CN109241442B - Project recommendation method based on predictive value filling, readable storage medium and terminal - Google Patents
Project recommendation method based on predictive value filling, readable storage medium and terminal Download PDFInfo
- Publication number
- CN109241442B CN109241442B CN201811181074.4A CN201811181074A CN109241442B CN 109241442 B CN109241442 B CN 109241442B CN 201811181074 A CN201811181074 A CN 201811181074A CN 109241442 B CN109241442 B CN 109241442B
- Authority
- CN
- China
- Prior art keywords
- matrix
- user
- neighbor
- filling
- trust
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 26
- 239000011159 matrix material Substances 0.000 claims abstract description 201
- 238000000354 decomposition reaction Methods 0.000 claims description 29
- 238000004364 calculation method Methods 0.000 description 7
- 238000001914 filtration Methods 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012546 transfer Methods 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
- 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
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Data Mining & Analysis (AREA)
- Game Theory and Decision Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
A project recommendation method based on predictive value filling, a readable storage medium and a terminal, the method comprising: acquiring an original user project scoring matrix; when the original user item scoring matrix is determined to be a sparse matrix, filling the original user item scoring matrix based on the similarity among the items to obtain a corresponding filling matrix; predicting the project scores of the target users based on the filling matrix; and recommending the corresponding item to the target user according to the predicted item score. By means of the scheme, the accuracy of project recommendation can be improved, and the user experience is improved.
Description
Technical Field
The invention belongs to the technical field of internet, and particularly relates to a project recommendation method based on predictive value filling, a readable storage medium and a terminal.
Background
With the rapid development of the internet, network information grows explosively, so that users cannot screen valuable information from huge data information, and the problem of information overload needs to be solved urgently. In this context, personalized recommendation systems have come to mind. Specifically, the recommendation system recommends an item meeting the needs of a user for the user through an effective algorithm based on historical behavior data or item data of the user. Recommendation system technology is currently used to varying degrees in the fields of e-commerce and social networking.
A Collaborative Filtering (CF) algorithm is one of the most widely used recommendation algorithms, and measures the interest and preference of a user according to the degree of the score of the user on an item, and recommends according to the score. Collaborative filtering recommendation algorithms are mainly classified into model-based and neighborhood-based.
Disclosure of Invention
The technical problem to be solved by the invention is how to improve the accuracy of project recommendation and improve the experience of users.
In order to achieve the above object, the present invention provides a project recommendation method based on predictive value filling, the method comprising:
acquiring an original user project scoring matrix;
when the original user item scoring matrix is determined to be a sparse matrix, filling the original user item scoring matrix based on the similarity among the items to obtain a corresponding filling matrix;
predicting the project scores of the target users based on the filling matrix;
and recommending the corresponding item to the target user according to the predicted item score.
Optionally, the populating the original user item rating matrix based on the similarity between items includes:
calculating the similarity between items in the original user item scoring matrix;
determining a plurality of scoring items with the highest similarity to the unscored items in the scoring matrix of the original user items based on the similarity between the calculated items;
calculating the prediction scores of the unscored items by adopting the scores of a plurality of scored items with the highest similarity to the unscored items in the original user item scoring matrix;
filling the prediction scores of the unscored items into corresponding positions in the original user item score matrix.
Optionally, the predicting the item score of the target user based on the filling matrix includes:
reducing the dimension of the filling matrix by singular value decomposition to obtain a corresponding left singular matrix and a corresponding singular value matrix;
constructing a user implicit characteristic space matrix based on the left singular matrix and the singular value matrix;
determining a neighbor user group of each user based on the user implicit feature space matrix of the user;
building a trust neighbor group set of a target user based on the trust relationship among the users in the neighbor user group;
and predicting the scoring items of the target user based on the constructed item scores of the users in the trust neighbor cluster of the target user.
Optionally, the following formula is adopted to perform dimension reduction on the filling matrix by using singular value decomposition:
wherein,is represented by P|U|×kRepresenting a left singular matrix obtained by dimensionality reduction of the filling matrix by singular value decomposition, Sk×kRepresenting a singular value matrix obtained by dimension reduction of the filling matrix by singular value decomposition,and D, performing dimensionality reduction on the filling matrix by adopting singular value decomposition to obtain a transposed matrix of a right singular matrix, wherein U represents the number of items, and k represents the number of users subjected to dimensionality reduction.
Optionally, a user implicit feature space matrix is constructed based on the left singular matrix and the singular value matrix by using the following formula:
Y=PKSK 1/2;
wherein Y represents the user implicit feature space matrix, PKRepresenting a left singular matrix obtained by dimensionality reduction of the filling matrix by singular value decomposition, SKAnd representing a singular value matrix obtained by dimension reduction of the filling matrix by singular value decomposition.
Optionally, the determining a neighbor user group of each user based on the user implicit feature space matrix includes:
calculating the similarity between every two users under the user implicit characteristic space matrix, and constructing a user similarity matrix;
converting the user similarity matrix into a corresponding fuzzy equivalent matrix;
and classifying users with similar characteristics into one class according to the equivalent relation value in the fuzzy equivalent matrix to obtain a neighbor user group corresponding to each user.
Optionally, the constructing a trust neighborhood group set of trust users based on trust relationships among users in the neighborhood user group includes:
when a target user has a direct trust relationship with a neighbor user in the trust neighbor group set, calculating a trust degree value between the target user and the neighbor user by adopting modified cosine similarity;
when a target user and a neighbor user in the trust neighbor group set have an indirect trust relationship, calculating a trust degree value between the target user and the neighbor user by adopting a trust propagation algorithm;
and acquiring a plurality of neighbor users with higher trust degree values with the target user to form a trust neighbor group set of the target user.
Optionally, the following formula is adopted to predict the scoring item of the target user based on the item score of the users in the constructed trust neighbor cluster of the target user:
wherein p isa,iRepresenting a target user uaThe predicted score of the non-scored item i of (a),representing a target user uaOf the neighbor users in the set of trusted neighbor clusters, rk,iNeighbor user k pairs in a trusted neighbor set representing a target userMean of the scores of item i, gtkDirect connection to user u in a trust network representing a set of trusted neighbors of a target useraIs the average of the local trust scores of the neighbor users of, B (u)a) Representing a target user uaTo trust neighbor users in the neighbor set.
The embodiment of the present invention further provides a computer-readable storage medium, on which computer instructions are stored, and when the computer instructions are executed, the method for recommending items based on filling of predicted values according to any of the above-mentioned steps is performed.
The embodiment of the invention also provides a terminal, which comprises a memory and a processor, wherein the memory is stored with computer instructions capable of being executed on the processor, and the processor executes the steps of any item recommendation method based on the filling of the predicted values when executing the computer instructions.
Compared with the prior art, the invention has the beneficial effects that:
according to the scheme, when the original user item scoring matrix is determined to be the sparse matrix, the original user item scoring matrix is filled based on the similarity among the items to obtain the corresponding filling matrix, item scoring of the target user is predicted based on the filling matrix, and finally the corresponding item is recommended to the target user according to the predicted item scoring, so that the problem of data sparseness can be effectively solved, and the accuracy of item recommendation is improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive labor.
FIG. 1 is a flowchart illustrating a project recommendation method based on predictive value filling according to an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of an item recommendation apparatus based on predictive value filling according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application. The directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship between the components, the movement, etc. in a specific posture (as shown in the drawings), and if the specific posture is changed, the directional indication is changed accordingly.
As described in the background, the collaborative filtering algorithm in the prior art is one of the most widely used recommendation algorithms, which measures the interest and preference of a user according to the rating of the user for an item, and makes a recommendation according to the rating. Collaborative filtering recommendation algorithms are mainly classified into model-based and neighborhood-based.
The collaborative filtering based on the model mainly trains out a corresponding model through the grading information of the product by a user, and then the model is used for predicting unknown data. Although the method has good expansibility and practicability and is widely used, the problem of sparsity of a user-scoring matrix is faced along with the rapid increase of the number of users and items, so that the recommendation quality is not high.
According to the technical scheme, when the original user item scoring matrix is determined to be the sparse matrix, the original user item scoring matrix is filled based on the similarity among the items to obtain the corresponding filling matrix, item scoring of the target user is predicted based on the filling matrix, and finally the corresponding item is recommended to the target user according to the predicted item scoring, so that the problem of data sparsity can be effectively solved, and the accuracy of item recommendation is improved.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
Fig. 1 is a flowchart illustrating a project recommendation method based on predictive value filling according to an embodiment of the present invention. Referring to fig. 1, a project recommendation method based on predictive value filling may specifically include the following steps:
step S101: and acquiring an original user item scoring matrix.
In a specific implementation, the original user item scoring matrix includes scoring information of a plurality of users and a plurality of corresponding items.
Step S102: and when the original user item scoring matrix is determined to be a sparse matrix, filling the original user item scoring matrix based on the similarity among the items to obtain a corresponding filling matrix.
In a specific implementation, when an original user item scoring matrix is determined to be a sparse matrix, if scoring prediction is performed on an unscored item of a target user under the sparse matrix, the problem of inaccurate recommendation caused by data sparseness occurs. Thus, to avoid data sparseness, the original user item scoring matrix may be populated first.
In a specific implementation, the scoring of the unscored items by the user may be predicted based on the similarity between the items in the original user item scoring matrix, so as to populate the original user item scoring matrix, which may specifically include the following operations:
first, the similarity between items in the original user item scoring matrix is calculated. In an embodiment of the present invention, the similarity between items in the original user item scoring matrix is calculated by using the following formula:
wherein sim (a, b) represents the item a and item b in the original user item scoring matrixSimilarity between, Ra,iRepresents the user a's score for item i, Rb,iIndicating the rating of the item i by the user b,represents the mean of the item scores for user a,represents the mean of the item scores for user b,represents the mean of the scores for item j.
After the similarity between the projects is obtained through calculation, determining a plurality of scoring projects with the highest similarity with the unscored projects in the original user project scoring matrix based on the similarity between the projects obtained through calculation, and calculating the prediction scores of the unscored projects by adopting the scores of the scoring projects with the highest similarity with the unscored projects in the original user project scoring matrix. In other words, the plurality of items S with the highest similarity to the unscored items will be searched in the whole item spaceP={I1,I2,..,ImAnd predicting the scoring items by adopting the user scores of the items with the highest similarity with the unscored items. In one embodiment of the present invention, the unscored items are predicted using the following formula:
wherein p isa,pRepresents the score prediction value, sim, of user a for the unscored item pp,cRepresenting the similarity of the unscored items p to the items c, Ra,cWhich represents the rating of the user a on item c, which is one of the items with the highest similarity to the unscored item p.
And repeating the steps, calculating the unscored items of the user in the original user item scoring matrix to obtain corresponding predicted scores, and filling the corresponding predicted scores into the corresponding positions in the original user item scoring matrix until all the items with the scores of zero in the original user item scoring matrix are filled, thereby obtaining a corresponding filling matrix.
Step S103: and predicting the project scores of the target users based on the filling matrix.
In an embodiment of the present invention, in order to reduce the complexity of the calculation, singular value decomposition (specifically, Truncated singular value decomposition, abbreviated as TSVD) may be first adopted to decompose and reduce the dimension of the padding matrix:
wherein,representing the original user rating matrix, P|U|×kRepresenting a left singular matrix obtained by dimensionality reduction of the filling matrix by singular value decomposition, Sk×kRepresenting a singular value matrix obtained by dimension reduction of the filling matrix by singular value decomposition,and D, performing dimensionality reduction on the filling matrix by adopting singular value decomposition to obtain a transposed matrix of a right singular matrix, wherein U represents the number of items in the filling matrix, and k represents the number of users subjected to dimensionality reduction.
Secondly, after the filling matrix is decomposed and dimensionality reduced to obtain a left singular matrix, a right singular matrix and a singular value matrix which correspond to each other, the left singular matrix and the singular value matrix which are obtained by decomposition and dimensionality reduction can be adopted to obtain a corresponding user implicit characteristic space matrix. In an embodiment of the present invention, the user implicit feature space matrix may be obtained by using the following formula:
Y=PKSK 1/2 (4)
wherein, Y tableIndicating said user implicit feature space matrix, PKRepresenting a left singular matrix obtained by dimensionality reduction of the filling matrix by singular value decomposition, SKAnd representing a singular value matrix obtained by dimension reduction of the filling matrix by singular value decomposition.
The user implicit characteristic space matrix obtained after the processing of the steps calculates the vector similarity for an m multiplied by k matrix (k-dimensional representation of m users) to form a neighborhood in the reduced space, so that the beneficial effects of reducing the calculation complexity, saving the calculation resources and improving the calculation efficiency can be achieved.
Then, when the user implicit feature space matrix is obtained, the neighbor user group of each user can be determined in the user implicit feature space. In an embodiment of the present invention, the following operations may be specifically adopted to determine the neighbor user group of each user:
(1) firstly, calculating the similarity between every two users under the user implicit characteristic space matrix, and constructing a user similarity matrix. Specifically, firstly, the similarity between every two users is calculated under the user implicit characteristic space matrix, and the calculated similarity of the users is filled in the corresponding position in the matrix, so that the user similarity matrix can be obtainedWhen calculating the similarity between every two users, the similarity between every two users can be calculated by using formula (1), and the similarity is only required to be converted into the corresponding user parameter, which is not described herein again.
(2) And converting the user similarity matrix into a corresponding fuzzy equivalent matrix. Specifically, the user similarity matrix can be converted into a fuzzy equivalent matrix by using the transfer closure, that is, a natural number K is found by using a flat method, so thatThereinRepresenting a fuzzy matrix multiplication operation.
(3) And classifying users with similar characteristics into one class according to the equivalent relation value in the fuzzy equivalent matrix to obtain a neighbor user group corresponding to each user. The fuzzy clustering algorithm is used for searching neighbor clients of a target client, the clients can be subjected to soft classification, one user is allowed to belong to different classes according to different probabilities, and the users of each class have similar preferences.
Then, based on the Trust relationship between the users in the neighbor user group, building a Trust neighbor group set of the target user, namely, in the user neighbor group of the target user, fusing the Trust factors into the similarity space, calculating the Trust between the target user and the neighbor users in the user neighbor group, and selecting a certain number of neighbor users with higher Trust to build the Trust neighbor group set Trust of the target usern. This is also an important step, since the data matrix is mostly sparse, the trusted neighbors of the high-confidence neighbors can be considered to be trustworthy, and this also conforms to the principle of trust network delivery.
Specifically, first, user u is definediI.e. directly connected to user u in the trust networkiIs given by the neighbor of (a):
whereiniRepresents the user uiThe global trust score of (a) is,representing user uiOf the user's neighborhood, tj→iRepresenting user uiWith user ujA local trust value between, n represents user uiThe total number of neighbors in the user's neighborhood group.
In the above equation (5), the user uiWith user ujBook in betweenNumber of degree of confidence of ground tj→iThe separate calculation can be divided into the following two cases:
(a) when target user uiWith user ujOnly one path is in the trust network, the two paths are in direct trust relationship, and the trust degree t can be calculated by using the modified cosine similarity measurementj→i。
(a) When target user uiWith user ujThere is no item with common rating between them, that is, the two are in indirect trust relationship in the trust network, and at this time, the indirect trust degree can be calculated through trust propagation. In particular, at the connection source user ujWith target user uiHas an intermediate user u in the trust pathmThen pass umRespectively with target users uiWith user ujDirect trust score t in betweenj→mAnd tm→iCarrying out weighted average and calculating to obtain a target user uiWith user ujLocal confidence score between:
wherein, IjRepresenting items j, ImRepresenting item m.
By analogy, the propagation algorithm can be repeatedly applied to calculate the indirect trust relationship between any two users in the trust network.
And finally, predicting the scoring items of the target user based on the item scores of the users in the constructed trust neighbor cluster of the target user. In an embodiment of the present invention, a formula as follows is adopted to predict a scoring item of a target user based on the item scores of users in the trust neighbor cluster set of the target user, wherein the item scores are set according to the following formula:
wherein p isa,iRepresenting a target user uaIs not scoredThe prediction score of the item i is calculated,representing a target user uaOf the neighbor users in the set of trusted neighbor clusters, rk,jMean of scores for item i, gt, for neighbor user k in the set of trusted neighbor clusters of target userkDirect connection to user u in a trust network representing a set of trusted neighbors of a target useraIs the average of the local trust scores of the neighbor users of, B (u)a) Representing a target user uaTo trust neighbor users in the neighbor set.
Step S104: and recommending the corresponding item to the target user according to the predicted item score.
In a specific implementation, when the item is scored according to the predicted target user, a plurality of items with the predicted item scoring values arranged in the front can be selected and recommended to the user.
The item recommendation method based on predictive value filling in the embodiment of the present invention is described in detail above, and apparatuses corresponding to the above method will be described below.
Fig. 2 is a schematic structural diagram of an item recommendation device based on predictive value filling in the embodiment of the invention in this year. Referring to fig. 2, an item recommendation apparatus 20 based on prediction value filling may include an acquisition unit 201, a filling unit 202, a prediction unit 203, and a recommendation unit 204, in which:
the obtaining unit 201 is adapted to obtain an original user item scoring matrix.
The filling unit 202 is adapted to, when it is determined that the original user item score matrix is a sparse matrix, fill the original user item score matrix based on the similarity between the items to obtain a corresponding filling matrix.
The prediction unit 203 is adapted to predict the item scores of the target users based on the filling matrix.
The recommending unit 204 is adapted to recommend the corresponding item to the target user according to the predicted item score.
In a specific implementation, the filling unit 202 is adapted to calculate similarity between items in the original user item scoring matrix; determining a plurality of scoring items with the highest similarity to the unscored items in the scoring matrix of the original user items based on the similarity between the calculated items; calculating the prediction scores of the unscored items by adopting the scores of a plurality of scored items with the highest similarity to the unscored items in the original user item scoring matrix; filling the prediction scores of the unscored items into corresponding positions in the original user item score matrix.
In a specific implementation, the prediction unit 203 is adapted to perform dimension reduction on the filling matrix by using singular value decomposition to obtain a corresponding left singular matrix and a corresponding singular value matrix; constructing a user implicit characteristic space matrix based on the left singular matrix and the singular value matrix; determining a neighbor user group of each user based on the user implicit feature space matrix of the user; building a trust neighbor group set of a target user based on the trust relationship among the users in the neighbor user group; and predicting the scoring items of the target user based on the constructed item scores of the users in the trust neighbor cluster of the target user.
In an embodiment of the present invention, the prediction unit 203 is adapted to perform dimension reduction on the filling matrix by using singular value decomposition using the following formula:
wherein,is represented by P|U|×kRepresenting a left singular matrix obtained by dimensionality reduction of the filling matrix by singular value decomposition, Sk×kRepresenting a singular value matrix obtained by dimension reduction of the filling matrix by singular value decomposition,and D, performing dimensionality reduction on the filling matrix by adopting singular value decomposition to obtain a transposed matrix of a right singular matrix, wherein U represents the number of items, and k represents the number of users subjected to dimensionality reduction.
In an embodiment of the present invention, the prediction unit 203 is adapted to construct a user implicit feature space matrix based on the left singular matrix and the singular value matrix by using the following formula:
Y=PKSK 1/2;
wherein Y represents the user implicit feature space matrix, PKRepresenting a left singular matrix obtained by dimensionality reduction of the filling matrix by singular value decomposition, SKAnd representing a singular value matrix obtained by dimension reduction of the filling matrix by singular value decomposition.
In an embodiment of the present invention, the prediction unit 203 is adapted to calculate a similarity between every two users under the user implicit feature space matrix, and construct a user similarity matrix; converting the user similarity matrix into a corresponding fuzzy equivalent matrix; and classifying users with similar characteristics into one class according to the equivalent relation value in the fuzzy equivalent matrix to obtain a neighbor user group corresponding to each user.
In an embodiment of the present invention, the predicting unit 203 is adapted to calculate a trust degree value between a target user and a neighbor user in the trust neighbor group set by using modified cosine similarity when the target user and the neighbor user have a direct trust relationship; when a target user and a neighbor user in the trust neighbor group set have an indirect trust relationship, calculating a trust degree value between the target user and the neighbor user by adopting a trust propagation algorithm; and acquiring a plurality of neighbor users with higher trust degree values with the target user to form a trust neighbor group set of the target user.
In an embodiment of the present invention, the predicting unit 203 is adapted to predict the item score of the target user based on the item score of the user in the trust neighbor cluster of the target user, which is constructed by using the following formula:
wherein p isa,iThe predicted score of the unscored item i representing the target user a,mean, r, representing the user's score for item i in the target user's trusted neighbor clusterk,iMean of user k's score for item i in the target user's trusted neighbor cluster, gtKDirect connection to user u in user trust network in trust neighbor cluster representing target useraIs the average of the local trust scores of the neighbors of, B (u)a) And (4) showing.
The embodiment of the present invention further provides a computer-readable storage medium, on which computer instructions are stored, and when the computer instructions are executed, the method for recommending items based on filling of predicted values according to any of the above-mentioned steps is performed.
The embodiment of the invention also provides a terminal, which comprises a memory and a processor, wherein the memory is stored with computer instructions capable of being executed on the processor, and the processor executes the steps of any item recommendation method based on the filling of the predicted values when executing the computer instructions.
By adopting the scheme in the embodiment of the invention, when the original user item scoring matrix is determined to be the sparse matrix, the original user item scoring matrix is filled based on the similarity among the items to obtain the corresponding filling matrix, the item scoring of the target user is predicted based on the filling matrix, and finally the corresponding item is recommended to the target user according to the predicted item scoring, so that the problem of data sparseness can be effectively solved, and the accuracy of item recommendation is improved.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the foregoing description only for the purpose of illustrating the principles of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the invention as defined by the appended claims, specification, and equivalents thereof.
Claims (8)
1. A project recommendation method based on predictive value filling is characterized by comprising the following steps:
acquiring an original user project scoring matrix;
when the original user item scoring matrix is determined to be a sparse matrix, filling the original user item scoring matrix based on the similarity among the items to obtain a corresponding filling matrix;
predicting the project scores of the target users based on the filling matrix; specifically, the filling matrix is subjected to dimensionality reduction by singular value decomposition to obtain a corresponding left singular matrix and a corresponding singular value matrix; constructing a user implicit characteristic space matrix based on the left singular matrix and the singular value matrix; determining a neighbor user group of each user based on the user implicit feature space matrix of the user; building a trust neighbor group set of a target user based on the trust relationship among the users in the neighbor user group; predicting the scoring items of the target user based on the item scores of the users in the trust neighbor cluster of the target user; determining a neighbor user group of each user based on the user implicit feature space matrix, wherein the determining comprises: calculating the similarity between every two users under the user implicit characteristic space matrix, and constructing a user similarity matrix; converting the user similarity matrix into a corresponding fuzzy equivalent matrix; classifying users with similar characteristics into one class according to the equivalent relation value in the fuzzy equivalent matrix to obtain a neighbor user group corresponding to each user;
and recommending the corresponding item to the target user according to the predicted item score.
2. The predictive value filling-based item recommendation method according to claim 1, wherein the filling of the original user item rating matrix based on the similarity between the items comprises:
calculating the similarity between items in the original user item scoring matrix;
determining a plurality of scoring items with the highest similarity to the unscored items in the scoring matrix of the original user items based on the similarity between the calculated items;
calculating the prediction scores of the unscored items by adopting the scores of a plurality of scored items with the highest similarity to the unscored items in the original user item scoring matrix;
filling the prediction scores of the unscored items into corresponding positions in the original user item score matrix.
3. The predictive value filling-based item recommendation method according to claim 1, wherein the filling matrix is dimension-reduced by singular value decomposition using the following formula:
wherein,representing the original user rating matrix, P|U|×kRepresenting a left singular matrix obtained by dimensionality reduction of the filling matrix by singular value decomposition, Sk×kRepresenting a singular value matrix obtained by dimension reduction of the filling matrix by singular value decomposition,and D, performing dimensionality reduction on the filling matrix by adopting singular value decomposition to obtain a transposed matrix of a right singular matrix, wherein U represents the number of items, and k represents the number of users subjected to dimensionality reduction.
4. The project recommendation method based on predictive value filling according to claim 1, characterized in that a user implicit feature space matrix is constructed based on the left singular matrix and singular value matrix by using the following formula:
wherein Y represents the user implicit feature space matrix, PKRepresenting a left singular matrix obtained by dimensionality reduction of the filling matrix by singular value decomposition, SKAnd representing a singular value matrix obtained by dimension reduction of the filling matrix by singular value decomposition.
5. The project recommendation method based on predictive value filling according to claim 1, characterized in that the building of the trust neighborhood group set of the trust users based on the trust relationship among the users in the neighborhood user group comprises:
when a target user has a direct trust relationship with a neighbor user in the trust neighbor group set, calculating a trust degree value between the target user and the neighbor user by adopting modified cosine similarity;
when a target user and a neighbor user in the trust neighbor group set have an indirect trust relationship, calculating a trust degree value between the target user and the neighbor user by adopting a trust propagation algorithm;
and acquiring a plurality of neighbor users with higher trust degree values with the target user to form a trust neighbor group set of the target user.
6. The project recommendation method based on predictive value filling according to claim 5, characterized in that the project scores of the users in the trust neighbor cluster of the target user are predicted based on the project scores of the users in the constructed trust neighbor cluster of the target user by adopting the following formula:
wherein p isa,iRepresenting a target user uaThe predicted score of the non-scored item i of (a),representing a target user uaOf the neighbor users in the set of trusted neighbor clusters, rk,iMean of scores for item i, gt, for neighbor user k in the set of trusted neighbor clusters of target userkDirect connection to user u in a trust network representing a set of trusted neighbors of a target useraIs the average of the local trust scores of the neighbor users of, B (u)a) Representing a target user uaTo trust neighbor users in the neighbor set.
7. A computer readable storage medium having stored thereon computer instructions, wherein the computer instructions when executed perform the steps of the predictive value based population item recommendation method of any of claims 1 to 6.
8. A terminal comprising a memory and a processor, the memory having stored thereon computer instructions capable of being executed on the processor, the processor when executing the computer instructions performing the steps of the project recommendation method based on predictive value population of any of claims 1 to 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811181074.4A CN109241442B (en) | 2018-10-10 | 2018-10-10 | Project recommendation method based on predictive value filling, readable storage medium and terminal |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811181074.4A CN109241442B (en) | 2018-10-10 | 2018-10-10 | Project recommendation method based on predictive value filling, readable storage medium and terminal |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109241442A CN109241442A (en) | 2019-01-18 |
CN109241442B true CN109241442B (en) | 2022-03-11 |
Family
ID=65052569
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811181074.4A Active CN109241442B (en) | 2018-10-10 | 2018-10-10 | Project recommendation method based on predictive value filling, readable storage medium and terminal |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109241442B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109857935B (en) * | 2019-01-24 | 2021-08-24 | 腾讯科技(深圳)有限公司 | Information recommendation method and device |
CN110288444B (en) * | 2019-06-28 | 2022-03-01 | 第四范式(北京)技术有限公司 | Method and system for realizing user related recommendation |
CN111833990A (en) * | 2020-07-17 | 2020-10-27 | 电子科技大学 | Method for filling missing items of psychological assessment scale |
CN113221003B (en) * | 2021-05-20 | 2023-05-02 | 北京建筑大学 | Mixed filtering recommendation method and system based on dual theory |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104317900A (en) * | 2014-10-24 | 2015-01-28 | 重庆邮电大学 | Multiattribute collaborative filtering recommendation method oriented to social network |
-
2018
- 2018-10-10 CN CN201811181074.4A patent/CN109241442B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104317900A (en) * | 2014-10-24 | 2015-01-28 | 重庆邮电大学 | Multiattribute collaborative filtering recommendation method oriented to social network |
Non-Patent Citations (2)
Title |
---|
"一种SVD和信任因子相结合的协同过滤算法";王建芳等;《小型微型计算机***》;20170615;第38卷(第6期);第1292页 * |
"基于项目评分预测的协同过滤推荐算法";邓爱林等;《软件学报》;20030923;第14卷(第9期);第1623-1625页 * |
Also Published As
Publication number | Publication date |
---|---|
CN109241442A (en) | 2019-01-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Solus et al. | Consistency guarantees for greedy permutation-based causal inference algorithms | |
CN109241442B (en) | Project recommendation method based on predictive value filling, readable storage medium and terminal | |
CN109190023B (en) | Collaborative recommendation method and device and terminal equipment | |
EP3327582A1 (en) | Method and apparatus for completing a knowledge graph | |
US10565518B2 (en) | Collaborative feature learning from social media | |
CN106055661B (en) | More interest resource recommendations based on more Markov chain models | |
CN110162711B (en) | Resource intelligent recommendation method and system based on network embedding method | |
Liang et al. | A probabilistic rating auto-encoder for personalized recommender systems | |
CN107943897B (en) | User recommendation method | |
CN115062732B (en) | Resource sharing cooperation recommendation method and system based on big data user tag information | |
CN108389113B (en) | Collaborative filtering recommendation method and system | |
CN104063555B (en) | The user model modeling method intelligently distributed towards remote sensing information | |
CN114637920A (en) | Object recommendation method and device | |
CN113239266B (en) | Personalized recommendation method and system based on local matrix decomposition | |
CN108427730A (en) | It is a kind of that method is recommended based on the Social Label of random walk and condition random field | |
CN110457387B (en) | Method and related device applied to user tag determination in network | |
Idrissi et al. | A new hybrid-enhanced recommender system for mitigating cold start issues | |
CN116467466A (en) | Knowledge graph-based code recommendation method, device, equipment and medium | |
CN114332472A (en) | Data processing method and device based on graph neural network | |
Spelta et al. | Wasserstein barycenter for link prediction in temporal networks | |
CN111428741B (en) | Network community discovery method and device, electronic equipment and readable storage medium | |
Abdullah et al. | Predicting qos for web service recommendations based on reputation and location clustering with collaborative filtering | |
CN109241443A (en) | Project recommendation device based on predicted value filling | |
Abdelwahab et al. | Alleviating the sparsity problem of collaborative filtering using an efficient iterative clustered prediction technique | |
CN113807749B (en) | Object scoring method and device |
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 |