CN113360759A - Crowd-sourcing task recommendation method based on dual timing sequence correlation of user and project - Google Patents
Crowd-sourcing task recommendation method based on dual timing sequence correlation of user and project Download PDFInfo
- Publication number
- CN113360759A CN113360759A CN202110641787.XA CN202110641787A CN113360759A CN 113360759 A CN113360759 A CN 113360759A CN 202110641787 A CN202110641787 A CN 202110641787A CN 113360759 A CN113360759 A CN 113360759A
- Authority
- CN
- China
- Prior art keywords
- user
- time slice
- project
- time
- users
- 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
- 238000000034 method Methods 0.000 title claims description 30
- 230000009977 dual effect Effects 0.000 title claims description 18
- 238000012358 sourcing Methods 0.000 title claims description 18
- 239000011159 matrix material Substances 0.000 claims abstract description 56
- 239000013598 vector Substances 0.000 claims abstract description 37
- 230000003993 interaction Effects 0.000 claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 27
- 230000006399 behavior Effects 0.000 claims abstract description 20
- 238000001914 filtration Methods 0.000 claims abstract description 15
- 230000001537 neural effect Effects 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims description 22
- 238000012163 sequencing technique Methods 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 238000013522 software testing Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 102100035115 Testin Human genes 0.000 description 1
- 101710070533 Testin Proteins 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
S1, calculating the relation weight between users and items, constructing an attribute network by using the partial sequence relation of historical interaction behaviors, and defining the similar relation between the users and the items through the attribute network; s2, calculating the relevance of the time slices, acquiring a similar time slice set of any time slice, and calculating the similar relation between the user time slices and the project time slices by using the relation weight between the users and the projects; s3, constructing a to-be-predicted time slice user item matrix, wherein the user item interaction matrix is obtained by multiplying a user time slice characteristic matrix and a project time slice characteristic matrix, and the two characteristic matrices are formed by similar time slice vectors of the to-be-predicted time slices; and S4, continuously training the neural collaborative filtering network along with the time slice updating until the network is updated to the time slice to be predicted, and then predicting the user behavior to obtain a prediction result.
Description
Technical Field
The invention relates to the technical field of crowd-sourcing task recommendation, in particular to a crowd-sourcing task recommendation method based on dual timing sequence correlation of users and projects.
Background
Since the twenty-first century, human society has entered the information-based era. The application of computer and internet technology is rapidly popularized, and the vigorous software industry needs to directly promote the academic development of related subjects until today, wherein software testing is one of the important links in the software life cycle, but the traditional software testing time rhythm is slow, so that the software quality problem cannot be more comprehensively covered, and the software testing task under the internet environment is difficult to adapt.
The crowd-sourcing workers are capable of distributing software testing tasks to non-specific groups using the internet, are of basic skill, and are willing to work with spare time to obtain corresponding rewards or other services. With the development of the many-test platform, the number of testers, task publishers and tasks to be tested increases dramatically, and if the help of an automatic recommendation method is not provided, a platform user can encounter the problem of information overload and can hardly acquire real interesting and valuable test tasks, so that the importance of a recommendation algorithm is emphasized. The interaction record formed by the user and the task in the crowd-sourced scene is implicit feedback data, more context information is utilized to solve the data sparsity problem and the cold start problem widely existing in the recommendation problem, and the utilization of the time sequence information is an important practice method.
A traditional matrix decomposition method and a collaborative filtering method based on a user are used as classical algorithms in recommendation problems and are often used for recommendation problem practice, but the collaborative rules of test tasks cannot be comprehensively reflected purely based on the user in the public testing field, and the nonlinear interaction relation between the user and the tasks cannot be effectively extracted through the traditional matrix decomposition method.
Disclosure of Invention
The purpose of the invention is as follows: in order to overcome the defects in the prior art, the invention provides a crowd-sourced test task recommendation method based on dual time sequence correlation of users and projects.
The technical scheme is as follows: in order to achieve the above object, the present invention provides a method for recommending a task for people testing based on dual timing correlations between users and items, which is characterized in that: comprises the following steps of (a) carrying out,
s1, calculating the relation weight between users and items, constructing an attribute network by using the partial order relation of historical interaction behaviors, and defining the similar relation between users and items through the attribute network;
s2, calculating the relevance of the time slices, acquiring a similar time slice set of any time slice, and calculating the similar relation between the user time slices and the project time slices by using the relation weight between the users and the projects;
s3, constructing a to-be-predicted time slice user item matrix, wherein the user item interaction matrix is obtained by multiplying a user time slice characteristic matrix and a project time slice characteristic matrix, and the two characteristic matrices are formed by similar time slice vectors of the to-be-predicted time slices; (ii) a
S4, continuously training the neural collaborative filtering network along with the updating of the time slice until the network is updated to the time slice to be predicted, and then predicting the user behavior to obtain a prediction result;
and S5, sequencing the prediction results to obtain a final prediction result of the behaviors between the future user and the project.
Further, in the present invention: the S1 further includes the following steps,
s11, traversing the user-project interaction records, and constructing a consumption network graph by taking the user as a node and the user consuming the same project successively as edges;
s12, acquiring the weight of the edge between two users according to the number of the partial order relationship, wherein the ratio of the weight of the edge between two users and the module of the set union of the two user consumption products represents the influence of the preorder user on the subsequent user in the partial order relationship, so that the influence of any user on all other users can be obtained;
and S13, randomly ordering all users, and then constructing an influence vector by using the influence weight of a certain user on other users, wherein each dimension in the influence vector represents the influence of the user on other specified users.
Further, in the present invention: the S2 further includes the following steps,
s21, the time slice relativity comprises user side time slice relativity and item side time slice relativity, and the user side time slice relativity is calculated according to the weighted cosine similarity by using the user influence vector and the user-time slice vector;
s22, calculating the relevance of the project side time slice according to the weighted cosine similarity by using the project influence vector and the project-time slice vector; the influence is divided into a user influence and a project influence, which are mixed in S23;
and S23, reconciling the user time slice correlation and the project time slice correlation by using a linear method to obtain the dual correlation of the time slices.
Further, in the present invention: the S3 further includes the following steps,
s31, traversing all user behavior time slice data before the time slice to be predicted, calculating a similar time slice set corresponding to each time slice, and sequencing elements in the sets according to the similarity;
s32, taking similar time slices of the pre-sequence time slices of the L time slices to be predicted to form a set, directly replacing each element in the set with the subsequent time slice, wherein the subsequent time slice set obtained by replacement is the similar time slice set of the time slices to be predicted;
and S33, dividing the similar time slice set into a user similar time slice set and a project similar time slice set, wherein the user characteristic matrix and the project characteristic matrix are respectively formed by the two sets, and multiplying the two matrixes to obtain the user project interaction matrix to be predicted.
Further, in the present invention: the S4 neural network collaborative filtering for predicting user behavior further includes,
establishing initial feature vectors for users and projects by utilizing an Embedding layer;
connecting the eigenvectors by utilizing a multilayer perceptron and generalized matrix decomposition, obtaining a prediction result through an output layer, calculating loss with a true value in a matrix to be predicted, and taking a minimum loss function as a target;
and training by continuously taking the time in the training data set as the time to be predicted, wherein the training result of each time is taken as the initial value of the network of the next training.
Further, in the present invention: and S5, sequencing the prediction data obtained after the neural network training in a descending order according to the interaction frequency of the user and the items, and pushing a plurality of items with the most interaction frequency as a recommendation list to the user.
Further, in the present invention: the item is a task to be recommended, and the task to be recommended is a task to be tested in the public testing platform.
Has the advantages that: the user project interaction matrix is obtained by multiplying a user time slice characteristic matrix and a project time slice characteristic matrix, and the two characteristic matrices are formed by similar time slice vectors of time slices to be predicted; continuously training the neural collaborative filtering network along with updating of the time slice until the network is updated to the time slice to be predicted, and then predicting the user behavior to obtain a prediction result; and sequencing the prediction results to obtain a final prediction result of the behavior between the future user and the project. According to the invention, implicit feedback information between the user and the test item can be better utilized, the recommendation effect is improved, and the development of a public test platform is promoted. According to the invention, based on the dual time sequence correlation of the user side and the project side and the neural collaborative filtering method, the test tasks on the many-test platform can be personalized and recommended; by using the method provided by the invention, the periodic hot spot problem on the public measurement platform can be recommended by fully utilizing the time sequence context information, and the time sequence data of the public measurement platform is utilized to the maximum extent; the interactive relation between the user and the task project can be fully extracted by neural collaborative filtering, the personalized characteristics of the user can be paid attention to efficiently, the testing efficiency and the testing quality of the user are improved, the task throughput of a testing platform is improved, the development of public testing is promoted, the recommendation result obtained by the recommendation method is high in fidelity and short in calculation time, and the recommendation method is beneficial to popularization in practical application.
Drawings
FIG. 1 is a schematic overall flow chart of a crowd-sourcing task recommendation method based on dual timing correlations of users and projects according to the present invention;
FIG. 2 is a schematic diagram illustrating a process of calculating weights of relationships between users and items according to the present invention;
FIG. 3 is a schematic flow chart of constructing a user item interaction matrix to be predicted for a time slice according to the present invention;
FIG. 4 is a schematic flow chart of the neural collaborative filtering network training according to the present invention.
Detailed Description
The technical scheme of the invention is further explained in detail by combining the attached drawings:
the present invention may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
As shown in fig. 1, an overall flow chart of the crowd-sourcing task recommendation method based on dual timing correlations of users and projects according to the present invention is shown, specifically, the method includes the following steps,
s1, calculating the relation weight between users and items, constructing an attribute network by using the partial order relation of historical interaction behaviors, and defining the similar relation between users and items through the attribute network;
specifically, referring to the illustration of fig. 2, the S1 further includes the following steps,
s11, traversing the user-project interaction records, and constructing a consumption network graph by taking the user as a node and the user consuming the same project successively as edges; the user-project interaction record in the embodiment is recorded through interaction of the user and the project in the real audience measurement platform, and the project is the task to be measured. The crowd-sourcing test platform of the embodiment is at least one of a hundredth crowd-sourcing test platform, a CNVD (common name distribution video distribution), a 360 crowd-sourcing test platform, a Testin, a pig eight-ring network and a crowd-sourcing test platform. The tasks to be tested are software testing tasks, including mobile terminal software testing tasks, personal computer terminal testing tasks, performance testing tasks and the like.
Further, traversing the interaction records to construct a consumption network, e.g., taking users i and j as network nodes to obtain a node Ui、UjWhen the user i and the user j successively consume the same project, two unidirectional edges E are generated between the two nodesi,j、Ej,iWhenever user i or user j consumes the same item before user j or user i, the corresponding one-way edge weight is increased. All the unidirectional edges with the user i as the starting point jointly form the representation of the influence of the user, and the influence of the user on other users such as j can be represented by the modulus ratio of the weight corresponding to the unidirectional edge to the user i and j consumption item complete set to obtain the influence relation weight W between the usersi,jAnd calculating the influence relation weight among the ratio items.
S12, acquiring the weight of the edge between two users according to the number of the partial order relationship, wherein the ratio of the weight of the edge between two users and the module of the set union of the two user consumption products represents the influence of the preorder user on the subsequent user in the partial order relationship, so that the influence of any user on all other users can be obtained;
and S13, randomly ordering all users, and then constructing an influence vector by using the influence weight of a certain user on other users, wherein each dimension in the influence vector represents the influence of the user on other specified users.
S2, calculating the relevance of the time slices, acquiring a similar time slice set of any time slice, and calculating the similar relation between the user time slices and the project time slices by using the relation weight between the users and the projects; the relevance between any two time slices is defined by the relevance of the time slices shown by the user side and the relevance of the time slices shown by the item side. User time slice formed by user and time sliceThe similarity measure of the matrices, time slices, being weighted cosine similarity, e.g. time slice vector TiAnd TjWhen calculating the similarity, each dimension is transformed by different influence weights of the same users in two time slices, and the same users have different global influence in different time slices, so the weight is necessary in weighting cosine similarity. Therefore, the similarity defined by the user group between any two time slices can be obtained, the similarity defined by the project group between any two time slices can be obtained in the same way, the double similarities between the two time slices are obtained through super-parameter alpha linear harmony, and the parameters are used for balancing the proportional relation of the two similarities.
Specifically, the step S2 further includes the step of,
s21, the time slice relativity comprises user side time slice relativity and item side time slice relativity, and the user side time slice relativity is calculated according to the weighted cosine similarity by using the user influence vector and the user-time slice vector;
s22, calculating the relevance of the project side time slice according to the weighted cosine similarity by using the project influence vector and the project-time slice vector;
and S23, reconciling the user time slice correlation and the project time slice correlation by using a linear method to obtain the dual correlation of the time slices.
S3, constructing a to-be-predicted time slice user item matrix, wherein the user item interaction matrix is obtained by multiplying a user time slice characteristic matrix and a project time slice characteristic matrix, and the two characteristic matrices are formed by similar time slice vectors of the to-be-predicted time slices;
specifically, the step S3 further includes the step of,
s31, traversing all user behavior time slice data before the time slice to be predicted, calculating a similar time slice set corresponding to each time slice, and sequencing elements in the sets according to the similarity; according to the double similarity between any two time slices calculated in the above embodiment, a similar time slice set of the current time slice can be obtained.
S32, taking similar time slices of the pre-sequence time slices of the L time slices to be predicted to form a set, directly replacing each element in the set with the subsequent time slice, wherein the subsequent time slice set obtained by replacement is the similar time slice set of the time slices to be predicted;
and S33, dividing the similar time slice set into a user similar time slice set and a project similar time slice set, wherein the user characteristic matrix and the project characteristic matrix are respectively formed by the two sets, and multiplying the two matrixes to obtain the user project interaction matrix to be predicted.
Furthermore, in this embodiment, L elements with the highest similarity are taken to form a new matrix, which is a user feature matrix in the current time slice and has dimensions of M × L, where M denotes the number of users; and then, constructing a project characteristic matrix, and replacing the column vector elements in the user characteristic matrix with corresponding column vectors in a project time slice matrix to obtain the project characteristic matrix of the current time slice, wherein the dimensionality is NxL, and N refers to the number of projects. The feature matrix of the current time slice is constructed in the embodiment, the time slice in the feature matrix is directly replaced by the direct subsequent time slice, and the obtained new feature matrix is the feature matrix of the next time slice, namely the feature matrix of the time slice to be predicted.
S4, continuously training the neural collaborative filtering network along with the updating of the time slice until the network is updated to the time slice to be predicted, and then predicting the user behavior to obtain a prediction result;
the neural collaborative filtering network comprises an embedding layer, a multi-layer perceptron layer, a generalized matrix grading layer and an output layer; the embedding layer extracts the features of the word embedding word vector and sends the word embedding word vector into the multilayer perceptron layer and the generalized matrix grading layer; the multilayer perceptron layer extracts the nonlinear interaction rule between the user and the project; the generalized matrix hierarchical layer extracts a linear or low-dimensional interaction rule between a user and a project; and the output layer receives and fuses the vector characteristics output by the multi-layer perceptron layer and the generalized matrix hierarchical layer and processes and outputs a final prediction result. In this embodiment, the model is trained based on the training result and the real result.
Specifically, the predicting the user behavior by the S4 neural network collaborative filtering further includes,
establishing initial feature vectors for users and projects by utilizing an Embedding layer; in this embodiment, data of a ONE-HOT (ONE-HOT) vector is utilized, and is converted into a word embedding word vector based on a word2vector method, for example, the above-mentioned ONE-HOT vector is expressed as an M × 100 dimensional matrix vector, where M is the number of users, and 100 is the size of the word embedding word vector.
Connecting the eigenvectors by utilizing a multilayer perceptron and generalized matrix decomposition, obtaining a prediction result through an output layer, calculating loss with a true value in a matrix to be predicted, and taking a minimum loss function as a target;
and training by continuously taking the time in the training data set as the time to be predicted, wherein the training result of each time is taken as the initial value of the network of the next training. Specifically, the neural collaborative network is continuously updated by using training data, and referring to the schematic diagram of fig. 4, the training data is sequence data obtained by sequencing time slices, starting with the L +1 time slice, and is used as the current time slice to calculate the user project influence weight of the L +2 time slice, construct a user and project characteristic matrix, construct a user project interaction matrix, and perform fitting training on the matrix through the neural collaborative filtering network; and then, taking the L +2 th time slice as the current time slice, predicting the L +3 th time slice so as to continuously train the neural collaborative filtering network, wherein each training result is used as the initial value of the network for the next training until the target time slice prediction result.
And S5, sequencing the prediction results to obtain a final prediction result of the behaviors between the future user and the project.
Specifically, the S5 performs descending order sorting on the prediction data obtained after the neural network training according to the interaction frequency of the user and the items, and pushes a plurality of items with the highest interaction frequency as a recommendation list to the user.
Further, the ranking basis is generally the interaction frequency, and the higher the frequency is, the more likely the interaction behavior is to occur with the user, so the more advanced the ranking in the recommendation list is, the first 5-10 in the list can be generally taken as the list to be recommended.
It should be noted that the above-mentioned examples only represent some embodiments of the present invention, and the description thereof should not be construed as limiting the scope of the present invention. It should be noted that, for those skilled in the art, various modifications can be made without departing from the spirit of the present invention, and these modifications should fall within the scope of the present invention.
Claims (7)
1. A crowd-sourcing task recommendation method based on dual timing sequence correlation of users and projects is characterized by comprising the following steps: comprises the following steps of (a) carrying out,
s1, calculating the relation weight between users and items, constructing an attribute network by using the partial order relation of historical interaction behaviors, and defining the similar relation between users and items through the attribute network;
s2, calculating the relevance of the time slices, acquiring a similar time slice set of any time slice, and calculating the similar relation between the user time slices and the project time slices by using the relation weight between the users and the projects;
s3, constructing a to-be-predicted time slice user item matrix, wherein the user item interaction matrix is obtained by multiplying a user time slice characteristic matrix and a project time slice characteristic matrix, and the two characteristic matrices are formed by similar time slice vectors of the to-be-predicted time slices;
s4, continuously training the neural collaborative filtering network along with the updating of the time slice until the network is updated to the time slice to be predicted, and then predicting the user behavior to obtain a prediction result;
and S5, sequencing the prediction results to obtain a final prediction result of the behaviors between the future user and the project.
2. The crowd-sourcing task recommendation method based on user and item dual timing correlations as claimed in claim 1, wherein: the S1 further includes the following steps,
s11, traversing the user-project interaction records, and constructing a consumption network graph by taking the user as a node and the user consuming the same project successively as edges;
s12, acquiring the weight of the edge between two users according to the number of the partial order relationship, wherein the ratio of the weight of the edge between two users and the module of the set union of the two user consumption products represents the influence of the preorder user on the subsequent user in the partial order relationship, so that the influence of any user on all other users can be obtained;
and S13, randomly ordering all users, and then constructing an influence vector by using the influence weight of a certain user on other users, wherein each dimension in the influence vector represents the influence of the user on other specified users.
3. The crowd-sourcing task recommendation method based on user and item dual timing correlations as claimed in claim 2, wherein: the S2 further includes the following steps,
s21, the time slice relativity comprises user side time slice relativity and item side time slice relativity, and the user side time slice relativity is calculated according to the weighted cosine similarity by using the user influence vector and the user-time slice vector;
s22, calculating the relevance of the project side time slice according to the weighted cosine similarity by using the project influence vector and the project-time slice vector; the influence is divided into a user influence and a project influence, which are mixed in S23;
and S23, reconciling the user time slice correlation and the project time slice correlation by using a linear method to obtain the dual correlation of the time slices.
4. The crowd-sourcing task recommendation method based on user and item dual timing correlations as claimed in claim 3, wherein: the S3 further includes the following steps,
s31, traversing all user behavior time slice data before the time slice to be predicted, calculating a similar time slice set corresponding to each time slice, and sequencing elements in the sets according to the similarity;
s32, taking similar time slices of the pre-sequence time slices of the L time slices to be predicted to form a set, directly replacing each element in the set with the subsequent time slice, wherein the subsequent time slice set obtained by replacement is the similar time slice set of the time slices to be predicted;
and S33, dividing the similar time slice set into a user similar time slice set and a project similar time slice set, wherein the user characteristic matrix and the project characteristic matrix are respectively formed by the two sets, and multiplying the two matrixes to obtain the user project interaction matrix to be predicted.
5. The crowd-sourcing task recommendation method based on user and item dual timing correlations as claimed in claim 4, wherein: the S4 neural network collaborative filtering for predicting user behavior further includes,
establishing initial feature vectors for users and projects by utilizing an Embedding layer;
connecting the eigenvectors by utilizing a multilayer perceptron and generalized matrix decomposition, obtaining a prediction result through an output layer, calculating loss with a true value in a matrix to be predicted, and taking a minimum loss function as a target;
and training by continuously taking the time in the training data set as the time to be predicted, wherein the training result of each time is taken as the initial value of the network of the next training.
6. The crowd-sourcing task recommendation method based on user and item dual timing correlations as claimed in claim 5, wherein: and S5, sequencing the prediction data obtained after the neural network training in a descending order according to the interaction frequency of the user and the items, and pushing a plurality of items with the most interaction frequency as a recommendation list to the user.
7. The crowd-sourcing task recommendation method based on user and item dual timing correlations as claimed in claim 6, wherein: the item is a task to be recommended, and the task to be recommended is a task to be tested in the public testing platform.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110641787.XA CN113360759B (en) | 2021-06-09 | 2021-06-09 | Crowd measurement task recommendation method based on user and project dual time sequence correlation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110641787.XA CN113360759B (en) | 2021-06-09 | 2021-06-09 | Crowd measurement task recommendation method based on user and project dual time sequence correlation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113360759A true CN113360759A (en) | 2021-09-07 |
CN113360759B CN113360759B (en) | 2023-08-25 |
Family
ID=77533318
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110641787.XA Active CN113360759B (en) | 2021-06-09 | 2021-06-09 | Crowd measurement task recommendation method based on user and project dual time sequence correlation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113360759B (en) |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109241399A (en) * | 2018-07-11 | 2019-01-18 | 杭州电子科技大学 | A kind of personalized recommendation method combined based on trust and influence power |
CN109543109A (en) * | 2018-11-27 | 2019-03-29 | 山东建筑大学 | A kind of proposed algorithm of time of fusion window setting technique and score in predicting model |
US20190188272A1 (en) * | 2017-12-18 | 2019-06-20 | Microsoft Technology Licensing, Llc | Personalized content authoring driven by recommendations |
CN110162706A (en) * | 2019-05-22 | 2019-08-23 | 南京邮电大学 | A kind of personalized recommendation method and system based on interaction data cluster |
CN110807154A (en) * | 2019-11-08 | 2020-02-18 | 内蒙古工业大学 | Recommendation method and system based on hybrid deep learning model |
CN111046294A (en) * | 2019-12-27 | 2020-04-21 | 支付宝(杭州)信息技术有限公司 | Click rate prediction method, recommendation method, model, device and equipment |
CN111079005A (en) * | 2019-12-06 | 2020-04-28 | 成都理工大学 | Recommendation method based on article time popularity |
CN112084407A (en) * | 2020-09-08 | 2020-12-15 | 辽宁工程技术大学 | Collaborative filtering recommendation method fusing graph neural network and attention mechanism |
CN112631560A (en) * | 2020-12-29 | 2021-04-09 | 上海海事大学 | Method and terminal for constructing objective function of recommendation model |
-
2021
- 2021-06-09 CN CN202110641787.XA patent/CN113360759B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190188272A1 (en) * | 2017-12-18 | 2019-06-20 | Microsoft Technology Licensing, Llc | Personalized content authoring driven by recommendations |
CN109241399A (en) * | 2018-07-11 | 2019-01-18 | 杭州电子科技大学 | A kind of personalized recommendation method combined based on trust and influence power |
CN109543109A (en) * | 2018-11-27 | 2019-03-29 | 山东建筑大学 | A kind of proposed algorithm of time of fusion window setting technique and score in predicting model |
CN110162706A (en) * | 2019-05-22 | 2019-08-23 | 南京邮电大学 | A kind of personalized recommendation method and system based on interaction data cluster |
CN110807154A (en) * | 2019-11-08 | 2020-02-18 | 内蒙古工业大学 | Recommendation method and system based on hybrid deep learning model |
CN111079005A (en) * | 2019-12-06 | 2020-04-28 | 成都理工大学 | Recommendation method based on article time popularity |
CN111046294A (en) * | 2019-12-27 | 2020-04-21 | 支付宝(杭州)信息技术有限公司 | Click rate prediction method, recommendation method, model, device and equipment |
CN112084407A (en) * | 2020-09-08 | 2020-12-15 | 辽宁工程技术大学 | Collaborative filtering recommendation method fusing graph neural network and attention mechanism |
CN112631560A (en) * | 2020-12-29 | 2021-04-09 | 上海海事大学 | Method and terminal for constructing objective function of recommendation model |
Non-Patent Citations (2)
Title |
---|
HOUDA KHROUF ET AL.: "Hybrid event recommendation using linked data and user diversity", 《PROCEEDINGS OF THE 7TH ACM CONFERENCE ON RECOMMENDER SYSTEMS》, pages 185 - 192 * |
康迪迪: "基于时序信息的推荐算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》, pages 138 - 1484 * |
Also Published As
Publication number | Publication date |
---|---|
CN113360759B (en) | 2023-08-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110008399B (en) | Recommendation model training method and device, and recommendation method and device | |
US20210271975A1 (en) | User tag generation method and apparatus, storage medium, and computer device | |
Sandryhaila et al. | Discrete signal processing on graphs | |
CN110503531B (en) | Dynamic social scene recommendation method based on time sequence perception | |
WO2019061905A1 (en) | User behavior prediction method and apparatus, and electronic device | |
CN109345302A (en) | Machine learning model training method, device, storage medium and computer equipment | |
CN110555753A (en) | recommendation-based ranking control method and device, computer equipment and storage medium | |
Huynh et al. | Context-similarity collaborative filtering recommendation | |
CN109189922B (en) | Comment evaluation model training method and device | |
Vilakone et al. | Movie recommendation system based on users’ personal information and movies rated using the method of k-clique and normalized discounted cumulative gain | |
CN108205537A (en) | A kind of video recommendation method and system | |
CN115438787A (en) | Training method and device of behavior prediction system | |
CN109657725B (en) | Service quality prediction method and system based on complex space-time context awareness | |
Lian et al. | Cross-device user matching based on massive browse logs: The runner-up solution for the 2016 cikm cup | |
Zhao et al. | FINT: field-aware interaction neural network for click-through rate prediction | |
Li et al. | Time-aware hyperbolic graph attention network for session-based recommendation | |
CN113850654A (en) | Training method of item recommendation model, item screening method, device and equipment | |
CN113763095B (en) | Information recommendation method and device and model training method and device | |
CN113360759A (en) | Crowd-sourcing task recommendation method based on dual timing sequence correlation of user and project | |
Mitra et al. | Recommendation system based on product purchase analysis | |
CN114429384B (en) | Intelligent product recommendation method and system based on e-commerce platform | |
Feng et al. | Retracted: Recommendation Algorithm Based on Tag Time Weighting | |
Nazari et al. | Increasing performance of recommender systems by combining deep learning and extreme learning machine | |
CN114782062A (en) | Commodity recall optimization method and device, equipment, medium and product thereof | |
Agagu et al. | Context-aware recommendation methods |
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 |