CN110851707B - Intelligent recommendation method for building material bidding platform - Google Patents

Intelligent recommendation method for building material bidding platform Download PDF

Info

Publication number
CN110851707B
CN110851707B CN201910963606.8A CN201910963606A CN110851707B CN 110851707 B CN110851707 B CN 110851707B CN 201910963606 A CN201910963606 A CN 201910963606A CN 110851707 B CN110851707 B CN 110851707B
Authority
CN
China
Prior art keywords
bidding
user
platform
documents
algorithm
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.)
Active
Application number
CN201910963606.8A
Other languages
Chinese (zh)
Other versions
CN110851707A (en
Inventor
须峰
宋安平
章珍顺
李传中
陈冉
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.)
Zhuke Network Technology Shanghai Co ltd
Original Assignee
Zhuke Network Technology Shanghai Co ltd
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 Zhuke Network Technology Shanghai Co ltd filed Critical Zhuke Network Technology Shanghai Co ltd
Priority to CN201910963606.8A priority Critical patent/CN110851707B/en
Publication of CN110851707A publication Critical patent/CN110851707A/en
Application granted granted Critical
Publication of CN110851707B publication Critical patent/CN110851707B/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
    • G06F16/90—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
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/906—Clustering; Classification
    • 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
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0631—Item recommendations
    • 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
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/08—Auctions

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • General Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention relates to an intelligent recommendation method for a building material bidding platform, which comprises the following steps: determining a clustering dimension of a bidding document in the bidding platform; clustering the bidding documents in the bidding platform by using a clustering algorithm; converting the behavior record of the user on the bidding platform into a user-project scoring matrix; populating the user-item scoring matrix with a factorizer-based negative sample pre-population algorithm; calculating the interest degree of the user to all the bidding documents by using an SVD algorithm; and sequencing the interest degrees of all the obtained bidding documents, and recommending n bidding documents with the highest interest degree for the user. The method and the device can effectively solve the problem that the single-type collaborative filtering negative sample is missing, and accurately recommend the interested bidding documents for the platform user.

Description

Intelligent recommendation method for building material bidding platform
Technical Field
The invention relates to the technical field of intelligent recommendation, in particular to an intelligent recommendation method for a building material bidding platform.
Background
In recent years, electronic commerce platforms, large video websites, even social networks, and the like, have widely used different recommendation algorithms in any scene where a special personalized service needs to be customized for a user. Therefore, in recent years, recommendation algorithms are developed quickly, and a wide variety of recommendation algorithms suitable for various scenes are developed. The current popular algorithms mainly include a rule-based recommendation algorithm, a content-based recommendation algorithm, a collaborative filtering-based recommendation algorithm and a hybrid recommendation algorithm.
The proposed algorithm is widely used in e-commerce platform, and the e-commerce platform in B2C (Business-to-Customer) model is developed very rapidly, and a large number of excellent B2C e-commerce platforms are developed in a short time, compared with the B2B (Business-to-Business) e-commerce platform, which is developed more slowly. Accordingly, the recommendation algorithm is mostly studied based on the B2C platform. Both buyers and sellers of the B2B platform are enterprises, which pay more attention to stability and long-term performance of transactions, product quality, the benefit of the enterprises needs to be ensured, and the requirements of buyers are relatively more fixed, while the recommendation algorithm based on the B2C platform is more inclined to recommend different and novel products for users. Therefore, the B2B platform service scenario is very different from the B2C platform, and therefore, the recommendation algorithm based on the B2C platform cannot be directly applied to the B2B platform.
Although domestic B2B E-commerce has been developed to a certain extent in recent years, the B2B E-commerce still has a large transformation space, and with the explosive increase of data volume, the B2B platform also needs a certain recommendation strategy in order to better meet the requirements of purchasing enterprises and suppliers. Currently, research on the recommendation system of the B2B platform is less, and the demand is more and more extensive with the development of the platform.
The building material bidding platform belongs to a typical B2B platform, and as the daily bid number of the platform and the number of registered suppliers increase, it is necessary to provide personalized intelligent bidding recommendation service for the suppliers. The bidding document as a recommended article has timeliness, and only the bidding document in the bidding stage has recommended value; due to the uniqueness of the bidding document, the personalized recommendation of the bidding document faces a serious cold start problem of the project.
Disclosure of Invention
The invention aims to provide an intelligent recommendation method for a building material bidding platform, which can effectively solve the problem of single collaborative filtering negative sample loss and accurately recommend interesting bidding documents for platform users.
The technical scheme adopted by the invention for solving the technical problems is as follows: the intelligent recommendation method for the building material bidding platform comprises the following steps:
(1) determining a clustering dimension of a bidding document in the bidding platform;
(2) clustering the bidding documents in the bidding platform by using a clustering algorithm;
(3) converting the behavior record of the user on the bidding platform into a user-project scoring matrix;
(4) populating the user-item scoring matrix with a factorizer-based negative sample pre-population algorithm;
(5) calculating the interest degree of the user to all the bidding documents by using an SVD algorithm;
(6) and sequencing the interest degrees of all the obtained bidding documents, and recommending n bidding documents with the highest interest degree for the user.
The clustering dimension of the bidding document in the step (1) comprises the bidding document type, the material category, the delivery area longitude and the delivery area latitude.
And (2) clustering the bidding documents in the bidding platform by adopting an improved k-protocols algorithm, dividing the dimensionality of the bidding documents into a numerical attribute, a classification attribute and a multi-level classification attribute during clustering, and calculating the distance of the bidding documents by mixing different distance formulas for different attributes.
When the bidding document distance is calculated, calculating the distance between the two bidding documents by adopting an Euclidean distance calculation formula for the numerical attribute; for the classification attribute, if the classification attributes of the two bidding documents are the same, the distance increment is zero, otherwise, the distance increment is I; and for the multi-stage classification attributes, sequentially judging whether each stage of classification attributes of the two labels are consistent, if M inconsistent classification attributes exist, increasing the distance by M/M I, wherein M is the stage number of the multi-stage classification attributes.
The step (4) specifically comprises the following substeps:
(41) filling corresponding interestingness for the user-item matrix according to different behavior categories by using the user behavior record;
(42) training the factorization machine model by using all positive samples to obtain the predicted score of the missing sample, and filling the user-project matrix;
(43) determining the number N of negative samples to be selected according to the activity of the useru；
(44) According to the popularity of the bidding documents, different weights are given to each bidding document;
(45) for each missing sample of each user, multiplying the predicted score by the weight of the standard corresponding to the missing sample to obtain a final score, and selecting N with the lowest scoreuTaking the individual bidding documents as negative samples;
(46) the resulting negative examples are populated into the original user-item matrix.
The intelligent recommendation method for the building material bidding platform further comprises the step of solving the interest degree of the user in the new bidding document through a clustering algorithm and an SVD algorithm when the new bidding document is released on the bidding platform, and specifically comprises the following steps:
(A) predicting the category of the new bidding document through the trained clustering model;
(B) calculating k historical bidding documents in the category which are closest to the new bidding document, and giving corresponding weights to the k historical bidding documents according to the distance;
(C) and (5) predicting the interest degree of the user in the new bidding document according to the interest degree of the user in the k historical bidding documents in the step (5).
Advantageous effects
Due to the adoption of the technical scheme, compared with the prior art, the invention has the following advantages and positive effects: the invention adopts the improved k-protocols algorithm, so that the k-protocols algorithm has wider applicability on the multi-level classification property. The negative sample pre-filling algorithm based on the factorization machine solves the problem of negative sample loss of single-type collaborative filtering, and compared with the existing negative sample pre-filling algorithm, the negative sample pre-filling algorithm based on the factorization machine has the advantages of more excellent effect and higher filling accuracy under the condition of sparse data. The method combines the clustering algorithm and the collaborative filtering recommendation algorithm, and effectively relieves the problem of cold start of the project in intelligent recommendation of the bidding documents.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of a negative sample pre-population algorithm based on a factorizer in the present invention;
FIG. 3 is a graph comparing the experimental results of the present invention method with the conventional SVD algorithm and other negative sample pre-population algorithms on the MovieLens dataset;
fig. 4 is a graph comparing the experimental effect of the method of the present invention on building a real-world data set with the conventional SVD algorithm and other negative sample pre-population algorithms.
Detailed Description
The invention will be further illustrated with reference to the following specific examples. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Further, it should be understood that various changes or modifications of the present invention may be made by those skilled in the art after reading the teaching of the present invention, and such equivalents may fall within the scope of the present invention as defined in the appended claims.
The embodiment of the invention relates to an intelligent recommendation method for a building material bidding platform, which is a method for learning interest preference of a user from data by a machine learning algorithm based on historical behavior data of the user on the bidding platform so as to intelligently recommend bidding documents to the user, and as shown in FIG. 1, the method comprises the following steps:
(1) and determining the clustering dimension of the bidding document according to the specific condition of the bidding platform. In this embodiment, the clustering dimension of the bidding document is selected from four dimensions of the bidding document type, the material type, the delivery area longitude and the delivery area latitude.
(2) And clustering the benchmarks by using an improved k-protocols algorithm, dividing the dimensionality of the benchmarks into a numerical attribute, a classification attribute and a multi-level classification attribute during clustering, and mixedly calculating the distance of the benchmarks by using different distance formulas for different attributes.
Specifically, the distance between the two books is initialized, so that the distance d between the two books is 0; the bidding document types are classified attributes, if the bidding document types of the two bidding documents are the same, the distance increment d + ═ 0, otherwise, the distance increment d + ═ 1; the material types (the shapes are 1101 and 1203, the first two digits represent the first-level type, the second two digits represent the second-level type, and the second-level type refers to the small type in the first-level type) of the label are classified attributes of two levels, whether the first-level type is consistent or not is judged firstly, and whether the second-level type is consistent or not is judged continuously if the first-level type is consistent; if the distance increment d + is consistent with 0, and only the first-level categories are consistent with 0.5, otherwise, the distance increment d + is 1; the regional longitude and latitude are numerical attributes and can be calculated by using an Euclidean distance formula. And finally, the distance between the two labels is the Euclidean distance plus the distance increment.
According to the embodiment, the distance of the bidding document is calculated by mixing different distance formulas for different attributes, so that the clustering algorithm can be suitable for multi-level classification attributes, and the applicability is improved.
(3) And converting the user behavior data into a user-item matrix R. In the embodiment, the records of checking, leaving a message, paying attention to and bidding for the bidding document by the user in the platform are extracted, if the behavior record exists on the bidding document by the user, 1 is filled in the corresponding position of the user-project matrix R, and otherwise, the record is defaulted to 0.
(4) The user-item scoring matrix is populated with a factoring machine based negative sample pre-population algorithm.
Specifically, the user behavior record is used for filling the corresponding interestingness for the user-item matrix according to different behavior categories, in the embodiment: recording interest degree of a user when viewing the bidding document as 1, recording interest degree of the user when leaving a message on the bidding document as 2, recording interest degree of the user when performing attention operation on the bidding document as 3, recording interest degree of the user when performing bidding operation on the bidding document as 5, and recording as 0 if the interest degree is lost; then, training the factorization machine model by using all positive samples to obtain the predicted score of the user on the missing samples, and filling a user-project matrix; determining the number N of negative samples to be selected according to the activity of the useru＝μ*PuWhere μ is the ratio of positive and negative samples, PuIs the number of positive samples of user u, representing the activity of user uJumping degree; according to the popularity of the bidding documents, different weights (the weight is between 0 and 1) are given to each bidding document, the bidding document with high popularity has low weight, otherwise, the weight is higher; for the missing samples of each user u in the original user-project matrix, multiplying the predicted score value obtained after the user-project matrix is filled according to the factorization machine by the weight of the standard book to obtain a final score, sequencing the final score, and selecting the N with the minimum score valueuThe individual item is taken as a negative sample; filling the negative samples into an original user-item matrix R to obtain a matrix
Figure BDA0002229765500000041
(5) And calculating the interest degree of the user in all the bidding documents by using an SVD algorithm.
(6) And if the new bidding document exists in the bidding platform and is issued, calculating the interest degree of the supplier in the new bidding document through a clustering algorithm and an SVD algorithm. In the embodiment, when a new bidding document is published, the new bidding document is firstly clustered into the kth class through a trained k-prototypes clustering model, and then n bidding documents closest to the newly published bidding document are found from the kth class. And then weights are respectively set for the n labels, and the closer the distance is, the greater the weight is, namely, the closer the similarity between the labels is. Calculating the interest degree of each supplier in the newly issued bidding documents according to the interest degree of each supplier in the n bidding documents
Figure BDA0002229765500000051
Wherein, wjThe weight of the jth label is in the range of 0-1, Pu,jAnd (5) the interest degree of the jth bidding document for the user u.
(7) And recommending the n bidding papers with the highest interest degree for the user.
The invention combines the clustering algorithm and the matrix decomposition algorithm, and effectively relieves the problem of severe cold start of the bidding platform applying the intelligent recommendation algorithm. As shown in fig. 3 and 4, the negative sample pre-filling algorithm a of the present invention has the lowest mean square error in the optimal segment part with hidden features of 15-25 compared to the conventional SVD algorithm B, the negative sample pre-filling algorithm C based on user activity and the negative sample pre-filling algorithm D based on matrix factorization, so that the negative sample pre-filling algorithm based on the factorization machine of the present invention performs better in the sparse data condition than the existing negative sample pre-filling algorithm.
The invention provides an intelligent recommendation method according with the actual situation of a bidding platform, which can learn the bidding preference of a supplier according to the behavior data of the supplier on the platform and accurately recommend a purchasing bidding document issued by the buyer to a supplier possibly interested in the bidding document so as to find a high-quality supplier and solve the problems of difficult bidding, difficult bidding and the like caused by insufficient information in the past. Meanwhile, the requirements of both suppliers and buyers are met, and the user experience of both parties is improved, so that the aims of promoting the viscosity of platform users and the benign development of platforms are fulfilled.

Claims (5)

1. An intelligent recommendation method for a building material bidding platform is characterized by comprising the following steps:
(1) determining a clustering dimension of a bidding document in the bidding platform;
(2) clustering the bidding documents in the bidding platform by using a clustering algorithm;
(3) converting the behavior record of the user on the bidding platform into a user-project scoring matrix;
(4) populating the user-item scoring matrix with a factorizer-based negative sample pre-population algorithm; the method specifically comprises the following steps:
(41) filling corresponding interestingness for the user-item matrix according to different behavior categories by using the user behavior record;
(42) training the factorization machine model by using all positive samples to obtain the predicted score of the missing sample, and filling the user-project matrix;
(43) determining the number N of negative samples to be selected according to the activity of the useru；
(44) According to the popularity of the bidding documents, different weights are given to each bidding document;
(45) for each userMissing samples, multiplying the predicted scores and the weights of the benchmarks corresponding to the missing samples to obtain final scores, and selecting N with the lowest scoreuTaking the individual bidding documents as negative samples;
(46) filling the obtained negative sample into an original user-item matrix;
(5) calculating the interest degree of the user to all the bidding documents by using an SVD algorithm;
(6) and sequencing the interest degrees of all the obtained bidding documents, and recommending n bidding documents with the highest interest degree for the user.
2. The intelligent recommendation method for a building material bidding platform according to claim 1, wherein the clustering dimension of the bidding in step (1) comprises bidding type, material category, shipping area longitude and shipping area latitude.
3. The intelligent recommendation method for building material bidding platforms according to claim 1, wherein in the step (2), the bidding documents in the bidding platform are clustered by using a modified k-protocols algorithm, and in the clustering, the dimensions of the bidding documents are divided into numerical attributes, classification attributes and multi-level classification attributes, and the distances of the bidding documents are mixedly calculated by using different distance formulas for different attributes.
4. The intelligent recommendation method for a building material bidding platform according to claim 3, wherein in calculating the bidding document distance, the distance between two bidding documents is calculated by using the Euclidean distance calculation formula for the numerical attribute; for the classification attribute, if the classification attributes of the two bidding documents are the same, the distance increment is zero, otherwise, the distance increment is I; and for the multi-stage classification attributes, sequentially judging whether each stage of classification attributes of the two labels are consistent, if M inconsistent classification attributes exist, increasing the distance by M/M I, wherein M is the stage number of the multi-stage classification attributes.
5. The intelligent recommendation method for the building material bidding platform according to claim 1, further comprising the step of calculating the interest degree of the user in the new bidding document through a clustering algorithm and an SVD algorithm when the bidding platform releases the new bidding document, specifically comprising the steps of:
(A) predicting the category of the new bidding document through the trained clustering model;
(B) calculating k historical bidding documents in the category which are closest to the new bidding document, and giving corresponding weights to the k historical bidding documents according to the distance;
(C) and (5) predicting the interest degree of the user in the new bidding document according to the interest degree of the user in the k historical bidding documents in the step (5).
CN201910963606.8A 2019-10-11 2019-10-11 Intelligent recommendation method for building material bidding platform Active CN110851707B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910963606.8A CN110851707B (en) 2019-10-11 2019-10-11 Intelligent recommendation method for building material bidding platform

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910963606.8A CN110851707B (en) 2019-10-11 2019-10-11 Intelligent recommendation method for building material bidding platform

Publications (2)

Publication Number Publication Date
CN110851707A CN110851707A (en) 2020-02-28
CN110851707B true CN110851707B (en) 2021-06-04

Family

ID=69597427

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910963606.8A Active CN110851707B (en) 2019-10-11 2019-10-11 Intelligent recommendation method for building material bidding platform

Country Status (1)

Country Link
CN (1) CN110851707B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112184401B (en) * 2020-09-22 2021-05-14 筑客网络技术（上海）有限公司 Intelligent matching method for building material bidding platform

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106326483A (en) * 2016-08-31 2017-01-11 华南理工大学 Collaborative recommendation method with user context information aggregation
CN107515909A (en) * 2017-08-11 2017-12-26 深圳市耐飞科技有限公司 A kind of video recommendation method and system
CN108334592A (en) * 2018-01-30 2018-07-27 南京邮电大学 A kind of personalized recommendation method being combined with collaborative filtering based on content

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8224818B2 (en) * 2010-01-22 2012-07-17 National Cheng Kung University Music recommendation method and computer readable recording medium storing computer program performing the method

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106326483A (en) * 2016-08-31 2017-01-11 华南理工大学 Collaborative recommendation method with user context information aggregation
CN107515909A (en) * 2017-08-11 2017-12-26 深圳市耐飞科技有限公司 A kind of video recommendation method and system
CN108334592A (en) * 2018-01-30 2018-07-27 南京邮电大学 A kind of personalized recommendation method being combined with collaborative filtering based on content

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
基于K-prototypes的混合属性数据聚类算法;陈韡等;《计算机应用》;20100801;第30卷(第8期);正文第1-3节 *
基于协同过滤算法的推荐系统研究与应用;李世伟;《中国优秀硕士学位论文全文数据库信息科技辑》;20180715;正文第3-4章 *
融合因子分解机和用户行为预测的音乐推荐;潘洋等;《计算机工程与应用》;20170901;第53卷(第17期);正文第2-4节 *

Also Published As

Publication number Publication date
CN110851707A (en) 2020-02-28

Similar Documents

Publication Publication Date Title
CN106651546B (en) Electronic commerce information recommendation method oriented to smart community
CN108629665B (en) Personalized commodity recommendation method and system
CN107833117B (en) Bayesian personalized sorting recommendation method considering tag information
JP6356744B2 (en) Method and system for displaying cross-website information
WO2020147594A1 (en) Method, system, and device for obtaining expression of relationship between entities, and advertisement retrieval system
CN109064285B (en) Commodity recommendation sequence and commodity recommendation method
CN111553754A (en) Updating method and device of behavior prediction system
CN101454771A (en) System and method of segmenting and tagging entities based on profile matching using a multi-media survey
CN102902691A (en) Recommending method and recommending system
US20180308152A1 (en) Data Processing Method and Apparatus
CN112184401B (en) Intelligent matching method for building material bidding platform
CN102542490A (en) Commodity recommendation method based on model matching
CN103559622A (en) Characteristic-based collaborative filtering recommendation method
CN108109058B (en) Single-classification collaborative filtering method integrating personality traits and article labels
CN113407834A (en) Knowledge graph-assisted user multi-dimensional interest extraction method
CN114969566B (en) Distance-measuring government affair service item collaborative filtering recommendation method
CN112102029A (en) Knowledge graph-based long-tail recommendation calculation method
CN116541607A (en) Intelligent recommendation method based on commodity retrieval data analysis
CN108920647A (en) Low-rank matrix based on spectral clustering fills TOP-N recommended method
WO2020119017A1 (en) System and method for achieving data asset sensing and pricing functions in big data background
CN110851707B (en) Intelligent recommendation method for building material bidding platform
CN113434778A (en) Recommendation method based on regularization framework and attention mechanism
CN110570226A (en) scoring prediction method combining topic model and heterogeneous information network
CN111368216A (en) Movie and television recommendation method based on mixed collaborative filtering
CN106021558A (en) Calculation method for user availability in collaborative filtering recommendation system

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