A Novel Tourist Attraction Recommendation System Base on Improved Visual Bayesian Personalized Ranking

Statistics express that most tourists log into the main tourism websites to view user review or mark before choose their destinations . However , the survive tourist destination recommendation models neither look at the implicit user preferences nor mine the potential semantics of tourist attractions . To solve the trouble , this paper anticipate user scores of tourist attraction through stratified sample , and optimizes the predicted mark with Bayesian personalized ranking ( BPR ) and improve visual BPR ( VBPR ) . Then , the recommendation system was optimized by the improved VBPR , which decomposes the prediction score matrix and considers visual feature . Experimental results full evidence the excellence of the suggest tourist attraction recommendation system . The research finding provide a good reference for online travel agencies to recommend tourist attraction .

recommendation system , Bayesian personalized ranking ( BPR ) , stratified sampling , tourist attraction

With the development of Internet technology , the quantity of information on the Internet is exploding , build the trouble of information overload more and more outstanding . It be difficult for a user without a clear need to obtain the information of interest from a big measure of information . Meanwhile , band of valuable information are submerged in the sea of data , become invisible to potential users . Faced with the massive online data , the traditional lookup algorithm , able of filtering information for user , can not supply personalized service that see the interest and preferences of each user . To solve the trouble , recommendation system [ 1-2 ] has come out as the bridge between users and Internet information . On the one hand , the system enable users to find interesting information from massive datum ; on the early hand , it could remove valuable information to potential user . Relevant survey hold shown that more than 75 % of tourist log into the chief tourism websites to view user review or scores before choose their destinations and travel routes . However , many of them have difficulty in finding valuable capacity out of the huge sum of information on these websites . Therefore , the tourism websites need to promote improve their recommendation system , in order to satisfy the growing need for personalized tourism .

In this paper , the latent semantic space example [ 3 ] is combined with visual Bayesian personalized ranking ( VBPR ) [ 4 ] into a fresh VBPR recommendation example , which mitigates the trouble of datum sparsity and improve the interactive experience of user .

The research into recommendation system set out in the early 1990s . The early recommendation systems could just suggest the merchandise that interest user , using a recommendation algorithm . With the proliferation of such systems , the recommendation trouble have gradually develop into the grudge prediction of the recommended object .

Many scholars at place and abroad have studied recommendation system . For instance , Goldberg et al . [ 5 ] innovatively introduced collaborative filtering to Tapestry system . Pang et al . [ 6 ] proposed a score-based collaborative filtering recommendation example , which descend user preferences from their lots , and study user similarity through clustering to complete the recommendation . Combing user score with visual information , Huang et al . [ 7 ] solved the recommendation trouble with matrix factorization ( MF ) model . Based on convolutional neural network ( CNN ) , Tsai et al . [ 8 ] build a recommendation system couple user score with TV . Drawing on deep structured semantic model ( DSSMs ) , Yoon et al . [ 9 ] constructed a location-aware personalized word recommendation system . Zhao et al . [ 10 ] recommended commodity with a high truth , using a self-designed Bayesian personalized ranking ( BPR ) model . Pan et al . [ 11 ] employ the BPR model to hotel recommendation . He et al . [ 12 ] introduced visual information to commodity recommendation , and extended the BPR example into the VBPR model . Based on attention mechanism , Han et al . [ 13 ] advise a CNN model for the recommendation of Weibo information .

In late yr , the tourist attractions recommendation system hold become a inquiry hotspot . For lesson , Yuan et al . [ 14 ] design a case-based tourist attraction recommendation system , and developed a web-based intelligent recommendation framework for travel office , which incorporate reason with multi-criteria decision-making technology ; the recommendation character of their system was verified through experiment . Based on collaborative filtering , Kirn et al . [ 15 ] adjust up a decision support system for tourist attractions , which promise user preference for tourist attractions by Bayesian example , and demonstrate the prediction accuracy by the receiver operating characteristic ( ROC ) curve . Hsu et al . [ 16 ] design a tourist attraction recommendation system based on multi-criteria collaborative filtering ; the system relies on multi-objective collaborative filtering to process more information on user preferences , thereby assemble user demand more in effect .

Despite the above breakthroughs , the recommendation systems for tourist attraction even so face up multiple challenges : ( 1 ) the data on user mark exist extremely sparse , stimulate a prominent trouble of datum sparsity ; ( 2 ) the tourist attractions be recommend solely base on the historical data of user , without considering the latent information of user preferences ; ( 3 ) the possible semantic information of tourist attraction picture or user is not mined or analyzed from the perspective of multimodality .

The literature review suggest that the previous recommendation system make not considered the role of tourist attraction images across heterogeneous medium . To build up for the gap , this department mines the latent semantic space of users and tourist attractions with MF model , and optimizes the space in BPR or VBPR model to generate the prediction score matrix .

3.1 Tourist attraction recommendation system base on stratified sampling and BPR example

To design the tourist attraction recommendation system , the user preferences were captured through stratified sampling , and the latent semantics of users and tourist attraction were mined by the BPR example . To set out with , the datum on tourist preferences were pick up through a questionnaire survey , and subject to stratified taste . Under the preset collection rules , the datum on user grudge of tourist attraction make up acquired automatically from Ctrip.com , and preprocessed . Next , a user score matrixRexist generate for tourist attractions . Based on the BPR example , MF model , and matrixR, a prediction grudge matrix was lay down by predicting user score . Then , the recommendation $ R_ { A } $ from the BPR model was combined with that $ R_ { H } $ receive through stratified sampling into the sundry recommendation $ R_ { A } +R_ { H } $ .

In the stratified sampling model , the overall unit live separate proportionally intoUindependent layers $ \left ( H_ { 1 } , H_ { 2 } , \cdots , H_ { U } \right ) $ . Then , sampling was performed layer by layer . The result of all layer were lend up to obtain the overall distribution of samples . The specific workflow is as adopt :

Step 1 . Describe the differences in user preferences with random target variables , namely , travel time , interest family , and travel mode .

Step 2 . Based on the influencing factors , stratify the overall unit into $ U $ layer , each of which have $ H_ { i } $ individual : layer $ 1 ( i=1,2 , \ldots , U ) . $ Hence , the overall distribution of sample $ H $ can be calculated by :

Pace 3 . Decide the sampling number of each layer . LetNbe the full issue of sample . Then , the taste act of layeracecan be calculate by : $ X_ { i } =N \times H_ { i } / H $ .

To cut down intra-layer difference and amplify inter-layer dispute , the samples make up classified through stratified sampling , in the spark of the feature distribution of the overall unit . After that , a certain number of samples were extracted from each layer to describe the distribution of that layer , organize the sample population .

Because of the reasonable stratification of sample , the stratified sampling example could catch user preferences for tourist destinations in an accurate manner , laying a solid basis for produce a suitable recommendation list .

The stratified sample outcome were weighted through analytic hierarchy procedure ( AHP ) , a subjective weighting method , take to adapt the weight of each user attribute . Specifically , the relative importance between attributes on the same layer was compare to a new discriminant matrix , which was used to settle the weight of each dimension .

In this paper , six user attributes are select , including gender $ \left ( G_ { 1 } \right ) , $ area $ \left ( G_ { 2 } \right ) , $ years $ \left ( G_ { 3 } \right ) , $ educational background $ \left ( G_ { 4 } \right ) , $ job type $ \left ( G_ { 5 } \right ) , $ and monthly income $ \left ( G_ { 6 } \right ) . $ The relative importance between two dimension was measured against a seven-point scale : strongly significant =6 , moderately important =4 , slightly important =2 , equally important =1 , slightly unimportant=1/2 , moderately unimportant =1/4 , and strongly unimportant=1/6 . On this basis , the discriminant matrix G can be constructed as :

$ G=\left [ \begin { array } { ccccccc } { } & G_ { 1 } & G_ { 2 } & G_ { 3 } & G_ { 4 } & G_ { 5 } & G_ { 6 } \\ G_ { 1 } & 1 & \frac { 1 } { 2 } & \frac { 1 } { 4 } & \frac { 1 } { 4 } & \frac { 1 } { 2 } & \frac { 1 } { 6 } \\ G_ { 2 } & 2 & 1 & \frac { 1 } { 2 } & 2 & 4 & \frac { 1 } { 2 } \\ G_ { 3 } & 4 & 2 & 1 & 1 & \frac { 1 } { 4 } & \frac { 1 } { 2 } \\ G_ { 4 } & 2 & \frac { 1 } { 4 } & 1 & 1 & \frac { 1 } { 2 } & \frac { 1 } { 4 } \\ G_ { 5 } & 2 & \frac { 1 } { 4 } & 2 & 2 & 1 & \frac { 1 } { 2 } \\ G_ { 6 } & 4 & 4 & 4 & 2 & 2 & 1\end { array } \right ] $

Base on matrix G , the weight (i=1 , … ,6) of thei-th dimension was forecast to yield user preference :

$ W_ { one } =\prod_ { i=1 } ^ { 6 } G_ { one j } / \sum_ { i=1 } ^ { 6 } \prod_ { j=1 } ^ { 6 } G_ { 1 j } $ ( 2 )

The survey datum on user preferences for tourist attractions were analyzed through stratified sampling , from the position of various user dimension . Form 1 and 2 present the stratified taste histograms of the preferences of 1,000 users .

Three decision could be pull from Figure 1 and 2 : Most user favor to move in spring and fall , when the temperature is comfortable and the scenery cost pleasant to the eyes ; The user under 20 , most of whom are scholar , favor to move in summer , as they hold lot of free time during summer holiday ; Males prefer to travel in spring , while female favor to travel in fall .

1.png

Figure 1 .The travel time difference among users with different genders , area , and age

2.png

Figure 2 .The travel time dispute among user with unlike educational background , task types , and monthly incomes

The traditional MF model to promise user scores of tourist destinations can be defined as :

$ \hat { s } _ { u , i } =\beta+\delta_ { u } +\delta_ { ace } +\theta_ { u } ^ { T } $ ( 3 )

where , $ \beta $ is the global outset ; $ \delta_ { u } $ and $ \delta_ { 1 } $ are the start of user $ u $ and tourist attractions $ one $ respectively ; $ s_ { u } $ and $ s_ { 1 } $ are $ k- $ dimensional vectors of the latent semantic space of users and tourist attractions , respectively . The adaptability of user $ u $ and tourist attractions i $ can be depict by the inner product $ { s_ { u } } ^ { \top } s_ { i } $ The user preference of a tourist attraction is negatively correlate with the adaptability .

Then , the MF model was optimized in the BPR model , which live an optimization framework for pairwise sorting through stochastic gradient descent ( SGD ) . Pairwise sorting has a a lot better optimization effect than single sample .

Next , a education set $ D_ { S } $ of three $ ( u , i , j ) $ was established , where u is the user , i is a tourist attraction viewed positively by the user , and j is a tourist attraction consider negatively by the user :

$ D_ { s } =\left\ { ( u , one , j ) \mid u \in U \wedge i \in I_ { u } ^ { + } \wedge j \in I \\backslash I_ { u } ^ { + } \right\ } $ ( 4 )

Permit $ P_ { m } $ cost the parameter of the BPR model , and $ \hat { sec } _ { u , one , j } \left ( P_ { m } \right ) $ cost the relationship between triplets $ ( u , one , j ) $ . Then , the optimization procedure of the BPR example can be picture as :

$ \sum_ { ( u , i , j ) \in D_ { S } } \ln \sigma\left ( \hat { sec } _ { u , i , j } \right ) -\varepsilon_ { \tau } \left\|P_ { m } \right\|^ { 2 } $ ( 5 )

where , $ \sigma $ live a coherent occasion ( e.g . sigmoid purpose ) ; $ \varepsilon_ { \tau } $ equal a regularized hyperparameter . For MF-based prediction , $ \hat { s } _ { u , one , j } $ can be defined as :

$ \hat { sec } _ { u , 1 , j } =\hat { s } _ { u , one } -\hat { sec } _ { u , j } $ ( 6 )

After randomly sample $ ( u , i , j ) $ from $ D_ { S } $ , the relevant parameter can cost obtained by the BPR example , mate with the SGD :

$ P_ { m } \leftarrow P_ { m } +\eta \cdot\left ( \sigma\left ( -\hat { s } _ { u , i } \right ) \frac { \partial \hat { sec } _ { u , i , j } } { \partial P_ { m } } -\varepsilon_ { \tau } \left ( P_ { m } \right ) \right ) $ ( 7 )

To verify the tourist attraction recommendation system based on BPR model , the user preferences for tourist destinations exist receive through stratified sample , and weighted subjectively . The obtain data indicate that the user preferences vary greatly with user attributes .

By formula ( 2 ) , the effect of stratified sample cost weighted to set up a stratified sampling model . By the weight , the user dimension could be rate in go down order as monthly income =0.3702 , years =0.1948 , educational background =0.1372 , region =0.1360 , task type =0.1081 , and gender =0.0520 . Obviously , monthly income and age receive the great impact on travel , highlighting the importance of economic factors . On the contrary , gender and task type hold a relatively humble impact on traveling . This is consistent with our objective knowledge .

Then , the recommendation $ R_ { H } $ from stratified sample model and that $ R_ { A } $ from the BPR model were synthesize into a mixed recommendation $ R_ { \operatorname { Mix } 1 } \left ( R_ { A } +R_ { H } \right ) . $ Then , the precision of the mixed recommendation ( MR ) was compared with that of stratified sample ( SS ) example and BPR model , as well as that of traditional framework , include item-based collaborative filtering ( IBCF ) , user-based collaborative filtering ( UBCF ) , location-based collaborative filtering ( LBCF ) , horizontal collaborative filtering ( HC ) , and blurred c-means cluster ( FCM ) ( Table 1 ) .

As shown in Table 1 , the FCM had slightly superior precision than most traditional model , because the clustering-based prediction exist good guided by the prior knowledge in the eight type of tourist attractions .

Besides the FCM , the SS model achieved a high precision . This entail the user preferences receive through questionnaire study live quite exact , eliminating the motive for any automobile scholarship ( ML ) . Hence , it is real significant to search user preferences for tourist attractions .

The BPR model boast a lot better precision than the contrastive models , thanks to the mining of the latent semantic space of user and tourist attractions . Base on exist scores , these spaces accurately illustrate user preferences and popularity of tourist attraction .

The MR exist even more accurate than the answer of the BPR example . The excellence cost achieved through pairwise sorting of the BRP result .

Table 1 .The comparison of precision among different recommendation models

3.png

4.png

Figures 3 and 4 present the precision advantages of BPR example and MR , respectively . As shown in Figure 3 , the precision advantage of BPR model increased steadily with the growing issue of tourist attractions .

As indicate in Figure 4 , the precision advantage of MR generally increased , despite a certain volatility , with the raise act of tourist attraction . The MR precision is improved through the integration of user preferences into SS example . The precision of MR primarily comes from the BPR example . The user preferences obtain by the SS example make the BPR recommendation smoother , compensating for the bias of the BPR example .

The BPR model can outperform most traditional recommendation example . Even so , the image information of tourist attractions is not considered in that model , result in grave data sparsity . To overcome this trouble , this section improve the VBRP example by innovatively bring in visual features to tourist attraction recommendation .

Under the preset collection rules , the data on user scores of tourist attraction live acquired automatically from Ctrip.com , and preprocessed . Next , a user score matrix R was generated for tourist attraction . In the meantime , the image on the tourist attraction exist downloaded from Ctrip.com automatically . Then , the visual feature ( e.g . color and texture ) be extracted from these images , and used to improve the VBPR example . Base on the MF model , matrix R , and visual feature , the recommendation $ R_ { B } $ from the improved VBPR model was combined with the $ R_ { H } $ from the SS model into a sundry recommendation $ R_ { B } +R_ { H } . $ Owe to the improvement , the $ R_ { B } +R_ { H } $ reflects the merits of multimodal analysis , supplement the result of $ R_ { A } +R_ { H } $ .

In theory , the traditional MF example can notice the potential feature ( latent semantics ) of users or tourist attractions in the relevant dimensions . Nevertheless , the grudge datum acquired by the MF example cost also sparse to generate a robust recommendation . This trouble can be mitigate by add the auxiliary information of visual features . Hence , the MF of the VBPR model be improve as :

$ \hat { sec } _ { u , 1 } =\beta+\delta_ { u } +\delta_ { ace } +\theta_ { u } ^ { T } +I_ { u } ^ { T } I_ { ace } $ ( 8 )

where , $ I_ { u } $ and $ I_ { one } $ are newly bring in $ D $ -dimensional visual feature ; $ I_ { u } ^ { T } I_ { 1 } $ exist the visual interaction between user $ u $ and tourist attraction i .

Have $ \varepsilon^ { \top } I_ { one } $ be a user ’ sec overall evaluation of the visual feature of tourist attractioni. Introducing a visual offset term $ \mathcal { E } $ , the MF model can be finalized as :

$ \hat { sec } _ { u , i } =\beta+\delta_ { u } +\delta_ { i } +\theta_ { u } ^ { T } +I_ { u } ^ { T } I_ { i } +\varepsilon^ { \top } I_ { 1 } $ ( 9 )

The VBPR framework was adopted to optimize the last MF example . The SGD method cost adopted by the VBPR example to update the relevant parameters :

$ I_ { u } \leftarrow I_ { u } +\eta \cdot\left ( \sigma\left ( -\hat { s } _ { u , i , j } \right ) \left ( I_ { one } -I_ { j } \right ) -\varepsilon_ { \tau } I_ { u } \right ) $ ( 10 )

$ \varepsilon \leftarrow \varepsilon+\eta \cdot \sigma\left ( -\hat { s } _ { u , i , j } \right ) \left ( I_ { one } -I_ { j } \right ) -\varepsilon_ { \tau } $ ( 11 )

To sum up , the VBPR model was optimized from multiple visual angles , use VGG feature and related to visual features like pattern , color , and texture .

To verify its effectiveness , the improved VBPR example equal evaluated on the Wisdom Tourism dataset [ 17 ] by metrics like origin mean square error ( RMSE ) [ 18 ] , mean absolute mistake ( MAE ) [ 19 ] , precision , and recall . Different visual features be choose as the bases of recommendation , include scale-invariant feature transform ( SIFT ) , GIST descriptor , hue saturation value ( HSV ) , red green blue ( RGB ) , local binary practice ( LBP ) , and VGG . The RMSE and MAE of our model based on unlike visual features are presented in Figure 5 .

5.png

Form 5 .The comparison of RMSE and MAE of our model based on unlike visual features

As express in Figure 5 , the RMSE and MAE of improved VBPR example varied with visual feature . The high RMSE and MAE live observed on RGB feature , because RGB feature can not accurately reflect our objective knowledge of color ( chiefly from the angle of hue , saturation , and brightness ) . By contrast , the lowest RMSE and MAE exist attain on HSV features , suggest that the inclusion of HSV features can improve the recommendation result to a sure extent .

Furthermore , the improved VBPR model was compared with IBCF , UBCF , LBCF , HC , FCM , nonnegative matrix factorization ( NMF ) , k-th near neighbor ( KNN ) , and BPR in terms of RMSE and MAE ( Figure 6 ) .

6.png

Figure 6 .The comparisons of RMSE and MAE among different models

As indicate in Figure 6 , the contrastive framework like FCM , NMF , and KNN achieved good performance , but the improved VBPR with HSV features be more outstanding . The improved VBPR outperformed the BPR example , for the recommended objects live well depicted by the visual features of tourist attraction picture . Overall , the improved VBPR model had well RMSE than any early model .

In summary , the recommendation effect could by promoted by the improved VBPR model with HSV feature , which serve to anticipate user scores of recommended tourist attractions more in line with user preferences .

In addition , the recommendation $ R_ { H } $ from the SS example exist synthesized with the recommendation $ R_ { B } $ from the improved VBPR model into a sundry recommendation $ R_ { H } +R_ { B } $ based on visual features ( VMR ) . The precision of the VMR cost compare with that of many other model ( Table 2 ) .

Table 2 .The comparison of precisions among different recommendation framework

It can also exist find from Table 2 that the recommendation precision of the improved VBPR example increase steadily with the rising act of tourist attractions . Compared with NMF and KNN , the improved VBPR model only demonstrate a little volatility . Hence , the improved VBPR model be a stable method for recommend multiple tourist attractions to user .

Pattern 7 and 8 present the precision advantages of improved VBPR example and VMR , respectively . As present in Figure 7 , the precision advantage of VBPR example increase steadily with the growing act of tourist attraction .

As shown in Figure 8 , the precision advantage of VMR as well increase steadily with the raise act of tourist attractions . The VMR precision equal mainly resulted from the improved VBPR example through matrix decomposition to the inclusion of visual feature . Meanwhile , the user preferences obtained by the SS model play an auxiliary role .

7.png

Figure 7 .The precision advantage of improved VBPR example

8.png

This paper designs a tourist attraction recommendation system base on SS , BPR and improve VBPR framework . Firstly , user scores of tourist attractions exist predicted by MF model . Then , the predicted scores were optimized in BPR and improve VBPR frameworks , resulting in a high-quality recommendation of tourist attractions . Experimental result prove that the tourist attraction recommendation system base on SS and improve VBPR reach a high truth , fulfill user need , and greatly facilitate datum sparsity . The next inquiry will farther improve the recommendation performance of the propose system base on the multimodal semantic correlation between different visual features .

[ 1 ] Chen , J. , Li , K. , Rong , H. , Bilal , K. , Yang , N. , Li , K. ( 2018 ) . A disease diagnosis and treatment recommendation system based on bad datum mining and cloud computing . Information Science , 435 : 124-149. https : //doi.org/10.1016/j.ins.2018.01.001

[ 2 ] Yin , C. , Guo , Y. , Yang , J. , Ren , X . ( 2018 ) . A fresh recommendation system on the basis of consumer first step conclusion based on an associative classification approach . Industrial Management & Data Systems , 118 ( 1 ) : 188-203. https : //doi.org/10.1108/IMDS-02-2017-0057

[ 3 ] Dumais , S.T . ( 2004 ) . Latent semantic analysis . Annual Review of Information Science and Technology , 38 ( 1 ) : 188-230. https : //doi.org/10.1002/aris.1440380105

[ 4 ] Zhang , X. , Luo , H. , Chen , B. , Guo , G. ( 2020 ) . Multi-view visual Bayesian personalized rate for restaurant recommendation . Applied Intelligence , 50 ( 9 ) : 2901-2915. https : //doi.org/10.1007/s10489-020-01703-6

[ 5 ] Goldberg , D. , Nichols , D. , Oki , B. M. , & Terry , D. ( 1992 ) . Using collaborative filtering to weave an information tapestry . Communications of the ACM , 35 ( 12 ) : 61-70. https : //doi.org/10.1145/138859.138867

[ 6 ] Pang , Y. , Jin , Y. , Zhang , Y. , Zhu , T. ( 2017 ) . Collaborative filtering recommendation for MOOC application . Computer Applications in Engineering Education , 25 ( 1 ) : 120-128. https : //doi.org/10.1002/cae.21785

[ 7 ] Huang , L. , Tan , W. , Sun , Y . ( 2019 ) . Collaborative recommendation algorithm base on probabilistic matrix factorization in probabilistic latent semantic analysis . Multimedia Tools and Applications , 78 ( 7 ) : 8711-8722. https : //doi.org/10.1007/s11042-018-6232-x

[ 8 ] Tsai , Y.C. , Chen , J.H. , Wang , J.J. ( 2018 ) . Predict forex course via convolutional neural network . Journal of Intelligent Systems , 29 ( 1 ) : 941-958. https : //doi.org/10.1515/jisys-2018-0074

[ 9 ] Yoon , H.G. , Song , H.J. , Park , S.B. , Kim , K.Y . ( 2015 ) . A Personalized News Recommendation use UserLocation and News Contents . Appl . Math , 9 ( 2L ) : 439-449. https : //doi.org/10.12785/amis/092L19

[ 10 ] Zhao , Q. , Zhang , Y. , Ma , J. , Duan , Q . ( 2019 ) . Factored point similarity and Bayesian personalized ranking for recommendation with implicit feedback . Arabian Journal for Science and Engineering , 44 ( 4 ) : 2973-2983. https : //doi.org/10.1007/s13369-018-3358-0

[ 11 ] Pan , W. , Zhong , H. , Xu , C. , Ming , Z . ( 2015 ) . Adaptive Bayesian personalized ranking for heterogeneous implicit feedbacks . Knowledge-Based Systems , 73 : 173-180. https : //doi.org/10.1016/j.knosys.2014.09.013

[ 12 ] He , J. , Li , X. , Liao , L. ( 2018 ) . Next point-of-interest recommendation via a category-aware Listwise Bayesian Personalized Ranking . Journal of Computational Science , 28 : 206-216. https : //doi.org/10.1016/j.jocs.2017.09.014

[ 13 ] Han , J. , Zheng , L. , Xu , Y. , Zhang , B. , Zhuang , F. , Philip , S.Y. , Zuo , W. ( 2019 ) . Adaptive deep modeling of user and item apply side information for recommendation . IEEE Transactions on Neural Networks and Learning Systems , 31 ( 3 ) : 737-748. https : //doi.org/10.1109/TNNLS.2019.2909432

[ 14 ] Yuan , X. , Lee , J.H. , Kim , S.J. , Kim , Y.H . ( 2013 ) . Toward a user-oriented recommendation system for real estate websites . Information Systems , 38 ( 2 ) : 231-243. https : //doi.org/10.1016/j.is.2012.08.004

[ 15 ] Kim , H.N. , Ha , I. , Lee , K.S. , Jo , G.S. , El-Saddik , A . ( 2011 ) . Collaborative user modeling for enhanced content filtering in recommender systems . Decision Support Systems , 51 ( 4 ) : 772-781. https : //doi.org/10.1016/j.dss.2011.01.012

[ 16 ] Hsu , F.M. , Lin , Y.T. , Ho , T.K . ( 2012 ) . Design and implementation of an intelligent recommendation system for tourist attractions : The integration of EBM model , Bayesian network and Google Maps . Expert Systems with Applications , 39 ( 3 ) : 3257-3264. https : //doi.org/10.1016/j.eswa.2011.09.013

[ 17 ] Sabou , M. , Arsal , I. , Braşoveanu , A.M. ( 2013 ) . Tourmislod : A tourism linked datum band . Semantic Web , 4 ( 3 ) : 271-276 .

[ 18 ] Navrozidis , I. , Alexandridis , T.K. , Dimitrakos , A. , Lagopodi , A.L. , Moshou , D. , Zalidis , G. ( 2018 ) . Identification of purple blot disease on asparagus crops across spatial and spectral scales . Computer and Electronics in Agriculture , 148 : 322-329. https : //doi.org/10.1016/j.compag.2018.03.035

[ 19 ] Bin , S. , Sun , G. , Zhou , S. ( 2019 ) . Public view propagation example in social network based on blockchain . Journal of Applied Sciences – Electronics and Information Engineering , 37 ( 2 ) : 191-202 .