CN114881689A - Building recommendation method and system based on matrix decomposition - Google Patents
Building recommendation method and system based on matrix decomposition Download PDFInfo
- Publication number
- CN114881689A CN114881689A CN202210446569.5A CN202210446569A CN114881689A CN 114881689 A CN114881689 A CN 114881689A CN 202210446569 A CN202210446569 A CN 202210446569A CN 114881689 A CN114881689 A CN 114881689A
- Authority
- CN
- China
- Prior art keywords
- building
- brand
- matrix
- scoring
- feature vector
- 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.)
- Pending
Links
- 239000011159 matrix material Substances 0.000 title claims abstract description 110
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 title claims abstract description 27
- 239000013598 vector Substances 0.000 claims abstract description 28
- 238000004364 calculation method Methods 0.000 claims abstract description 21
- 230000004927 fusion Effects 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 238000012163 sequencing technique Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000009825 accumulation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 239000006227 byproduct Substances 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Images
Classifications
-
- 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0242—Determining effectiveness of advertisements
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/16—Real estate
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Tourism & Hospitality (AREA)
- Educational Administration (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a building recommendation method and system based on matrix decomposition, which comprises the following steps: step 1: acquiring a building portrait and a brand scheme; step 2: establishing a brand-building scoring matrix according to the building portrait and the brand scheme; and step 3: performing scoring prediction on brand-building through matrix decomposition; and 4, step 4: fusing the scoring matrix and the weight matrix, and weighting the scores of all the brands for the building; and 5: using the building dense feature vector after matrix decomposition for building similarity calculation, and using the brand dense feature vector after matrix decomposition for brand similarity calculation; step 6: and recalling the buildings for the brands according to the extracted strong rules, sorting the buildings from high to low according to the scores, and then recommending the buildings. The method and the system can be suitable for the scene of offline advertisement delivery, and can be widely used for the recommendation tasks of online business operations to the recommendation of offline advertisement resources, thereby realizing the perfect combination of online models and offline services.
Description
Technical Field
The invention relates to the technical field of matrix decomposition, in particular to a building recommendation method and system based on matrix decomposition.
Background
With the rapid development of offline advertising services in recent years, advertisers have made more demands for offline advertising refinement services. Especially in the scene of resource screening of the offline advertisements, a clear reference scheme is lacked all the time, and an intelligent one-stop recommendation service is provided for advertisers.
Patent document CN113868462A (application number: CN202111065891.5) discloses a song recommendation system and method based on matrix decomposition, relating to the technical field of data analysis; performing matrix decomposition on the preprocessed song characteristic data and the user behavior data by using a matrix decomposition model, acquiring a preference index of the user to the song by using a Bernoulli probability distribution model according to the decomposed data matrix, and recommending the song to the user according to the preference index.
With the continuous improvement of the data quantity and quality of offline advertisement services, the existing data can already meet the fine requirements of advertisers on resource layers, so if the intelligent recommendation of offline resources can be completed by combining the recommendation schemes of online merchants and other platforms and the inherent characteristics of the offline resources, the optimal solution for advertisement placement is undoubtedly provided, and meanwhile, great benefits are brought to resource allocation and integration, so that an intelligent recommendation scheme based on the offline resources is urgently needed.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a building recommendation method and system based on matrix decomposition.
The building recommendation method based on matrix decomposition provided by the invention comprises the following steps:
step S1: acquiring a building portrait and a brand scheme, wherein the building portrait and the brand scheme comprise building audience crowd attributes, self tags, advertisement owner releasing preferences, brand rules and historical releasing plans;
step S2: establishing a brand-building scoring matrix according to the building portrait and the brand scheme;
step S3: performing scoring prediction on brand-building through matrix decomposition;
step S4: fusing the scoring matrix and the weight matrix, and weighting the scores of all the brands for the building;
step S5: using the building dense feature vector after matrix decomposition for building similarity calculation, and using the brand dense feature vector after matrix decomposition for brand similarity calculation;
step S6: and recalling the buildings for the brands according to the extracted strong rules, sorting the buildings from high to low according to the scores, and then recommending the buildings.
Preferably, the expression of the brand-building scoring matrix is as follows:
wherein: m is the generated scoring matrix; P/T, where P is the number of selected points and T is the total number of points in the building;
establishing a rule matrix according to a brand rule, wherein the expression is as follows:
wherein: w represents the rule weight of the brand, and when the building which does not meet the preset condition needs to be avoided, w is set to be 0. Preferably, the expression of the scoring prediction is:
y(t)＝g(t)+s(t)+h(t)+ε t
wherein: g (t) is a trend term, s (t) is a period term, h (t) is a holiday term, epsilon t And (4) bringing the data into a model for a noise item to generate a predicted value y at the predicted value t.
Preferably, the weighting expression in step S4 is:
preferably, the similarity calculation expression is:
wherein: the cos θ is the degree of similarity,
in order to be the feature vector n,
is a feature vector n-1, | | B n I and B n- And | represents their modulus, respectively.
The building recommendation system based on matrix decomposition provided by the invention comprises the following components:
module M1: acquiring a building portrait and a brand scheme, wherein the building portrait and the brand scheme comprise building audience crowd attributes, self tags, advertisement owner releasing preferences, brand rules and historical releasing plans;
module M2: establishing a brand-building scoring matrix according to the building portrait and the brand scheme;
module M3: performing scoring prediction on brand-building through matrix decomposition;
module M4: fusing the scoring matrix and the weight matrix, and weighting the scores of all the brands for the building;
module M5: using the building dense feature vector after matrix decomposition for building similarity calculation, and using the brand dense feature vector after matrix decomposition for brand similarity calculation;
module M6: and recalling the buildings for the brands according to the extracted strong rules, sorting the buildings from high to low according to the scores, and then recommending the buildings.
Preferably, the expression of the brand-building scoring matrix is as follows:
wherein: m is the generated scoring matrix; P/T, where P is the number of selected points and T is the total number of points in the building;
establishing a rule matrix according to a brand rule, wherein the expression is as follows:
wherein: w represents the rule weight of the brand, and when the building which does not meet the preset condition needs to be avoided, w is set to be 0. Preferably, the expression of the scoring prediction is:
y(t)＝g(t)+s(t)+h(t)+ε t
wherein: g (t) is a trend term, s (t) is a period term, h (t) is a holiday term, epsilon t And (4) bringing the data into a model for a noise item to generate a predicted value y at the predicted value t.
Preferably, the weighting expression in the module M4 is: e ═ M ° R.
Preferably, the similarity calculation expression is:
wherein: the cos θ is the degree of similarity,
in order to be the feature vector n,
is a feature vector n-1, | | B n I and B n- And | represents their modulus, respectively.
Compared with the prior art, the invention has the following beneficial effects:
1. the method can be suitable for the scene of offline advertisement delivery, and can be widely used for the recommendation tasks of online business operations to the recommendation of offline advertisement resources, so that the perfect combination of an online model and offline services is realized;
2. the weight module designed by the invention is a powerful supplement to the application of the traditional recommendation algorithm to the offline advertisement delivery scene, the external characteristics such as strong rules, preference scores and the like are usually added to the service requirements of advertisers, and the supplement of the weight matrix enables the recommendation result to be closer to the service per se according to the interpretability;
3. the data byproduct of the invention can be used for cluster analysis and similarity calculation of the advertiser and the project resource, which is more beneficial to the resource integration of the company;
4. the method designed by the invention has a set of complete closed-loop links to correct the model, can continuously strengthen and correct by self, and finally continuously improves the service development.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a block diagram of the core algorithm of the present invention;
FIG. 3 is a flow chart of model modification according to the present invention.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
Example (b):
as shown in fig. 1, the invention provides a building recommendation scheme based on matrix decomposition, and the implementation and deployment mode of the method comprises the following steps:
step S1: and generating a building portrait and a brand scheme, wherein the building portrait and the brand scheme comprise building audience crowd attributes, self tags, advertisement owner delivery preferences, brand rules and historical delivery plans. Building portrait, including building nature statistics, DMP label and audience attribute, building portrait needs accumulation of data in time and fusion in space, including building marking task, audience statistics task, etc., and after fusion, obtaining project resource label vector, the dimension in the vector represents the confidence of the label corresponding to the building;
step S2: calculating a brand-building scoring matrix according to the building portrait and the brand scheme, establishing a rule matrix according to a brand rule, fusing the matrix of the building and the matrix of the brand, linking according to a historical release scheme and historical preference expressed by a customer, and calculating a score;
step S3: and (3) forecasting the brand-building score is completed through matrix decomposition, and extraction of strong rules is completed according to the rule matrix. Completing calculation of matrix decomposition through spark ALS, and extracting a strong rule through a characteristic that a threshold value in a rule matrix is equal to 1;
step S4: fusing the scoring matrix and the weight matrix, weighting the scores of all the brands for the building, adding the matrixes and bringing the added scores into a kernel function, namely multiplying the corresponding scores by the corresponding weights to finally obtain a comprehensive score;
step S5: the building dense feature vectors after matrix decomposition can be continuously used for building similarity calculation, and the brand dense feature vectors are used for brand similarity calculation. Except for the recommendation score, the brand similarity and the building similarity can be calculated according to the decomposed vector to perfect the cold start problem;
step S6: and recalling part of buildings for the brand according to the extracted strong rules, and sequencing the buildings from high to low according to the scores, recommending Top 10 buildings, wherein the number of recalls and the sequencing rules can be dynamically adjusted.
The expression of the brand-building scoring matrix is as follows:
wherein: m is the generated scoring matrix; P/T, where P is the number of selected points and T is the total number of points in the building;
establishing a rule matrix according to a brand rule, wherein the expression is as follows:
wherein: w represents the rule weight of the brand, and when the building which does not meet the preset condition needs to be avoided, w is set to be 0.
The expression of the scoring prediction is：y(t)＝g(t)+s(t)+h(t)+ε t (ii) a Wherein: g (t) is a trend term, s (t) is a period term, h (t) is a holiday term, epsilon t And (4) bringing the data into a model for a noise item to generate a predicted value y at the predicted value t.
The weighting expression in step S4 is: e ═ M ° R.
The similarity calculation expression is as follows:
wherein: the cos θ is the degree of similarity,
in order to be the feature vector n,
is a feature vector n-1, | | B n I and B n- And | represents their modulus, respectively.
The steps not only introduce the mainstream recommendation model to score and predict the building resources, but also customize the offline advertisement putting scene. Through data accumulation and building portrait data, scoring systems before advertisers and projects can be established, but the characteristic of a sparse matrix still exists, namely no brand scoring is carried out on projects in projects which are not put, and prediction cannot be finished. The scores are predicted through traditional matrix decomposition, and because the offline advertisements and the online advertisements have many differences, including many limits on cities, places, competitive products and the like, the limits become weight matrixes and recall rule engines respectively to be integrated into the whole recommendation system. The method can well combine the online and offline modes together, and the recommendation system is close to the actual existing business.
The method is applied to data recommendation of various resources, can be further expanded to recommend point locations, can also recommend areas and business circles, and can be applied to the mode for any data.
The method comprises the steps of firstly completing drawing of building figures and brand schemes, then completing construction of a real building matrix correlation matrix, and finally completing recall and sequencing. Taking the matrix decomposition procedure of buildings and brands as an example, as shown in fig. 2, the design of a matrix algorithm is completed, including a basic scoring matrix and a weight matrix, and these data are the basis of all calculations. Then, a Spark matrix decomposition program is designed and developed, a parameter rank is designed to be 50, the parameter is a characteristic vector dimension, the fitting degree is not enough when the value is too small, the generalization capability is weak when the value is too large, and the loss after the test 50 is the minimum; iteration is 10, and the parameter is Iteration number; lambda is 0.0001, lambda is a regularization coefficient, the larger the coefficient is, the greater the penalty degree is, and the greater the influence on dimensionality reduction and weakening index is. And after the super-parameters are set, decomposing and calculating the existing matrix to obtain dense features related to the building and the brand, performing cross multiplication on the two feature matrices to complete the restoration of the matrix, and generating a new scoring item in a prediction unit in the newly generated matrix. The score is typically used to recall and rank the brand for the building, but due to the specificity of the offline advertisement, consideration is needed for both strong and weak rules. And adding general measurable rule indexes such as audience requirements and rent requirements into the weight matrix, generating the data from the DMP system, and fusing the data into the prediction matrix to finish final scoring. And for the proprietary strong rules, a rule engine is arranged in the recall stage, the rules are recalled for different customers, and then score sequencing is carried out to meet the pushing requirements of different customers in different scenes.
The above is only a design and development sample of one scenario in the recommendation system, and the same is true of the recommendation practice of other scenarios. The invention has lower development cost and can be completed only by the existing big data frame; the invention can effectively solve the business problem of offline delivery, and realizes online and offline combination; the method has strong expansibility, can finish the construction of any Spark cluster, and only needs to carry out customized cleaning on the existing data. In conclusion, the building recommendation scheme based on matrix decomposition has a very wide application prospect. It should be noted that, in the implementation process of the invention, besides the establishment of the Spark general environment, the method only needs to customize the data matrix conversion pipeline, the model can be migrated, and the method can be reused.
The invention provides a building recommendation system based on matrix decomposition, which comprises:
module M1: the data cleaning module is used for customizing building portrait and brand scheme through the DMP platform;
module M2: the matrix generation module is used for fusing various data indexes and converting the data indexes into various types of matrixes;
module M3: the Spark algorithm module is used for constructing a matrix decomposition module through Spark to complete ALS calculation;
module M4: a weighted fusion module to highlight the current preferences of the user;
module M5: the recall sorting module is used for fusing the weight matrix module and the scoring matrix, then completing recall filtering through a strong rule, and sorting recommendation through scores;
module M6: and the closed loop correction module dynamically and continuously updates the three matrixes through the buried point data and the user tendency to complete the iteration of the model.
The module M1 can clean and fuse various data, collect building portrait related data, collect brand release strategy related data, and finally fuse into building dimensionality and brand dimensionality portrait data.
The module M2 is a fusion of building portrait data and brand impression data to generate a scoring matrix of brand buildings, which represents the preference of the brand for the building through a single score dimension.
The module M3 performs decomposition of the M2 output matrix by developing the Spark task and generates byproduct building dense vectors and brand dense vectors while performing prediction of scores.
The module M4 is to generate a weighting matrix with the same dimension as the M3 prediction matrix, which weights the regular preference of the brand.
The module M5 completes generation of the final matrix by fusion of the weight matrix and the prediction matrix, and completes filtering in a way of recalling the matching rule for the restriction rule, the industry taboo and the like.
The above-mentionedThe module M6 influences the prediction matrix and the weight matrix by embedding data and supplementing offline rules, and each time the result is added again
Generating
Thus continuously optimizing the existing recommendation model, as in fig. 3.
Those skilled in the art will appreciate that, in addition to implementing the systems, apparatus, and various modules thereof provided by the present invention in purely computer readable program code, the same procedures can be implemented entirely by logically programming method steps such that the systems, apparatus, and various modules thereof are provided in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system, the device and the modules thereof provided by the present invention can be considered as a hardware component, and the modules included in the system, the device and the modules thereof for implementing various programs can also be considered as structures in the hardware component; modules for performing various functions may also be considered to be both software programs for performing the methods and structures within hardware components.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (10)
1. A building recommendation method based on matrix decomposition is characterized by comprising the following steps:
step S1: acquiring a building portrait and a brand scheme, wherein the building portrait and the brand scheme comprise building audience crowd attributes, self tags, advertisement owner releasing preferences, brand rules and historical releasing plans;
step S2: establishing a brand-building scoring matrix according to the building portrait and the brand scheme;
step S3: performing scoring prediction on brand-building through matrix decomposition;
step S4: fusing the scoring matrix and the weight matrix, and weighting the scores of all the brands for the building;
step S5: using the building dense feature vector after matrix decomposition for building similarity calculation, and using the brand dense feature vector after matrix decomposition for brand similarity calculation;
step S6: and recalling the buildings for the brands according to the extracted strong rules, sorting the buildings from high to low according to the scores, and then recommending the buildings.
2. The matrix factorization based building recommendation method of claim 1, wherein said brand-building scoring matrix is expressed as:
wherein: m is the generated scoring matrix; P/T, where P is the number of selected points and T is the total number of points in the building;
establishing a rule matrix according to a brand rule, wherein the expression is as follows:
wherein: w represents the rule weight of the brand, and when the building which does not meet the preset condition needs to be avoided, w is set to be 0.
3. The matrix factorization based building recommendation method of claim 1, wherein the scoring prediction is expressed as:
y(t)＝g(t)+s(t)+h(t)+ε t
wherein: g (t) is a trend term, s (t) is a period term, h (t) is a holiday term, epsilon t For noise terms, the data is brought into the model to generate predicted valuesAnd predicting the value y at the time t.
4. The matrix factorization-based building recommendation method of claim 2, wherein the weighting expression in step S4 is:
5. the matrix factorization based building recommendation method of claim 1, wherein the similarity calculation expression is:
wherein: the cos θ is the degree of similarity,
in order to be the feature vector n,
is a feature vector n-1, | | B n I and B n- And | represents their modulus, respectively.
6. A matrix factorization based building recommendation system, comprising:
module M1: acquiring a building portrait and a brand scheme, wherein the building portrait and the brand scheme comprise building audience crowd attributes, self tags, advertisement owner releasing preferences, brand rules and historical releasing plans;
module M2: establishing a brand-building scoring matrix according to the building portrait and the brand scheme;
module M3: performing scoring prediction on brand-building through matrix decomposition;
module M4: fusing the scoring matrix and the weight matrix, and weighting the scores of all the brands for the building;
module M5: using the building dense feature vector after matrix decomposition for building similarity calculation, and using the brand dense feature vector after matrix decomposition for brand similarity calculation;
module M6: and recalling the buildings for the brands according to the extracted strong rules, sorting the buildings from high to low according to the scores, and then recommending the buildings.
7. The matrix factorization based building recommendation system of claim 6, wherein the expression of the brand-building scoring matrix is:
wherein: m is the generated scoring matrix; P/T, where P is the number of selected points and T is the total number of points in the building;
establishing a rule matrix according to a brand rule, wherein the expression is as follows:
wherein: w represents the rule weight of the brand, and when the building which does not meet the preset condition needs to be avoided, w is set to be 0.
8. The matrix factorization based building recommendation system of claim 6 wherein the scoring prediction is expressed as:
y(t)＝g(t)+s(t)+h(t)+ε t
wherein: g (t) is a trend term, s (t) is a period term, h (t) is a holiday term, epsilon t And (4) bringing the data into a model for a noise item to generate a predicted value y at the predicted value t.
9. The matrix factorization based building recommendation system of claim 7 wherein the weighting expression in module M4 is:
10. the matrix factorization based building recommendation system of claim 6, wherein the similarity calculation expression is:
wherein: the cos θ is the degree of similarity,
in order to be the feature vector n,
is a feature vector n-1, | | B n I and B n- And | represents their modulus, respectively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210446569.5A CN114881689A (en) | 2022-04-26 | 2022-04-26 | Building recommendation method and system based on matrix decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210446569.5A CN114881689A (en) | 2022-04-26 | 2022-04-26 | Building recommendation method and system based on matrix decomposition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114881689A true CN114881689A (en) | 2022-08-09 |
Family
ID=82672404
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210446569.5A Pending CN114881689A (en) | 2022-04-26 | 2022-04-26 | Building recommendation method and system based on matrix decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114881689A (en) |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133279A (en) * | 2017-04-13 | 2017-09-05 | 西安电子科技大学 | A kind of intelligent recommendation method and system based on cloud computing |
| CN107944898A (en) * | 2016-10-13 | 2018-04-20 | 驰众信息技术（上海）有限公司 | The automatic discovery of advertisement putting building information and sort method |
| CN108090229A (en) * | 2018-01-10 | 2018-05-29 | 广东工业大学 | A kind of method and apparatus that rating matrix is determined based on convolutional neural networks |
| CN108921670A (en) * | 2018-07-04 | 2018-11-30 | 重庆大学 | A kind of potential interest of fusion user, the Drug trading recommended method of space-time data and classification popularity |
| CN109271582A (en) * | 2018-08-20 | 2019-01-25 | 东南大学 | It is a kind of based on the recommendation method for personalized information with attribute member path |
| CN109885719A (en) * | 2019-01-24 | 2019-06-14 | 广州小鹏汽车科技有限公司 | A kind of song recommendations method, system, terminal and storage medium |
| CN110413880A (en) * | 2019-07-10 | 2019-11-05 | 合肥工业大学 | Based on user's personality single classification collaborative filtering method layered |
| CN110851700A (en) * | 2019-09-25 | 2020-02-28 | 浙江工业大学 | Probability matrix decomposition cold start recommendation method integrating attributes and semantics |
| CN110929459A (en) * | 2019-11-28 | 2020-03-27 | 中国大唐集团科学技术研究院有限公司西北电力试验研究院 | Site selection method for wind measuring tower of wind power plant in complex terrain |
| CN111144978A (en) * | 2019-12-10 | 2020-05-12 | 支付宝(杭州)信息技术有限公司 | Fusion recommendation method, system and equipment |
| CN111160859A (en) * | 2019-12-26 | 2020-05-15 | 淮阴工学院 | Human resource post recommendation method based on SVD + + and collaborative filtering |
| WO2020135144A1 (en) * | 2018-12-29 | 2020-07-02 | 北京世纪好未来教育科技有限公司 | Method and apparatus for predicting object preference, and computer-readable medium |
| CN111475744A (en) * | 2020-04-03 | 2020-07-31 | 南京理工大学紫金学院 | Personalized position recommendation method based on ensemble learning |
| CN111667344A (en) * | 2020-06-08 | 2020-09-15 | 中森云链(成都)科技有限责任公司 | Personalized recommendation method integrating comments and scores |
| CN111695024A (en) * | 2019-03-11 | 2020-09-22 | 北京京东尚科信息技术有限公司 | Object evaluation value prediction method and system, and recommendation method and system |
| CN113392920A (en) * | 2021-06-25 | 2021-09-14 | 北京百度网讯科技有限公司 | Method, apparatus, device, medium, and program product for generating cheating prediction model |
| CN113868462A (en) * | 2021-09-13 | 2021-12-31 | 浪潮通信信息系统有限公司 | Song recommendation system and method based on matrix decomposition |
-
2022
- 2022-04-26 CN CN202210446569.5A patent/CN114881689A/en active Pending
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107944898A (en) * | 2016-10-13 | 2018-04-20 | 驰众信息技术（上海）有限公司 | The automatic discovery of advertisement putting building information and sort method |
| CN107133279A (en) * | 2017-04-13 | 2017-09-05 | 西安电子科技大学 | A kind of intelligent recommendation method and system based on cloud computing |
| CN108090229A (en) * | 2018-01-10 | 2018-05-29 | 广东工业大学 | A kind of method and apparatus that rating matrix is determined based on convolutional neural networks |
| CN108921670A (en) * | 2018-07-04 | 2018-11-30 | 重庆大学 | A kind of potential interest of fusion user, the Drug trading recommended method of space-time data and classification popularity |
| CN109271582A (en) * | 2018-08-20 | 2019-01-25 | 东南大学 | It is a kind of based on the recommendation method for personalized information with attribute member path |
| WO2020135144A1 (en) * | 2018-12-29 | 2020-07-02 | 北京世纪好未来教育科技有限公司 | Method and apparatus for predicting object preference, and computer-readable medium |
| CN109885719A (en) * | 2019-01-24 | 2019-06-14 | 广州小鹏汽车科技有限公司 | A kind of song recommendations method, system, terminal and storage medium |
| CN111695024A (en) * | 2019-03-11 | 2020-09-22 | 北京京东尚科信息技术有限公司 | Object evaluation value prediction method and system, and recommendation method and system |
| CN110413880A (en) * | 2019-07-10 | 2019-11-05 | 合肥工业大学 | Based on user's personality single classification collaborative filtering method layered |
| CN110851700A (en) * | 2019-09-25 | 2020-02-28 | 浙江工业大学 | Probability matrix decomposition cold start recommendation method integrating attributes and semantics |
| CN110929459A (en) * | 2019-11-28 | 2020-03-27 | 中国大唐集团科学技术研究院有限公司西北电力试验研究院 | Site selection method for wind measuring tower of wind power plant in complex terrain |
| CN111144978A (en) * | 2019-12-10 | 2020-05-12 | 支付宝(杭州)信息技术有限公司 | Fusion recommendation method, system and equipment |
| CN111160859A (en) * | 2019-12-26 | 2020-05-15 | 淮阴工学院 | Human resource post recommendation method based on SVD + + and collaborative filtering |
| CN111475744A (en) * | 2020-04-03 | 2020-07-31 | 南京理工大学紫金学院 | Personalized position recommendation method based on ensemble learning |
| CN111667344A (en) * | 2020-06-08 | 2020-09-15 | 中森云链(成都)科技有限责任公司 | Personalized recommendation method integrating comments and scores |
| CN113392920A (en) * | 2021-06-25 | 2021-09-14 | 北京百度网讯科技有限公司 | Method, apparatus, device, medium, and program product for generating cheating prediction model |
| CN113868462A (en) * | 2021-09-13 | 2021-12-31 | 浪潮通信信息系统有限公司 | Song recommendation system and method based on matrix decomposition |
Non-Patent Citations (6)
| Title |
|---|
| KENNEY_QIN: "推荐系统之矩阵分解", Retrieved from the Internet <URL:https://blog.csdn.net/OrthocenterChocolate/article/details/38540871> * |
| RAVI, A 等: "Predicting Consumer Level Brand Preferences Using Persistent Mobility Patterns", IEEE INTERNATIONAL CONFERENCE ON BIG DATA (BIG DATA), 31 December 2018 (2018-12-31) * |
| 严素蓉;冯小青;廖一星;: "基于矩阵分解的社会化推荐模型", 清华大学学报(自然科学版), no. 07, 15 July 2016 (2016-07-15) * |
| 孙海威;钟绍波;曹曦文;高云云;李思培;常泉;: "基于矩阵分解的兴趣点推荐算法研究", 软件导刊, no. 09, 21 September 2016 (2016-09-21) * |
| 张博;周瑞瑞;鱼冰;: "协同过滤推荐算法专利综述", 河南科技, no. 19, 5 October 2015 (2015-10-05) * |
| 曹小华 等: "Prophet框架下生产物流运输节点的阻塞预警方法", 武汉理工大学学报（交通科学与工程版）, vol. 44, no. 2, 15 May 2020 (2020-05-15) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111291266A (en) | Artificial intelligence based recommendation method and device, electronic equipment and storage medium | |
| CN103473291B (en) | Personalized service recommendation system and method based on latent semantic probability models | |
| CN111382350B (en) | Multi-task television program recommendation method integrating user click behavior and user interest preference | |
| CN109255586A (en) | A kind of online personalized recommendation method that E-Governance Oriented is handled affairs | |
| CN112700274A (en) | Advertisement click rate estimation method based on user preference | |
| CN114519435A (en) | Model parameter updating method, model parameter updating device and electronic equipment | |
| CN113868537B (en) | Recommendation method based on multi-behavior session graph fusion | |
| Wang et al. | Hybrid customer requirements rating method for customer-oriented product design using QFD | |
| CN116862454A (en) | Indoor building design management method and system | |
| CN115203550A (en) | Social recommendation method and system for enhancing neighbor relation | |
| CN112269943B (en) | Information recommendation system and method | |
| CN110633401A (en) | Prediction model of store data and establishment method thereof | |
| CN114491236A (en) | Intelligent supply chain matching technology based on knowledge learning | |
| CN116629338A (en) | Model data processing method, related device, equipment and storage medium | |
| CN114881689A (en) | Building recommendation method and system based on matrix decomposition | |
| Kumar et al. | Advertising data analysis using rough sets model | |
| CN113128597B (en) | Method and device for extracting user behavior characteristics and classifying and predicting user behavior characteristics | |
| CN113095883A (en) | Video payment user prediction method and system based on deep cross attention network | |
| CN113011921B (en) | Information pushing method and device, electronic equipment and readable storage medium | |
| CN115619238B (en) | Method for establishing inter-enterprise cooperative relationship for non-specific B2B platform | |
| CN114781625B (en) | Network model training and push content determining method and device | |
| CN113688315B (en) | Sequence recommendation method based on no-information-loss graph coding | |
| CN115375364A (en) | Advertisement putting and selecting method and device, equipment and medium thereof | |
| CN117992658A (en) | Automatic architecture searching method for multi-layer perceptron module of recommendation system | |
| CN116976992A (en) | Target object group determination method and device and computer equipment |
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 |