Research On Kunming Tourism Route Recommendation Algorithm Based On Deep Learning

With the continuous development of the economy,people's living standards have also been continuously improved,and residents' travel styles are also changing.The traditional travel mode of travel agencies planning travel routes is no longer the first choice for people to travel.On the one hand,the rapid development of the Internet has brought convenience to people,and on the other hand,it has also brought a lot of information to the public.It takes a lot of time and energy for people to select valuable and suitable travel information from the huge amount of information.Planning a detailed and personalized travel strategy has become an indispensable job in free travel.Therefore,the planning of smart tourism routes also needs to follow the development of the tourism industry,and plan a detailed and accurate tourism route for tourists,which can not only improve the tourist experience of tourists,but also increase the user stickiness of tourists to the recommendation system.Therefore,no matter from the enterprise level or from the user level,this work is valuable.This article takes short-distance travel as the starting point and restricts the research scope to one tourist city.It is assumed that tourists can visit a limited number of scenic spots in a day.Through the mining of tourism big data,the popular deep learning model is established,and Internet map service data is called to extract relevant information.At the same time,using tourist's big data information as experimental samples,a tourism route planning algorithm based on neural network and tourist income function is designed.Plan for users a travel route that suits the user's own situation,budget cost,time and interest preferences,and provide tourists with a variety of sub-optimal routes for decision-making.The research work of this article is as follows:"Mafengwo" is a large-scale domestic tourism portal website.On the platform,visitors can write travel notes,exchange experiences,share experiences,obtain customized travel services,etc.Therefore,the experimental data source of the research is selected as the ‘horse honeycomb'.Use the crawler tool to first obtain the travel notes,travel logs,user characteristics and other information of senior users above level 30 on the ''Mafengwo''.After data cleaning,combination,and storage in a unified format,this information includes the demographic characteristics of the tourist travel,Travel history,income,interest preferences,etc.Tourist attractions information is obtained from professional tourism websites,and the 31 most popular tourist attractions in the urban area of Kunming,a popular tourist city,are selected as the research objects.It covers 7 types of tourist attractions,including natural scenery,famous mountains and rivers,historical sites,amusement parks,and shopping centers.Extract tourist attractions information from professional Internet travel websites and local tourist information networks,including the latitude and longitude of the tourist attractions,heat index,traffic information,etc.The user data and tourism information data are modeled,and the acquired user data and tourist attractions data are first combined to form a user interest sample.This is the basis and core part of establishing an interest mining model based on convolutional neural networks.Store the combined data in the text in a uniform format.Next,the convolutional neural network model is designed.This paper designs a convolutional neural network model with 15 convolutional layers,2 pooling layers,and2 fully connected layers.20000 data samples are used as the ratio of 9:1.Training samples and test samples.The model is obtained by training the training samples for 20 epoches and stored.In order to call the model when predicting.Design the optimal tourist attraction mining algorithm.The neural network prediction module outputs the descending matrix of tourist attractions classification that tourists are most interested in.According to the actual situation,the number of tourist attractions visited by tourists in a day is limited.Therefore,according to the descending order of interest classification matrix,the tourist attractions classification and the number of tourist attractions distribution matrix are designed.Call it the classification distribution matrix of tourist attractions.The recommendation system will focus on recommending the top planning schemes in the matrix rows,and visitors can also choose by themselves.When a tourist chooses a certain line,the tourist experience income value of each tourist attraction section is calculated according to the designed tourist income function.Under the arrangement of tourists' interest and time,the larger the income value,the better the travel experience.By calculating the target maximum value based on the income function,the seed tourist attractions that tourists are most interested in are discovered.The set of optimal tourist attractions is discrete in geographical distribution.According to the principle of permutation and combination,these tourist attractions can be combined into multiple routes,that is,which attraction is visited first and which attraction is visited.The travel experience for tourists is different,Only planning from the perspective of distance is too mechanical.Considering that the factors that affect the travel experience of tourists include not only distance,but also transportation waiting time,budget cost and other factors,therefore,the positive and negative impact factors of the income function are introduced for iterative calculations,and the tourist routes formed by each attraction are calculated separately The iterative function value of tourist income,and finally the route with the smallest function value is selected as the optimal tourist route to recommend tourists,and at the same time,the second best route is provided to tourists for decision-making.At the end of the article,a more detailed analysis of the experimental part is carried out.Explains the application value of the scheme proposed in this article.At the same time,it analyzes and summarizes the defects and parts to be improved in the tourism route planning algorithm based on neural network,and makes a prospect for future work.