CN108665323A - A kind of integrated approach for finance product commending system - Google Patents
A kind of integrated approach for finance product commending system Download PDFInfo
- Publication number
- CN108665323A CN108665323A CN201810484714.2A CN201810484714A CN108665323A CN 108665323 A CN108665323 A CN 108665323A CN 201810484714 A CN201810484714 A CN 201810484714A CN 108665323 A CN108665323 A CN 108665323A
- Authority
- CN
- China
- Prior art keywords
- user
- article
- preference
- similarity
- indicate
- 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
Links
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
- G06Q30/0251—Targeted advertisements
- G06Q30/0255—Targeted advertisements based on user history
-
- 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
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/06—Asset management; Financial planning or analysis
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Human Resources & Organizations (AREA)
- Operations Research (AREA)
- Technology Law (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention discloses a kind of integrated approach for finance product commending system, and the collaborative filtering based on data smoothing can be filled sparse data, reduce the sparse sex chromosome mosaicism of data.Historical data is not needed based on demographic proposed algorithm, does not depend on the attribute of article yet, can solve the problems, such as the cold start-up of user;Two kinds of algorithms and the good proposed algorithm based on item cluster and matrix decomposition of performance capabilities are integrated, expands the usage scenario of proposed algorithm, improves the adaptivity of proposed algorithm.Integrated approach of the present invention can efficiently reduce the sparsity of data and solve the problems, such as cold start-up, promote the recommendation performance to each user.
Description
Technical field
The invention belongs to online Products Show technical field more particularly to a kind of integrating for finance product commending system
Method.
Background technology
Traditional proposed algorithm mostly calculates the interest preference and resource similarity of user with user's score data, to sparse
Data and the recommendation quality of new user are relatively low, can not maximize the information for excavating recessive data institute band.
In recent years, the problem of being brought for Deta sparseness, in order to promote recommendation effect, scholars by principal component analysis,
Clustering, singular value decomposition scheduling algorithm are introduced into traditional Collaborative Filtering Recommendation Algorithm, by dimensionality reduction, reduce target user
Search for the range of nearest-neighbors so that the precision and real-time of recommendation, which have, to be obviously improved, but cold start-up problem still remains.
The primary structure and principle of the prior art:
1. generating user-item attribute preference pattern.User to the preference pattern of item attribute be carry out user clustering and
User is established to going out in project in the basis of similarity calculation by analyzing user-project rating matrix and project-attribute matrix
The preference weight matrix of existing all properties.
2. user clustering.User is clustered with the mixing Clustering Model that K-means clusters are combined using SOM:
2.1 train input using obtained user-item attribute preference matrix as the input data of cluster, by SOM
Less number is slightly clustered, output clustering cluster ClusterSOM, neuron weights ωSOM, clustering cluster number K;
2.2 by ωSOMAs original barycenter Ooriginal, for each cluster interior element be 0 clustering cluster, find with
OoriginalThe barycenter O final as the cluster apart from nearest elementSOM;
2.3 with K, OSOMAs the clustering cluster number and initial clustering barycenter of K-means clusters, user is further clustered,
Export user clustering result ClusterResult.
3. user's similarity calculation and Nearest neighbor queries.Target user U is calculated using cosineiWith place clustering cluster
cindexNearest-neighbors set M is calculated in the similarity of middle other usersKnear。
4. score in predicting.Find target user UiFor destination item IijNearest neighbor set MKnearAfterwards, pass through collection
Close MKnearIn user to destination item IijThe weighted average of scoring describes target user UiTo destination item IijComment
Point.Shown in score in predicting formula such as formula (1):
Recommend 5. generating.Repeat step (3) and step (4), prediction target user UiScoring to all non-scoring items,
Target user U is given in the highest N number of project recommendation of selection prediction scoringi。
There are many proposed algorithms at present, but all there is no a kind of algorithms to be always better than under any background or any data
Other proposed algorithms.Existing commending system is mostly a kind of single method cannot neatly apply with the limitation of itself
In all kinds of scenes.Good algorithm is showed in terms of recommendation cannot efficiently solve the sparsity and cold start-up problem of data.
Invention content
The present invention provides a kind of integrated approach for finance product commending system, reduces the shadow that Deta sparseness is brought
It rings, solves the problems, such as the cold start-up of commending system.The adaptivity for improving proposed algorithm, expands the applicable scene of proposed algorithm.
Collaborative filtering based on data smoothing can be filled sparse data, and the sparsity for reducing data is asked
Topic.Historical data is not needed based on demographic proposed algorithm, does not depend on the attribute of article yet, the cold of user can be solved
Starting problem.Two kinds of algorithms and the good proposed algorithm based on item cluster and matrix decomposition of performance capabilities are integrated,
The usage scenario for expanding proposed algorithm, improves the adaptivity of proposed algorithm.Integrated approach of the present invention can efficiently reduce number
According to sparsity and solves the problems, such as cold start-up, recommendation performance of the promotion to each user.
Description of the drawings
Fig. 1 is the integrated approach flow chart that the present invention is used for finance product commending system.
Specific implementation mode
As shown in Figure 1, the present invention provides a kind of integrated approach for finance product commending system, include the following steps:
Input data is:User's characteristic information, project-attribute matrix, user-project rating matrix;Output data is:Production
Product recommended models, user's recommendation results.
Step 1:Based on demographic proposed algorithm.
It is improved based on demographic proposed algorithm to traditional, different power is assigned for different user properties
Value.The present invention chooses age, gender, occupation and 4 kinds of features of hobby as the range considered, and each attribute information is pre-processed into number
The form of font representation.The similarity calculated between user obtains user preference.
1. age attribute, the present invention is 5 years old to be increment, for example 32 years old age can be denoted as 7, and 56 years old age was denoted as 12.It utilizes
Euclidean distance calculates the similarity at age between user s and user t, as shown in formula (2).
Wherein A (s, t) indicates user s and the similarities of user t in years, xs、xtThe respectively age of user s and t point
Segment value.
2. gender is a kind of symmetrical double attributes, i.e., two states are all of equal importance.Male is denoted as 1, Nv Xingji
It is 0, the list of two rows two row can be obtained according to the value of user's gender in this way, as shown in table 1.Here S (s, t) is used
Indicate user s and similarities of the user t in gender, as shown in formula (3), wherein a is the attribute that object s and t take 1 situation
Value, b are the attribute values that object t takes 1 and object s takes 0 situation, and c is that 0 and object s is taken to take the attribute value of 1 situation, d to be in object t
Object s and t take the attribute value of 0 situation.
1 double attributes value of table
3. occupation, hobby belong to the attributes of tag types, i.e., this attribute is described with certain fields, to this generic attribute,
The similarity between user s and t is calculated using formula (4).K indicates user s and the jointly owned same attributes of t in formula (4)
Label number, n indicate the alternative label number of the attribute.
4. the calculating of user's similarity.To user u and its k neighborhood Uk, formula (5) calculate user information it is similar
Degree, and preserve Top-N user set Us similar with target user uk.Under user's Demographic's method, feature is got over
It is more, it more can Accurate Prediction user interest.
Calculate user preference
After obtaining the similarity between user, the article liked with K most like user of his interest can be recommended to user,
Preferences of the user u to article i is calculated using following formula:
Wherein, S (u, k) includes and the most similar K user of user u, N (i) are the user's collection for having behavior to article i
It closes, wuvIt is the Interest Similarity of user u and user v, rviIndicate preferences of the user v to article i.
Obtain prediction preference matrix
N is the quantity of article, and m is the quantity of user.
Step 2:Proposed algorithm based on item cluster and matrix decomposition
1. calculating the similarity between article
The distance between article is calculated by using manhatton distance.
ruiIndicate favorable ratings of the user u to article i.dijIndicate the distance between article i and article j.
Similarity between article i and article j is expressed as formula (9).
ciIndicate the popularity of article i, cjIndicate the popularity of article j.The popularity of article is exactly the people for clicking the article
Number.
Then article is classified, obtains different cluster centre { c1,c2,Λ,ck, k is the number of cluster.
2. building article vector.
Based on k cluster centre, K is set as 200, and article vector is defined as
Wherein,
Article vector is normalized:
Finally, the vector of article i is:
pi=(pi1,pi2,Λ,pik,Λ,piK) (13)
Wherein,
3. calculating prediction preference matrix.Based on article vector sum singular value decomposition (SVD), prediction preference square can be obtained
Battle array:
N is the quantity of article, and m is the quantity of user,It indicates prediction preference, is defined as:
Wherein, allMean is the average value of preference, buIndicate the deviation between user and allMean, biIndicate article and
Deviation between allMean, quIt is the vector of user u, is initialized by random value.
Step 3:Collaborative filtering based on data smoothing.
1. calculating the similarity of user.Similarity is calculated using Pearson correlation coefficient.
Similarity between user u and user u' is:
Ru(t) preferences of the user u to article t is indicated,Indicate user u to the average preference of all items, Ru’(t) it indicates
User u ' to the preference of article t,User u ' is indicated to the average preference of all items, t is that user u and user u ' were clicked
Article.
User is defined as U={ u1,u2,Λ,un, user is divided into n cluster, is expressed as
2. being based on previous step, the data set that smooth user not yet clicks.The preference of user is expressed as:
Wherein ruiIt is to be calculated by function,It is the i that was not clicked for user u by smoothly obtaining.
User u, the u cluster belonged to are expressed asIn view of individual difference, lead to
Formula (19) is crossed to calculate
It is average preference of all users to article i, is calculated as follows:
Wherein, Cu(i)∈CuIt indicates in cluster CuIn click article i user collection, | Cu(i) | it indicates in cluster CuIn
Click article i number of users.
Prediction preference can be obtained by calculating weighted sum:
It is preferences of the user u to article i,It is the average preference of article i,It is the average preference of article j, wuj
It is the weight between u and j, sim (i, j) is the similarity of i and j.
3. obtaining prediction preference matrix.
N is the quantity of article, and m is the quantity of user.
Step 4:Algorithm is integrated
According to Step 1: two and three, prediction preference of the user to each product has been obtained, these preferences has been based on, is calculated
Method is integrated.
1. linear weighted function fusion method
One model of cover sheet as a result, different weights is then assigned by algorithms of different, by the result of multiple proposed algorithms
It is weighted, you can obtain result:
WhereinIt is the final preference of prediction, wkIt is weight corresponding with k algorithms.
2. mixing together method
In recommendation results, intert different recommended models as a result, to ensure the diversity of result.
Rec (u) indicates the article recommended user u, reck(u) article that algorithm k recommends user u is indicated.
3. waterfall fusion method
Waterfall type fusion method is used the concatenated method of multiple models.Each proposed algorithm is considered as a filtering
Device, by the way that the successive method of varigrained filter is carried out, in the method, previous recommendation method filtering
As a result, the candidate collection of method will be recommended to input as the latter, progressive, candidate result can be selected for a post gradually in the process
Choosing finally obtains the high recommendation results set of the few matter of an amount.
Claims (3)
1. a kind of integrated approach for finance product commending system, which is characterized in that include the following steps:
Step 1:Based on demographic proposed algorithm.
4 kinds of age, gender, occupation and hobby features are chosen, by the pretreatment of each attribute information at the form of numeric type representation, meter
The similarity between user is calculated to obtain user preference and obtain prediction preference matrix;
Step 2:Proposed algorithm based on item cluster and matrix decomposition
Similarity between step 2.1, calculating article
The distance between article is calculated by using manhatton distance.
Wherein, ruiIndicate user u to the favorable rating of article i, dijIndicate the distance between article i and article j,
Similarity between article i and article j is expressed as formula (9).
Wherein, ciIndicate the popularity of article i, cjIndicate the popularity of article j.The popularity of article is exactly to click the article
Number,
Then article is classified, obtains different cluster centre { c1,c2,…,ck, k is the number of cluster.
Step 2.2, structure article vector
Based on k cluster centre, K is set as 200, and article vector is defined as
Wherein,
Article vector is normalized:
Finally, the vector of article i is:
pi=(pi1,pi2,…,pik,…,piK) (13)
Wherein,
Step 2.3 calculates prediction preference matrix
Based on article vector sum singular value decomposition (SVD), prediction preference matrix can be obtained:
Wherein, n is the quantity of article, and m is the quantity of user,It indicates prediction preference, is defined as:
Wherein, allMean is the average value of preference, buIndicate the deviation between user and allMean, biIndicate article and
Deviation between allMean, quIt is the vector of user u, is initialized by random value;
Step 3:Collaborative filtering based on data smoothing
Step 3.1, the similarity for calculating user
Similarity is calculated using Pearson correlation coefficient, the similarity between user u and user u ' is:
Wherein, Ru(t) preferences of the user u to article t is indicated,Indicate user u to the average preference of all items, Ru, (t) table
Show user u, to the preference of article t,Indicate user u, to the average preference of all items, t is user u and user u, is all clicked
The article crossed,
User is defined as U={ u1,u2,…,un, user is divided into n cluster, is expressed as
Step 3.2 is based on previous step, the data set that smooth user not yet clicks
The preference of user is expressed as:
Wherein, ruiIt is to be calculated by function,It is the i that was not clicked for user u by smoothly obtaining,
User u, the u cluster belonged to are expressed asIn view of individual difference, pass through formula
(19) it calculates
It is average preference of all users to article i, is calculated as follows:
Wherein, Cu(i)∈CuIt indicates in cluster CuIn click article i user collection, | Cu(i) | it indicates in cluster CuIn point
The number of users of article i was hit,
Prediction preference can be obtained by calculating weighted sum:
Wherein,It is preferences of the user u to article i,It is the average preference of article i,It is the average preference of article j, wuj
It is the weight between u and j, sim (i, j) is the similarity of i and j.
Step 3.3 obtains prediction preference matrix
N is the quantity of article, and m is the quantity of user.
Step 4:Algorithm is integrated
According to Step 1: two and three, prediction preference of the user to each product has been obtained, has been based on these preferences, has carried out set of algorithms
At.
2. being used for the integrated approach of finance product commending system as described in claim 1, which is characterized in that in step 1, user
The calculating process of preference is as follows:
After obtaining the similarity between user, the article liked with K most like user of his interest can be recommended to user, used
Following formula calculates preferences of the user u to article i:
Wherein, S (u, k) includes and the most similar K user of user u, N (i) are the user's set for having behavior to article i, wuv
It is the Interest Similarity of user u and user v, rviIndicate preferences of the user v to article i.
Predict that the calculating process of preference matrix is as follows:
Wherein, n is the quantity of article, and m is the quantity of user.
3. being used for the integrated approach of finance product commending system as described in claim 1, which is characterized in that step 4 is using such as
Lower method is integrated:
1, linear weighted function fusion method
One model of cover sheet as a result, then assign different weights by algorithms of different, the result of multiple proposed algorithms is carried out
Weighting, you can obtain result:
Wherein,It is the final preference of prediction, wkIt is weight corresponding with k algorithms.
2, mixing together method
In recommendation results, intert different recommended models as a result, to ensure the diversity of result,
Wherein, rec (u) indicates the article recommended user u, reck(u) article that algorithm k recommends user u is indicated.
3. waterfall fusion method
Waterfall type fusion method is used the concatenated method of multiple models, and each proposed algorithm is considered as a filter, is led to
Cross and carry out the successive method of varigrained filter, in the method, the filtering of previous recommendation method as a result,
The candidate collection of method will be recommended to input as the latter, progressive, candidate result can be selected gradually in the process, most
The high recommendation results set of the few matter of an amount is obtained eventually.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810484714.2A CN108665323B (en) | 2018-05-20 | 2018-05-20 | Integration method for financial product recommendation system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810484714.2A CN108665323B (en) | 2018-05-20 | 2018-05-20 | Integration method for financial product recommendation system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108665323A true CN108665323A (en) | 2018-10-16 |
CN108665323B CN108665323B (en) | 2021-01-05 |
Family
ID=63776873
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810484714.2A Active CN108665323B (en) | 2018-05-20 | 2018-05-20 | Integration method for financial product recommendation system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108665323B (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109544231A (en) * | 2018-11-21 | 2019-03-29 | 浙江工商大学 | Logistics distribution service personalization recommended method based on Logistics Information Platform |
CN109902235A (en) * | 2019-03-06 | 2019-06-18 | 太原理工大学 | User preference based on bat optimization clusters Collaborative Filtering Recommendation Algorithm |
CN109977315A (en) * | 2019-03-29 | 2019-07-05 | 厦门铠甲网络股份有限公司 | A kind of article recommended method, device, equipment and storage medium |
CN110930226A (en) * | 2019-11-26 | 2020-03-27 | 中国建设银行股份有限公司 | Financial product recommendation method and device, electronic equipment and storage medium |
CN110930259A (en) * | 2019-11-15 | 2020-03-27 | 安徽海汇金融投资集团有限公司 | Creditor right recommendation method and system based on mixed strategy |
CN111179037A (en) * | 2019-12-31 | 2020-05-19 | 中国银行股份有限公司 | Financial product recommendation method and device |
CN111552852A (en) * | 2020-04-27 | 2020-08-18 | 北京交通大学 | Article recommendation method based on semi-discrete matrix decomposition |
CN111652669A (en) * | 2020-04-15 | 2020-09-11 | 珠海小礼鱼科技有限公司 | Intelligent terminal shopping guide system and method |
CN111966907A (en) * | 2020-08-21 | 2020-11-20 | 贝壳技术有限公司 | User preference cold start method, device, medium and electronic equipment |
CN112241894A (en) * | 2019-07-16 | 2021-01-19 | 百度时代网络技术(北京)有限公司 | Content delivery method and device and terminal |
CN113379482A (en) * | 2021-05-28 | 2021-09-10 | 车智互联(北京)科技有限公司 | Item recommendation method, computing device and storage medium |
CN113393297A (en) * | 2021-06-17 | 2021-09-14 | 中国工商银行股份有限公司 | Financial product pushing method and device |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2011096255A (en) * | 2009-10-30 | 2011-05-12 | Nec (China) Co Ltd | Ranking oriented cooperative filtering recommendation method and device |
CN102968506A (en) * | 2012-12-14 | 2013-03-13 | 北京理工大学 | Personalized collaborative filtering recommendation method based on extension characteristic vectors |
CN103246672A (en) * | 2012-02-09 | 2013-08-14 | 中国科学技术大学 | Method and device for performing personalized recommendation on users |
CN106846106A (en) * | 2017-01-16 | 2017-06-13 | 南京师范大学 | A kind of figure builds the matrix decomposition recommendation method under framework |
CN106980648A (en) * | 2017-02-27 | 2017-07-25 | 南京邮电大学 | It is a kind of that the personalized recommendation method for combining similarity is decomposed based on probability matrix |
CN106997347A (en) * | 2016-01-22 | 2017-08-01 | 华为技术有限公司 | Information recommendation method and server |
-
2018
- 2018-05-20 CN CN201810484714.2A patent/CN108665323B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2011096255A (en) * | 2009-10-30 | 2011-05-12 | Nec (China) Co Ltd | Ranking oriented cooperative filtering recommendation method and device |
CN103246672A (en) * | 2012-02-09 | 2013-08-14 | 中国科学技术大学 | Method and device for performing personalized recommendation on users |
CN102968506A (en) * | 2012-12-14 | 2013-03-13 | 北京理工大学 | Personalized collaborative filtering recommendation method based on extension characteristic vectors |
CN106997347A (en) * | 2016-01-22 | 2017-08-01 | 华为技术有限公司 | Information recommendation method and server |
CN106846106A (en) * | 2017-01-16 | 2017-06-13 | 南京师范大学 | A kind of figure builds the matrix decomposition recommendation method under framework |
CN106980648A (en) * | 2017-02-27 | 2017-07-25 | 南京邮电大学 | It is a kind of that the personalized recommendation method for combining similarity is decomposed based on probability matrix |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109544231A (en) * | 2018-11-21 | 2019-03-29 | 浙江工商大学 | Logistics distribution service personalization recommended method based on Logistics Information Platform |
CN109902235A (en) * | 2019-03-06 | 2019-06-18 | 太原理工大学 | User preference based on bat optimization clusters Collaborative Filtering Recommendation Algorithm |
CN109902235B (en) * | 2019-03-06 | 2023-07-07 | 太原理工大学 | User preference clustering collaborative filtering recommendation algorithm based on bat optimization |
CN109977315A (en) * | 2019-03-29 | 2019-07-05 | 厦门铠甲网络股份有限公司 | A kind of article recommended method, device, equipment and storage medium |
CN112241894A (en) * | 2019-07-16 | 2021-01-19 | 百度时代网络技术(北京)有限公司 | Content delivery method and device and terminal |
CN110930259B (en) * | 2019-11-15 | 2023-05-26 | 安徽海汇金融投资集团有限公司 | Credited recommendation method and system based on mixed strategy |
CN110930259A (en) * | 2019-11-15 | 2020-03-27 | 安徽海汇金融投资集团有限公司 | Creditor right recommendation method and system based on mixed strategy |
CN110930226A (en) * | 2019-11-26 | 2020-03-27 | 中国建设银行股份有限公司 | Financial product recommendation method and device, electronic equipment and storage medium |
CN111179037A (en) * | 2019-12-31 | 2020-05-19 | 中国银行股份有限公司 | Financial product recommendation method and device |
CN111179037B (en) * | 2019-12-31 | 2023-07-25 | 中国银行股份有限公司 | Financial product recommendation method and device |
CN111652669A (en) * | 2020-04-15 | 2020-09-11 | 珠海小礼鱼科技有限公司 | Intelligent terminal shopping guide system and method |
CN111552852A (en) * | 2020-04-27 | 2020-08-18 | 北京交通大学 | Article recommendation method based on semi-discrete matrix decomposition |
CN111966907A (en) * | 2020-08-21 | 2020-11-20 | 贝壳技术有限公司 | User preference cold start method, device, medium and electronic equipment |
CN113379482A (en) * | 2021-05-28 | 2021-09-10 | 车智互联(北京)科技有限公司 | Item recommendation method, computing device and storage medium |
CN113379482B (en) * | 2021-05-28 | 2023-12-01 | 车智互联(北京)科技有限公司 | Article recommendation method, computing device and storage medium |
CN113393297A (en) * | 2021-06-17 | 2021-09-14 | 中国工商银行股份有限公司 | Financial product pushing method and device |
Also Published As
Publication number | Publication date |
---|---|
CN108665323B (en) | 2021-01-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108665323A (en) | A kind of integrated approach for finance product commending system | |
Zarzour et al. | A new collaborative filtering recommendation algorithm based on dimensionality reduction and clustering techniques | |
Wu et al. | Robust latent factor analysis for precise representation of high-dimensional and sparse data | |
CN110162693B (en) | Information recommendation method and server | |
Lyu et al. | Deep match to rank model for personalized click-through rate prediction | |
CN107833117B (en) | Bayesian personalized sorting recommendation method considering tag information | |
CN112232925A (en) | Method for carrying out personalized recommendation on commodities by fusing knowledge maps | |
CN107894998B (en) | Video recommendation method and device | |
Vargas et al. | Exploiting the diversity of user preferences for recommendation | |
CN108363804A (en) | Partial model Weighted Fusion Top-N films based on user clustering recommend method | |
CN110020128B (en) | Search result ordering method and device | |
CN109241366B (en) | Hybrid recommendation system and method based on multitask deep learning | |
CN107220365A (en) | Accurate commending system and method based on collaborative filtering and correlation rule parallel processing | |
CN106777051A (en) | A kind of many feedback collaborative filtering recommending methods based on user's group | |
CN107545471B (en) | Big data intelligent recommendation method based on Gaussian mixture | |
CN101454771A (en) | System and method of segmenting and tagging entities based on profile matching using a multi-media survey | |
CN108509573A (en) | Book recommendation method based on matrix decomposition collaborative filtering and system | |
CN106127506B (en) | recommendation method for solving cold start problem of commodity based on active learning | |
Rokach et al. | Initial profile generation in recommender systems using pairwise comparison | |
CN106951471A (en) | A kind of construction method of the label prediction of the development trend model based on SVM | |
CN106777359B (en) | A kind of text services recommended method based on limited Boltzmann machine | |
CN109903138A (en) | A kind of individual commodity recommendation method | |
CN108389113B (en) | Collaborative filtering recommendation method and system | |
CN109977299A (en) | A kind of proposed algorithm of convergence project temperature and expert's coefficient | |
CN110334278A (en) | A kind of web services recommended method based on improvement deep learning |
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 | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20210617 Address after: 200335 Room 301, building 1, 111 Fuquan Road, Changning District, Shanghai Patentee after: SHANGHAI DIGITAL CHINA INFORMATION TECHNOLOGY SERVICE Co.,Ltd. Address before: 100124 No. 100 Chaoyang District Ping Tian Park, Beijing Patentee before: Beijing University of Technology |