the recommended effect.The final experimental results and the recommended results represent that the improved algorithm is feasible.