CN112269882A - Tourist attraction recommendation method oriented to knowledge map - Google Patents

Tourist attraction recommendation method oriented to knowledge map Download PDF

Info

Publication number
CN112269882A
CN112269882A CN202011084884.5A CN202011084884A CN112269882A CN 112269882 A CN112269882 A CN 112269882A CN 202011084884 A CN202011084884 A CN 202011084884A CN 112269882 A CN112269882 A CN 112269882A
Authority
CN
China
Prior art keywords
scenic spot
tourist
recommendation
attraction
user
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202011084884.5A
Other languages
Chinese (zh)
Other versions
CN112269882B (en
Inventor
苏雪平
朱丹尧
任劼
陈宁
李云红
高蒙
段嘉伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Polytechnic University
Original Assignee
Xian Polytechnic University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Polytechnic University filed Critical Xian Polytechnic University
Priority to CN202011084884.5A priority Critical patent/CN112269882B/en
Publication of CN112269882A publication Critical patent/CN112269882A/en
Application granted granted Critical
Publication of CN112269882B publication Critical patent/CN112269882B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/951—Indexing; Web crawling techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/14—Travel agencies

Abstract

The invention discloses a tourist attraction recommendation method facing to a knowledge map, which is specifically carried out according to the following steps; step 1: collecting scenic spot data, determining entities in the tourist field, defining entity attributes and value domains, and importing a Neo4j database to generate a tourist scenic spot knowledge map; step 2: training a Transd knowledge representation model; and step 3: constructing a tourist attraction recommendation model; the tourist attraction recommendation model specifically comprises the steps of constructing an attraction and attraction score matrix, constructing a user interest model, performing fusion calculation and generating a recommendation list. The tourist attraction recommendation method oriented to the knowledge map provides auxiliary information by utilizing semantic information of the knowledge map, supplements and describes relevant information of a user and an attraction, enhances the consistency of data, and improves the accuracy of a recommendation algorithm; therefore, accuracy and diversity of scenic spot recommendation are improved.

Description

Tourist attraction recommendation method oriented to knowledge map
Technical Field
The invention belongs to the technical field of tourist attraction recommendation, and particularly relates to a tourist attraction recommendation method oriented to a knowledge map.
Background
With the increasing economic level of people, people are pursuing higher quality of life. The tourism not only can release the working pressure and relieve the mood, but also can widen the eyes. At present, tourist attractions are recommended to bring convenience to travel of users. Although with the help of the tourism website, the user can know the tourism information of all places at any time and any place, select favorite tourism products and make a travel plan meeting the requirements. However, due to the serious information overload and information dispersion, users cannot quickly find out the required travel information from the internet, and the efficiency of making travel decisions is low, so that the travel industry faces a great challenge. Tourist attraction recommendation services of large websites at home and abroad are analyzed, most recommendation modes are concentrated on packaged services, services such as individuation, socialization and instantaneity are lacked, and in combination, the user experience of the current tourist attraction recommendation needs to be further improved.
Disclosure of Invention
The invention aims to provide a scenic spot recommendation method facing a knowledge map, and solves a part of problems that user experience of current scenic spot recommendation needs to be further improved.
The technical scheme adopted by the invention is that,
a tourist attraction recommendation method facing to a knowledge map specifically comprises the following steps;
step 1: collecting scenic spot data, determining entities in the tourist field, defining entity attributes and value domains, and importing a Neo4j database to generate a tourist scenic spot knowledge map;
step 2: training a Transd knowledge representation model;
and step 3: constructing a tourist attraction recommendation model;
the tourist attraction recommendation model specifically comprises the steps of constructing an attraction and attraction score matrix, constructing a user interest model, performing fusion calculation and generating a recommendation list.
The present invention is also characterized in that,
in the step 1, the scenic spot data acquisition comprises the steps of acquiring required scenic spot information data from the Internet by utilizing a web crawler technology, and crawling data by using a Scapy frame in python.
In step 1, the entities in the tourism field include the region, the type of the scenic spot, the dynasty, the price of the entrance ticket, the official network of the scenic spot, the average score and the opening time.
In step 2, training a Transd knowledge representation model specifically according to the following formula (1):
hr＝hMr，tr＝tMr (1)，
wherein h and t are vector representations in the entity space, hr、trIs a vector representation of an entity in a hyperplane, MrIs a mapping matrix that maps entities from entity space to hyperplane r.
In step 3, the construction of the scenic spot and the scenic spot scoring matrix specifically comprises the following steps: analyzing the behavior data of the user, corresponding the scenic spots in the matrix with the entities in the knowledge map, and using a TransD model to learn and represent the entities and attributes in the knowledge map to obtain a group of low-dimensional dense real value vectors representing entities Ii＝(e1i，e2i，...，eni)TAnd facing to the entities with similarity when different relations, calculating the distance between the entities by using an Euclidean distance formula, and reducing the value range to (0, 1)]The similarity between entities is calculated by the following formula (4):
Figure BDA0002720039570000031
wherein, sim (I)i，Ij) Is an entity IiAnd entity IjSimilarity of (e)kiAnd ekjAre respectively entity IiAnd entity IjVector representation of
Then, the scenery and a semantic similarity matrix of the scenery are generated.
In step 3, generating a recommendation list specifically comprises:
calculating the score of the user on the scenic spot, forming a recommendation list according to the score, obtaining a final scenic spot and scenic spot similarity matrix by combining TransD model learning and similarity weight calculation, and calculating the similarity between the scenic spot to be predicted and each scenic spot in the scenic spot set scored by the user, wherein the predicted score is calculated according to the following formula (10):
Figure BDA0002720039570000032
wherein, sim (I)i，Ij) Is a scenery spot IiHe-scenic spots IjSimilarity of (D), Su，jFor users to scenery spots IjThe score of (a), (b), (c) is the set of sights scored by user u, and S (I, k) is the first k and IiThe most similar attractions.
The tourist attraction recommendation method based on the knowledge graph has the advantages that aiming at the problems of sparsity and cold start existing in a recommendation system, the knowledge graph of the tourist attraction is constructed, entity attributes in the knowledge graph are vectorized, and a user interest model fusing the entity attributes in the knowledge graph is designed. Experiments prove that the accuracy and diversity of the traditional recommendation algorithm are effectively improved, and the recommendation accuracy rate reaches 86%.
Drawings
FIG. 1 is a schematic diagram of an overall flow of a scenic spot recommendation method oriented to a knowledge base;
FIG. 2 is a schematic diagram of data collected in the scenic spot recommendation method oriented to the knowledge base.
Detailed Description
The tourist attraction recommendation method based on the knowledge map is described in detail below with reference to the accompanying drawings and specific embodiments.
The invention relates to a tourist attraction recommendation method facing to a knowledge map, which is specifically carried out according to the following steps;
step 1: collecting scenic spot data, determining entities in the tourist field, defining entity attributes and value domains, and importing a Neo4j database to generate a tourist scenic spot knowledge map;
step 2: training a Transd knowledge representation model;
and step 3: constructing a tourist attraction recommendation model;
the tourist attraction recommendation model specifically comprises the steps of constructing an attraction and attraction score matrix, constructing a user interest model, performing fusion calculation and generating a recommendation list.
The scenic spot recommendation method based on the knowledge map introduces the knowledge map into the scenic spot recommendation field, and effectively improves the accuracy and diversity of recommendation results.
The invention mainly comprises the following steps:
as shown in fig. 1 and fig. 2, (1) constructing a tourist attraction knowledge map: the method comprises the steps of collecting scenic spot data, determining entities and related concepts of a tourist field, defining entity attributes and value domains, and importing a Neo4j database to generate a tourist map; (2) training a TransD knowledge representation model; (3) and (3) personalized tourist attraction recommendation: the method comprises the steps of constructing a scenery spot and scenery spot scoring matrix, constructing a user and scenery spot scoring matrix, modeling user interest, performing fusion calculation and generating a recommendation list. The specific implementation mode is as follows:
1. construction of tourist attractions knowledge map
1.1 gathering data of scenic spots
A large amount of sight spot data on the network can be freely obtained. This patent utilizes web crawler technique to obtain required sight spot information data from the internet. This patent mainly uses the Scapy frame in python to carry out the data crawl.
1.2 determining entities and related concepts in the area of travel
The tourist attraction attributes comprise belonged areas, attraction types, belonged dynasties, ticket prices, attraction official nets, average scores, opening time and the like. The type of sight, the score of sight, and the attribute of sight location all affect the recommendation, for example, a sight may belong to different types. Therefore, the patent determines four concepts of 'sight', 'sight type', 'sight score' and 'sight position' as entities. The method adopts a Bi-LSTM + CRF (Bidirectional Long Short-Term Memory Conditional Random Fields) method to carry out entity identification.
1.3 defining entity attributes and value ranges
As shown in FIG. 1, the entity attributes in the ontology are divided into two categories, namely object attributes and data attributes. The object attribute is used for describing the relationship between the objects, and the value range of the object attribute is an entity object; the data attribute is used to describe the inherent attribute of the entity, and has transitivity, that is, the data attribute owned by the upper entity, and the lower entity also inherits the attribute, and the value range of the attribute is generally String.
Table 1 entity attribute table
Figure BDA0002720039570000051
Figure BDA0002720039570000061
1.4 importing Neo4j database to generate tourist attraction knowledge map
Neo4j is a graph-based, non-relational database that graphically visualizes structured data. When processing data, the speed is far higher than that of a relational database, and the data can be efficiently searched. Thus, this patent uses Neo4j to construct a travel knowledge map.
2. Training a Transd knowledge representation model;
h and t in the TransD model are vector representations in a solid space, hr、trIs a vector representation of an entity in a hyperplane, MrIs a mapping matrix that maps entities from entity space to hyperplane r. Their relationship is as follows:
hr＝hMr，tr＝tMr (1)，
in the relation space, through continuous adjustment, the head entity vector hrTranslating along the relation vector r can obtain a tail entity vector trI.e. so that hr+ r and trAs equal as possible. The distance between vectors is measured in terms of euclidean distance, which represents the distance between two points in an n-dimensional space. The loss function of the TransD model is defined as follows:
fr(h，t)＝||hr+r-tr||2 (2)，
wherein | | | purple hair2Is the 2 norm of the vector. In the actual training, all vectors in the model are normalized, so that | | h | survival of the naked eye2≤1，||t||2≤1，||r||2≤1，||hMr||2≤1，||tMr||2≤1。
3. Personalized tourist attraction recommendation
3.1. Constructing the scenic spots and a scenic spot scoring matrix,
analyzing the behavior data of the user, corresponding the scenic spots in the matrix with the entities in the knowledge map, and using a TransD model to learn and represent the entities and attributes in the knowledge map to obtain a group of low-dimensional dense real value vectors representing entities Ii＝(e1i，e2i...，eni)TAnd facing to the entities with similarity when different relations, calculating the distance between the entities by using an Euclidean distance formula, and reducing the value range to (0, 1)]The calculation formula of the similarity between the entities is as follows:
Figure BDA0002720039570000071
sim(Ii，Ij) Representing an entity IiAnd entity IjSimilarity of (e)kiAnd ekjAre respectively entity IiAnd entity IjIs represented by a vector of (a). Secondly, generating a scenery spot and a semantic similarity matrix of the scenery spot.
3.2. A user and sight spot scoring matrix is constructed,
the user and scenery spot scoring matrix is constructed by scoring scenery spots according to all users, scoring data of the user can reflect the preference degree of the scenery spots, and the preference degree of the user to each scenery spot is represented by calculating the weight of the scenery spot in all the scoring scenery spots of the user through the scoring matrix. Assume that in the recommendation algorithm, the user set is denoted as U ═ U1，U2...，UmThe tourist attraction set is expressed as I ═ I1，I2...，InAnd the constructed scoring matrix of the users and the scenic spots is represented as a matrix R of i multiplied by jm×n。
Using wpujRepresents the weight of the scoring scenery j in all scoring scenery of the user u, RujThe score of the user u to the scenery j is expressed, N (u) is expressed as a scenery set scored by the user u, the score of the scenery is divided by the sum of the scores of all scored scenery of the user, and the calculation formula is as follows:
Figure BDA0002720039570000081
3.3. the modeling of the interests of the user is performed,
aiming at the problem that entity attributes are not considered in recommendation based on the knowledge graph, the patent provides a novel user interest modeling method for fusing the entity attributes in the knowledge graph. By using the tourist attraction knowledge map, the interest of the user on the attraction attributes is mined from the attribute level, a user interest model is built, the user interest is represented in a finer granularity mode, the user interest is integrated into a recommendation system, and the recommendation performance is further improved.
The attention of the user to the scenic spot is often focused on the attributes of the scenic spot, and the interestingness of the user is calculated, that is, the vector representation of the user under the attributes of the scenic spot is calculated. And calculating the user interest vector according to the B-TransD model. First, define the user attribute three-tuple set:
Figure BDA0002720039570000082
wherein the content of the first and second substances,
Figure BDA0002720039570000083
the historical visiting scene set of the user u, and the h is the corresponding entity of the scene in the knowledge map.
In order to obtain finer granularity of user interest, the user's preference for the sight attributes needs to be known. The preference indicates that the user has different degrees of attention to different attributes, and obtaining the preference of the user first requires calculating the attention of the user to the attributes, that is, calculating the weight of each attribute in the historical visit sights. Given a vector of sights, attributes, and attribute values, attribute r in sight vaThe weights of (d) are as follows:
Figure BDA0002720039570000091
wherein h and raRespectively in the form of vectors of the scenic spots and the attributes obtained by the TransD training,
Figure BDA0002720039570000092
is a mapping matrix in TransD. Respectively calculating the weight of each attribute of the user, and expressing the weighted sum of all attribute values in the historical visit sight spot set as a user interest vector, wherein the expression is as follows:
Figure BDA0002720039570000093
wherein, taThe amount of attribute values is obtained for the TransD training. User interest is represented by vectors of related terms, rather than individual feature words, with fixed vector dimensions reducing the size of the parameters.
3.4. The fusion calculation is carried out on the basis of the calculation,
the interest level information of the user about the scenic spot is contained in the weight. The calculation of the similarity weight mainly comes from two parts: (1) calculating the weight of each scenic spot evaluated by the user in all the scored scenic spots according to the scoring matrix of the user and the scenic spots, and recording the weight as wp; (2) mining the interest of a user on the attributes of the scenic spots from the attribute level by using the tourist attraction knowledge map, and constructing a user interest model as ws; and finally, fusing weights wp and ws by adopting a proportion eta epsilon (0, 1) to obtain a corresponding weight value, wherein the value range is (0, 1), and the fusion weight formula is as follows:
W＝η*wp+(1-η)*ws (9)，
3.5. a list of recommendations is generated and,
and predicting the scores of the users to the scenic spots, and forming a recommendation list according to the predicted scores. The method comprises the steps of obtaining a final scenery spot and scenery spot similarity matrix by combining with TransD model learning and similarity weight calculation, wherein the essence of the matrix is that k nearest scenery spots of each scenery spot in a user scoring scenery spot set are constructed, wherein the scenery spots scored by a user are removed, then the similarity between a scenery spot to be predicted and each scenery spot in a user scoring scenery spot set is calculated, and a prediction scoring calculation formula is as follows:
Figure BDA0002720039570000101
wherein, sim (I)i，Ij) Is a scenery spot IiScenic spot IjSimilarity of (D), Su，jRepresenting the user's subtended view point IjN (u) represents the set of sights scored by user u, and S (I, k) represents the first k and IiThe most similar attractions. User's opposite scenery spot IjHigher score of (c), and scenic spots IiAnd IjThe higher the similarity, the higher PuiThe larger the value of (c).
The tourist attraction recommendation method oriented to the knowledge base is further described in detail through specific embodiments.
(1) Construction of tourist attractions knowledge map
1) The data flow in the script is controlled by an engine, and the whole process is as follows:
a. the engine first opens the travel website, starts resolving the domain name, obtains the server host address, then finds the spider that handles the website, and requests the first link to crawl from this spider.
b. The engine will fetch the links to be crawled from the spider and send them to the crawling module via the dispatcher to download the web page data in the form of requests.
c. The capturing module downloads the webpage data and transmits the data to the analyzing module.
d. The analysis module extracts webpage data, sends the project data to the storage module for output, sends the webpage links to the link filtering module, filters useless links, sends the useless links to the duplication elimination module, and conducts duplication elimination processing.
e. And adding the new link subjected to filtering and de-duplication into a download queue to continue to capture.
f. Repeating a through e until the scheduler has no more request inputs, the engine shuts down, and the crawl ends.
2) Entity identification:
a. and segmenting the collected data by utilizing a third-party Chinese word segmentation packet jieba provided by a python language.
b. And (3) sequence labeling:
sequence tagging is simply the giving of a string of sequences, tagging each element present in the sequence with a corresponding tag, and performing a deep analysis of the string of sequences by the tag.
c. And marking the related training and testing data to finish the primary extraction of the text data set.
d. And in the marking process of the initial extraction, marking the category of the entity according to the well-defined classification condition.
e. Adopting a BIO labeling method for training and testing text data to finish labeled words, and embedding the words by using a word2vec method to generate a 300-dimensional word vector matrix;
f. identification was performed using BilSTM-CRF.
(2) Training a TransD knowledge representation model;
and during model training, the positive and negative triples are effectively separated in a vector space by adopting the idea of maximum distance. The distance-based ranking error function is used as an optimization objective function for model training:
L＝∑(h，r，t)∈T∑(h′，r′，t′)∈T′[fr(h，t)+γ-fr′(h′，t′)]+ (3)，
in the objective function, [ alpha ]]+The method is a hinge loss function, and ensures that each accumulated subentry value is not a negative number, T is a set of positive triples in the knowledge graph, T' is a set of negative triples, gamma is a distance separating the positive triples from the negative triples, and is generally set to be 1, the loss of a correct triplet is close to 0, and the loss of an error triplet tends to be infinite. And aiming at the target function, a random gradient descent algorithm is adopted, and a minimized loss function and updated model parameters are obtained through iterative solution. The specific training steps are as follows:
a. randomly traversing a positive triple from the knowledge graph, and constructing a corresponding negative triple;
b. embedding the entity normalization in the positive and negative triples into a vector, setting a learning rate epsilon, calculating a gradient direction by learning each sample, and updating an iterative model parameter;
c. and (c) repeating the steps a and b to reach the maximum iteration times to obtain a minimized loss function.
(3) Personalized tourist attraction recommendation
Firstly, relieving the sparsity problem of data by utilizing semantic information rich in a knowledge map, embedding the scenery spot information into a low-dimensional vector space, and expressing by using a distributed vector to obtain a scenery spot-scenery spot similar matrix; secondly, constructing a user interest model and a user-scenery spot scoring matrix; and then, combining the similarity weight with the scenery spot-scenery spot similarity matrix for calculation, and obtaining a scenery spot prediction score to generate a recommendation list.
The invention relates to a tourist attraction recommendation method facing a knowledge map, which is used for mining the interests of a user by a recommendation system and providing personalized information service and decision support for the user when facing mass tourist data. The knowledge graph is a structured semantic knowledge base and contains abundant entity information and association information among entities. The semantic information of the knowledge graph is used for providing auxiliary information, the related information of the user and the scenic spots is described in a supplementing mode, the density of data is enhanced, and the accuracy of a recommendation algorithm is improved; and performing divergence mining according to different semantic relations of the project to improve a recommendation result.
In conclusion, the scenic spot recommendation method and the scenic spot recommendation system utilize the scenic spot knowledge map, and accuracy and diversity of scenic spot recommendation are improved.

Claims (6)

1. A tourist attraction recommendation method oriented to a knowledge graph is characterized by comprising the following steps of;
step 1: collecting scenic spot data, determining entities in the tourist field, defining entity attributes and value domains, and importing a Neo4j database to generate a tourist scenic spot knowledge map;
step 2: training a Transd knowledge representation model;
and step 3: constructing a tourist attraction recommendation model;
the tourist attraction recommendation model specifically comprises the steps of constructing an attraction and attraction score matrix, constructing a user interest model, performing fusion calculation and generating a recommendation list.
2. The method as claimed in claim 1, wherein in step 1, the gathering of the scenic spot data includes obtaining the required scenic spot information data from the internet by using web crawler technology, and using script framework in python to perform data crawling.
3. The method as claimed in claim 1, wherein in step 1, the entities of the tourist area include belonged area, type of scenic spot, era, ticket price, official network of scenic spot, average score and open time.
4. The method for recommending tourist attractions based on knowledge-graph as claimed in claim 1, wherein in step 2, the training fransd knowledge representation model is specifically represented by the following formula (1):
hr＝hMr，tr＝tMr (1)，
wherein h and t are vector representations in the entity space, hr、trIs a vector representation of an entity in a hyperplane, MrIs a mapping matrix that maps entities from entity space to hyperplane r.
5. The method for recommending tourist attractions based on knowledge-graph as claimed in claim 1, wherein in step 3, constructing the attraction and attraction scoring matrix specifically comprises: analyzing the behavior data of the user, corresponding the scenic spots in the matrix with the entities in the knowledge map, and using a TransD model to learn and represent the entities and attributes in the knowledge map to obtain a group of low-dimensional dense real value vectors representing entities Ii＝(e1i，e2i，...，eni)TAnd facing to the entities with similarity when different relations, calculating the distance between the entities by using an Euclidean distance formula, and reducing the value range to (0, 1)]The similarity between entities is calculated by the following formula (4):
Figure FDA0002720039560000021
wherein, sim (I)i，Ij) Is an entity IiAnd entity IjSimilarity of (e)kiAnd ekjAre respectively entity IiAnd entity IjVector representation of
Then, the scenery and a semantic similarity matrix of the scenery are generated.
6. The method as claimed in claim 1, wherein in step 3, the generating of the recommendation list specifically comprises:
calculating the score of the user on the scenic spot, forming a recommendation list according to the score, obtaining a final scenic spot and scenic spot similarity matrix by combining TransD model learning and similarity weight calculation, and calculating the similarity between the scenic spot to be predicted and each scenic spot in the scenic spot set scored by the user, wherein the predicted score is calculated according to the following formula (10):
Figure FDA0002720039560000022
wherein, sim (I)i，Ij) Is a scenery spot IiHe-scenic spots IjSimilarity of (D), Su，jFor users to scenery spots IjThe score of (a), (b), (c) is the set of sights scored by user u, and S (I, k) is the first k and IiThe most similar attractions.
CN202011084884.5A 2020-10-12 2020-10-12 Tourist attraction recommendation method oriented to knowledge map Active CN112269882B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011084884.5A CN112269882B (en) 2020-10-12 2020-10-12 Tourist attraction recommendation method oriented to knowledge map

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011084884.5A CN112269882B (en) 2020-10-12 2020-10-12 Tourist attraction recommendation method oriented to knowledge map

Publications (2)

Publication Number Publication Date
CN112269882A true CN112269882A (en) 2021-01-26
CN112269882B CN112269882B (en) 2022-10-18

Family

ID=74337824

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011084884.5A Active CN112269882B (en) 2020-10-12 2020-10-12 Tourist attraction recommendation method oriented to knowledge map

Country Status (1)

Country Link
CN (1) CN112269882B (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112801751A (en) * 2021-02-04 2021-05-14 重庆邮电大学 Personalized scenic spot recommendation method of multitask graph neural network
CN112992021A (en) * 2021-03-08 2021-06-18 合肥工业大学 Intelligent scenic spot navigation system and use method
CN113112058A (en) * 2021-03-30 2021-07-13 西安理工大学 Travel route recommendation method based on knowledge graph and ant colony algorithm
CN113220905A (en) * 2021-05-27 2021-08-06 哈尔滨理工大学 Service recommendation method fusing knowledge graph
CN113902354A (en) * 2021-12-03 2022-01-07 环球数科集团有限公司 Travel evaluation data processing method and device and computer equipment
CN115018474A (en) * 2022-08-03 2022-09-06 山东美丽乡村云计算有限公司 Text and travel consumption heat degree analysis method based on big data
CN116662663A (en) * 2023-06-14 2023-08-29 广州市派客朴食信息科技有限责任公司 User dish recommending method based on knowledge graph and deep learning
WO2023179689A1 (en) * 2022-03-24 2023-09-28 重庆邮电大学工业互联网研究院 Knowledge graph-based recommendation method for internet of things
CN116821692A (en) * 2023-08-28 2023-09-29 北京化工大学 Method, device and storage medium for constructing descriptive text and space scene sample set

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016107326A1 (en) * 2014-12-30 2016-07-07 北京奇虎科技有限公司 Search recommending method and device based on search terms
CN107729444A (en) * 2017-09-30 2018-02-23 桂林电子科技大学 Recommend method in a kind of personalized tourist attractions of knowledge based collection of illustrative plates
CN109255033A (en) * 2018-11-05 2019-01-22 桂林电子科技大学 A kind of recommended method of the knowledge mapping based on location-based service field
CN110288436A (en) * 2019-06-19 2019-09-27 桂林电子科技大学 A kind of personalized recommending scenery spot method based on the modeling of tourist's preference
CN110287336A (en) * 2019-06-19 2019-09-27 桂林电子科技大学 A kind of tourist's portrait construction method recommended towards tourist attractions
CN110532464A (en) * 2019-08-13 2019-12-03 桂林电子科技大学 A kind of tourism recommended method based on more tourism context modelings

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016107326A1 (en) * 2014-12-30 2016-07-07 北京奇虎科技有限公司 Search recommending method and device based on search terms
CN107729444A (en) * 2017-09-30 2018-02-23 桂林电子科技大学 Recommend method in a kind of personalized tourist attractions of knowledge based collection of illustrative plates
CN109255033A (en) * 2018-11-05 2019-01-22 桂林电子科技大学 A kind of recommended method of the knowledge mapping based on location-based service field
CN110288436A (en) * 2019-06-19 2019-09-27 桂林电子科技大学 A kind of personalized recommending scenery spot method based on the modeling of tourist's preference
CN110287336A (en) * 2019-06-19 2019-09-27 桂林电子科技大学 A kind of tourist's portrait construction method recommended towards tourist attractions
CN110532464A (en) * 2019-08-13 2019-12-03 桂林电子科技大学 A kind of tourism recommended method based on more tourism context modelings

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
DEHUA CHEN等: ""Clinical Knowledge Graph Embeddings with Hierarchical Structure for Thyroid Treatment Recommendation"", 《2019 IEEE INTL CONF ON DEPENDABLE, AUTONOMIC AND SECURE COMPUTING, INTL CONF ON PERVASIVE INTELLIGENCE AND COMPUTING, INTL CONF ON CLOUD AND BIG DATA COMPUTING, INTL CONF ON CYBER SCIENCE AND TECHNOLOGY CONGRESS (DASC/PICOM/CBDCOM/CYBERSCITECH)》 *
贾中浩等: "旅游知识图谱特征学习的景点推荐", 《智能系统学报》 *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112801751A (en) * 2021-02-04 2021-05-14 重庆邮电大学 Personalized scenic spot recommendation method of multitask graph neural network
CN112801751B (en) * 2021-02-04 2022-12-23 重庆邮电大学 Personalized scenic spot recommendation method of multitask graph neural network
CN112992021A (en) * 2021-03-08 2021-06-18 合肥工业大学 Intelligent scenic spot navigation system and use method
CN113112058B (en) * 2021-03-30 2023-07-18 西安理工大学 Tourism route recommendation method based on knowledge graph and ant colony algorithm
CN113112058A (en) * 2021-03-30 2021-07-13 西安理工大学 Travel route recommendation method based on knowledge graph and ant colony algorithm
CN113220905A (en) * 2021-05-27 2021-08-06 哈尔滨理工大学 Service recommendation method fusing knowledge graph
CN113902354A (en) * 2021-12-03 2022-01-07 环球数科集团有限公司 Travel evaluation data processing method and device and computer equipment
WO2023179689A1 (en) * 2022-03-24 2023-09-28 重庆邮电大学工业互联网研究院 Knowledge graph-based recommendation method for internet of things
CN115018474B (en) * 2022-08-03 2022-11-08 山东美丽乡村云计算有限公司 Text and travel consumption heat degree analysis method based on big data
CN115018474A (en) * 2022-08-03 2022-09-06 山东美丽乡村云计算有限公司 Text and travel consumption heat degree analysis method based on big data
CN116662663A (en) * 2023-06-14 2023-08-29 广州市派客朴食信息科技有限责任公司 User dish recommending method based on knowledge graph and deep learning
CN116662663B (en) * 2023-06-14 2024-04-30 广州市派客朴食信息科技有限责任公司 User dish recommending method based on knowledge graph and deep learning
CN116821692A (en) * 2023-08-28 2023-09-29 北京化工大学 Method, device and storage medium for constructing descriptive text and space scene sample set

Also Published As

Publication number Publication date
CN112269882B (en) 2022-10-18

Similar Documents

Publication Publication Date Title
CN112269882B (en) Tourist attraction recommendation method oriented to knowledge map
CN109492157B (en) News recommendation method and theme characterization method based on RNN and attention mechanism
Wang et al. What your images reveal: Exploiting visual contents for point-of-interest recommendation
An et al. Design of recommendation system for tourist spot using sentiment analysis based on CNN-LSTM
CN108363804B (en) Local model weighted fusion Top-N movie recommendation method based on user clustering
Shen et al. Attraction recommendation: Towards personalized tourism via collective intelligence
CN110555112A (en) interest point recommendation method based on user positive and negative preference learning
CN113158023B (en) Public digital life accurate classification service method based on mixed recommendation algorithm
CN113505204B (en) Recall model training method, search recall device and computer equipment
Li et al. Where you instagram? associating your instagram photos with points of interest
CN108897750B (en) Personalized place recommendation method and device integrating multiple contextual information
Chen et al. Joint text embedding for personalized content-based recommendation
Kumar et al. Word semantics based 3-d convolutional neural networks for news recommendation
Jiang et al. Boosting facial expression recognition by a semi-supervised progressive teacher
CN115687760A (en) User learning interest label prediction method based on graph neural network
Zhang et al. SEMA: Deeply learning semantic meanings and temporal dynamics for recommendations
Yin et al. Gps2vec: Pre-trained semantic embeddings for worldwide gps coordinates
CN113239159A (en) Cross-modal retrieval method of videos and texts based on relational inference network
Mikhailov et al. Smartphone-based tourist trip planning system: a context-based approach to offline attraction recommendation
Zhang et al. Bilinear graph neural network-enhanced Web services classification
Kim et al. High-quality train data generation for deep learning-based web page classification models
Valliyammai et al. An intelligent personalized recommendation for travel group planning based on reviews
CN113869034B (en) Aspect emotion classification method based on reinforced dependency graph
Feng Design of tourism intelligent recommendation model of mount tai scenic area based on knowledge graph
CN112257517B (en) Tourist attraction recommendation system based on attraction clustering and group emotion recognition

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant