CN112529350A - Developer recommendation method for cold start task - Google Patents
Developer recommendation method for cold start task Download PDFInfo
- Publication number
- CN112529350A CN112529350A CN202010538552.3A CN202010538552A CN112529350A CN 112529350 A CN112529350 A CN 112529350A CN 202010538552 A CN202010538552 A CN 202010538552A CN 112529350 A CN112529350 A CN 112529350A
- Authority
- CN
- China
- Prior art keywords
- cold start
- implicit
- task
- model
- explicit
- 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 abstract description 31
- 238000012549 training Methods 0.000 claims abstract description 28
- 230000004927 fusion Effects 0.000 claims abstract description 26
- 238000013507 mapping Methods 0.000 claims abstract description 15
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 13
- 238000004364 calculation method Methods 0.000 claims abstract description 5
- 230000009467 reduction Effects 0.000 claims description 22
- 238000012417 linear regression Methods 0.000 claims description 15
- 239000013598 vector Substances 0.000 claims description 13
- 238000005457 optimization Methods 0.000 claims description 8
- 230000003993 interaction Effects 0.000 claims description 7
- 230000008569 process Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 101150104012 TOP2 gene Proteins 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction 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
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- G06Q10/105—Human resources
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Machine Translation (AREA)
Abstract
The invention discloses a developer recommendation method for a cold start task, which is realized by the following steps: fusing the explicit characteristics, the ID characteristics and the corresponding grading information of the existing tasks and the existing developers on the training set by utilizing an FM algorithm, and establishing an FM-based fusion model; based on implicit characteristics of a finished task obtained through training, a mapping relation from the explicit characteristics to the implicit characteristics is modeled by using a deep regression model; based on the mapping relation, calculating the implicit characteristic of the cold start task according to the explicit characteristic of the cold start task; and carrying out scoring prediction by utilizing an FM fusion model according to the explicit characteristics of the cold start task and the implicit characteristics of the cold start task obtained by calculation. The recommendation method is oriented to a cold start scene, the explicit characteristics and the ID characteristics of the task are comprehensively considered, the problem of the absence of the implicit characteristics of the cold start task in the FM model is solved by constructing the mapping relation from the explicit characteristics to the implicit characteristics, and the cold start problem is effectively solved.
Description
Technical Field
The invention belongs to the field of artificial intelligence, and particularly relates to a developer recommendation method for a cold start task.
Background
Crowd-sourced software engineering typically involves three different types of participants: employers (Requesters) who have software development work to be done; workers (Workers) who participate in developing software; there is also a platform (platfoms) which provides an online marketplace.
The development of various crowdsourced software platforms has been rapid in recent years, and the number of released tasks and the number of registered developers have increased dramatically. Thus, the problem of "information overload" on crowdsourced software platforms is becoming more severe, which makes both task parties and developers faced with serious selection difficulties. In this context, the recommendation of developers facing a crowdsourced software development platform has important research and application values, and has attracted the attention of some researchers in recent years.
However, the recommendation performance of existing developer recommendation algorithms for cold start tasks (new release tasks) is not ideal.
Disclosure of Invention
The invention aims to provide a developer recommendation method for a cold start task, which can not only utilize the explicit characteristics of tasks and developers, but also consider the ID characteristics of the tasks and the developers closely related to scoring (evaluation results of the tasks to the developers), and can effectively solve the technical problem of poor recommendation performance of the developers on the existing cold start task.
The invention is realized by adopting the following technical scheme:
a developer recommendation method for a cold start task is provided, and is realized by the following steps: step 1, fusing the explicit characteristics, the ID characteristics and the corresponding grading information of the existing tasks and the existing developers by utilizing an FM algorithm, and establishing an FM-based fusion model; step 2, based on the implicit characteristics of the completed task obtained by training, utilizing a depth regression model to model a mapping relation from the explicit characteristics to the implicit characteristics; step 3, calculating the implicit characteristic of the cold start task according to the explicit characteristic of the cold start task based on the mapping relation; step 4, according to the explicit characteristics of the cold start task and the implicit characteristics of the cold start task obtained through calculation, the FM fusion model is used for carrying out scoring prediction; wherein, the implicit characteristic refers to: and one-hot of the task ID represents the implicit factor vector corresponding to the one-dimensional feature with the numerical value of 1 in the fusion model.
Further, the fusion model of the FM in step 1 adopts a second-order FM model to perform fusion of heterogeneous information; the existing tasks and the explicit characteristics and the ID characteristics of the existing developers are used as the input characteristics of the regression problem, the scoring information is used as the regression variable, and the scenes are recommended to the developers, wherein the expression is as follows:
wherein the model parameter w0，wiAnd wijRespectively representing global bias, weight corresponding to the feature i and weight of an interaction item of the feature i and the feature j, n and m respectively representing the number of existing tasks and developers, and n1And m1Explicit feature dimensions representing existing tasks and developers, respectively; interaction term weight wijExpressed as:
viand vjRespectively represent the feature xiAnd xjThe corresponding hidden factor vector.
Further, in step 1, in order to train the FM-based fusion model, a loss function is defined based on a sum of squares error, and the following optimization problem is solved:
wherein D represents a training set,
λwand λvAnd respectively representing regularization coefficients corresponding to the three types of model parameters.
Furthermore, the stacked noise reduction self-encoder consists of a plurality of noise reduction self-encoders, and the feature dimension after dimension reduction is set to be gradually reduced so as to obtain a high-level feature with a low dimension; the learning output of one noise-reducing autoencoder is used as the input to train the next noise-reducing autoencoder.
Further, the linear regression unit obtains a predicted value of the implicit feature of the task i (i is 1, …, n) by weighted summation of the high-level features after dimensionality reduction
Where l is the dimension of the high level feature, wjIs the weight of the j-th dimension feature, w0Is an offset; the linear regression unit correlation weight coefficients are then determined by solving the following optimization problem by gradient descent:
wherein v isiFor the implicit characteristic value of the task i,
the implicit characteristic value of the task i is predicted by the depth regression model, and n is the number of the existing tasks.
Further, in the whole depth regression model, the stacked noise reduction self-encoder is initialized by using pre-trained weights, and an outer layer linear regression unit is initialized randomly; all weights are fine-tuned using a back-propagation algorithm.
Compared with the prior art, the invention has the advantages and positive effects that: the developer recommendation method of the cold start task, provided by the invention, comprises the steps of firstly fusing the explicit characteristics, the ID characteristics and the corresponding grading information of the existing task and the existing developer on a training set by utilizing an FM (frequency modulation) algorithm, and establishing an FM-based fusion model; however, in the process of fusion (training), only the implicit features of the known tasks can be obtained based on the data of the known tasks (the implicit features refer to the implicit factor vectors corresponding to the one-dimensional features with the numerical value of 1 in the one-hot representation of the known task ID in the fusion model), but the implicit features of the cold start tasks cannot be obtained; then, based on the mapping relation, calculating the implicit characteristic of the cold start task according to the explicit characteristic of the cold start task; and finally, according to the explicit characteristics of the cold start task and the implicit characteristics of the cold start task obtained through calculation, carrying out scoring prediction by using a trained FM fusion model. The recommendation method comprehensively considers the explicit characteristics and the ID characteristics of the task aiming at the cold start problem, solves the problem that the FM model lacks hidden factor vectors corresponding to the dimension of the ID of the cold start task by constructing the mapping relation from the display characteristics to the implicit characteristics, has a better recommendation effect compared with the prior method for recommending by only utilizing the display characteristics aiming at the cold start problem, and has obvious advantages in accuracy and recall rate.
Other features and advantages of the present invention will become more apparent from the detailed description of the embodiments of the present invention when taken in conjunction with the accompanying drawings.
Drawings
FIG. 1 is a method flow diagram of a developer recommendation method for cold start tasks in accordance with the present invention;
FIG. 2 is a computational model illustration of a developer recommendation method for cold start tasks in accordance with the present invention;
FIG. 3 is a schematic diagram of a training process of a stacked noise reduction self-encoder in the present invention on the existing task explicit feature data;
FIG. 4 is a schematic diagram of a deep regression model trained based on explicit and implicit features of existing tasks in the present invention;
FIG. 5 is a diagram illustrating a comparison result between the recommendation method and the comparison algorithm of the present invention.
Detailed Description
The following describes embodiments of the present invention in further detail with reference to the accompanying drawings.
With reference to fig. 1 and fig. 2, the developer recommendation method for the cold start task provided by the present invention is implemented by the following steps:
step S1: and fusing the explicit characteristics, the ID characteristics and the corresponding grading information of the existing tasks and the existing developers by utilizing an FM algorithm to establish an FM-based fusion model.
In the embodiment of the invention, the FM fusion model 1 adopts a second-order FM model to perform fusion of heterogeneous information, the explicit characteristics and the ID characteristics of the existing tasks and the existing developers are used as the input characteristics of the regression problem, the scoring information is used as the regression variable, and the expression is as follows aiming at the recommended scenes of the developers:
wherein the model parameter w0，wjAnd wijRespectively representing global bias, weight corresponding to the feature i and weight of an interaction item of the feature i and the feature j, n and m respectively representing the number of existing tasks and developers, and n1And m1Respectively, represent the explicit feature dimensions of the existing task and developer.
In the above FM model, the sparsity problem is alleviated by decomposing the weight of the interaction term into the product of implicit factor vectors, the interaction term x in formula (1)iAnd xjWeight w ofijCan be expressed as:
wherein v isiAnd vjRespectively represent the feature xiAnd xjCorresponding implicit factor directionAn amount; the dimensions of the hidden factor vector are typically specified manually by the user.
To train the FM regression model, a loss function is defined based on the sum of squares error and the following optimization problem is solved using stochastic gradient descent:
wherein D represents a training set,
λwand λvAnd respectively representing regularization coefficients corresponding to the three types of model parameters.
In a specific embodiment of the present invention, the input features are composed of a task ID, a task display feature, a developer ID, and a developer display feature, where the task ID and the developer ID are represented in a one-hot coding form (a 0,1 binary vector, where only one bit is 1 and the other bits are 0), for example, if there are 100 tasks, the task with ID number 1 is represented as 10000 … … 0 (99 0 bits after 1), the task with ID number 2 is represented as 01000 … … 0 (98 0 bits after 1), that is, only the corresponding position of the number is 1, and the rest bits are 0.
In the FM training process, when encoding the task ID (number) in the training set, the number of the cold start task to be predicted needs to be considered. For example, if the number of finished tasks is 99 (numbered 1 to 99) and the number of cold start task 1 to be predicted is 100, the number of finished tasks needs to be encoded by 100-dimensional one-hot encoding, but only 99-dimensional one-hot encoding cannot be adopted. If the number of the completed task is coded by only 99-dimensional one-hot coding, the task with the number of 100 cannot be coded in the prediction stage, so the trained FM model cannot process the cold start task.
In addition, for the solution of the optimization problem (3), since the cold start task (task number 100) has no score data, the implicit factor vector corresponding to the dimension with the value of 1 in the one-hot encoding of the cold start task (task number 100) cannot be obtained by training the FM model. This is because in training the FM fusion model, it is necessary to first assign an initial value to the unknowns in equation (3), and then perform iterative optimization based on a random gradient descent. However, as can be seen from the solution principle of stochastic gradient descent, only the unknowns associated with the training samples can be optimized. Therefore, as for the unknown quantity of the hidden factor vector of the one-dimensional feature (namely, the 100-dimensional feature) with the value of 1 in the one-hot coding of the cold start task ID, the unknown quantity has no corresponding training sample because no score is found on the cold start task, and therefore the solution cannot be carried out through iteration.
For convenience of description, in the embodiment of the present invention, the implicit factor vector corresponding to the one-dimensional feature with a value of 1 in the one-hot representation of the task ID in the FM fusion model is defined as the implicit feature of the task, so based on the analysis, in the FM fusion model training, since the cold start task has no score data, we cannot find the implicit feature corresponding to the cold start task.
Therefore, in order to predict the score of the cold start task for the developer by using the FM fusion model, the implicit characteristic of the cold start task needs to be calculated, and the method is implemented by adopting the following steps 2 to 3.
Step S2: and based on the implicit characteristics of the finished task obtained by training, utilizing a deep regression model to model the mapping relation from the explicit characteristics to the implicit characteristics.
In the embodiment of the invention, the mapping relation is modeled by combining a stacked noise reduction self-encoder and based on a depth regression model, and the specific implementation steps comprise:
1. reducing dimension of explicit characteristics based on a stack type noise reduction self-encoder: representing original explicit characteristics of tasks on a training set as low-dimensional high-level characteristics by using a stacked noise reduction self-encoder;
2. linear regression analysis: taking low-dimensional high-level features as input and implicit features of tasks on a training set as output to construct a linear regression model;
the two processes are completed in a deep neural network, namely a layer of linear regression unit is added on the outer layer of the stacked noise reduction self-encoder.
Specifically, as shown in fig. 3, in the training process of the stacked noise reduction self-encoder, the stacked noise reduction self-encoder is composed of a plurality of (3 in the figure) noise reduction self-encoders (DAE), and k is set1＞k2＞k3＞k4I.e., the dimensionality of the reduced features gradually decreases, the learning output of one DAE is used as an input to train the next DAE by training the DAE layer-by-layer to obtain the high-level features of the lower dimension. And then, expanding the trained DAE to construct an initial stacked noise reduction self-encoder, and performing fine adjustment on the stacked noise reduction self-encoder on a training set by using a back propagation algorithm to obtain the trained stacked noise reduction self-encoder.
As shown in fig. 4, in the linear regression analysis, a layer of conventional linear regression model is added to the outer layer of the stacked noise reduction self-encoder network, and the predicted value of the implicit feature of task i (i ═ 1, …, n) is obtained by weighted summation of the high-layer features after dimensionality reduction
Where l is the dimension of the high level feature, wjIs the weight of the j-th dimension feature, w0Is an offset; the linear regression unit correlation weight coefficients are then determined by solving the following optimization problem by gradient descent:
wherein v isiFor the implicit characteristic value of the task i,
the implicit characteristic value of the task i is predicted by the depth regression model, and n is the number of the existing tasks.
In the training process of the depth regression model of the embodiment of the invention, the stack type noise reduction self-encoder is initialized by using the pre-trained weights, the outermost layer traditional regression model is initialized randomly, and then all the weights are finely adjusted by using a back propagation algorithm, so that the final depth regression model is obtained.
Step S3: and calculating the implicit characteristics of the cold start task according to the explicit characteristics of the cold start task based on the mapping relation.
Step S4: and carrying out scoring prediction by utilizing an FM fusion model according to the explicit characteristics of the cold start task and the implicit characteristics of the cold start task obtained by calculation.
In a specific embodiment of the invention, let E be the explicit feature of the cold start task iiSetting the mapping relation from the explicit characteristic to the implicit characteristic as F1Then, according to the mapping relation, the implicit characteristic v of the cold start task i can be knowniIs shown as vi＝F1(Ei) (ii) a According to the definition, the implicit feature of the cold start task i is the implicit factor vector of the one-dimensional feature with the value of 1 in the one-hot coding of the cold start task i, so that the weights of all the interactive terms related to the one-dimensional feature with the value of 1 in the one-hot coding can be obtained according to the formula (2).
Further, the score of the cold start task on the existing developer can be obtained by using the obtained FM model according to the ID feature and the explicit feature of the cold start task and the ID feature and the explicit feature of the existing developer.
The developer recommendation method for the cold start task, provided by the invention, is oriented to the cold start scene on the crowdsourced software platform, the explicit characteristics and the ID characteristics are comprehensively considered, the cold start problem is effectively solved, and as shown in FIG. 5, the recommendation algorithm has obvious advantages in accuracy and recall rate through wide comparison experiments carried out on a topcoder crowdsourced software development platform data set. The MRFMRec curve is a developer recommendation result for cold starting proposed according to the invention, FM is a developer recommendation result for cold starting under a traditional display characteristic algorithm only considered, top2, top4 and top6 respectively represent the number of recommended developers, TR20, TR40, TR60 and TR80 respectively represent different training sets, and obviously the effect of the developer recommendation method for cold starting proposed according to the invention is better based on the principle that the higher the accuracy is, the better the recommendation is, and the higher the recall rate is, the better the recommendation is.
It should be noted that the above description is not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art should also make changes, modifications, additions or substitutions within the spirit and scope of the present invention.
Claims (6)
1. A developer recommendation method for cold start tasks is characterized by comprising the following steps:
step 1, fusing the explicit characteristics, the ID characteristics and the corresponding grading information of the existing tasks and the existing developers by utilizing an FM algorithm, and establishing an FM-based fusion model;
step 2, based on the implicit characteristics of the completed task obtained by training, utilizing a depth regression model to model a mapping relation from the explicit characteristics to the implicit characteristics;
step 3, calculating the implicit characteristic of the cold start task according to the explicit characteristic of the cold start task based on the mapping relation;
step 4, according to the explicit characteristics of the cold start task and the implicit characteristics of the cold start task obtained through calculation, the FM fusion model is used for carrying out scoring prediction;
wherein, the implicit characteristic refers to: and one-hot of the task ID represents the implicit factor vector corresponding to the one-dimensional feature with the numerical value of 1 in the fusion model.
2. The developer recommendation method for cold start task according to claim 1, wherein the fusion model of FM in step 1 is a second-order FM model for fusion of heterogeneous information; the existing tasks and the explicit characteristics and the ID characteristics of the existing developers are used as the input characteristics of the regression problem, the scoring information is used as the regression variable, and the scenes are recommended to the developers, wherein the expression is as follows:
wherein the model parameter w0，wiAnd wijRespectively representing global bias, featuresi and i are corresponding to the weight and the weight of the interaction items of the characteristics i and j, n and m respectively represent the number of the existing tasks and developers, and n1And m1Explicit feature dimensions representing existing tasks and developers, respectively; interaction term weight wijExpressed as:
viand vjRespectively represent the feature xiAnd xjThe corresponding hidden factor vector.
3. The developer recommendation method for cold start task according to claim 2, wherein in step 1, for training the FM-based fusion model, a loss function is defined based on the sum of squares error, and the following optimization problem is solved:
wherein D represents a training set,
λwand λvAnd respectively representing regularization coefficients corresponding to the three types of model parameters.
4. The developer recommendation method for a cold start task according to claim 1, wherein in the step 2, a mapping relationship modeling based on a depth regression model is performed in combination with a stacked noise reduction self-encoder, and specifically includes:
reducing dimension of explicit characteristics based on a stack type noise reduction self-encoder: representing original explicit characteristics of tasks on a training set as low-dimensional high-level characteristics by using the stacked noise reduction self-encoder;
linear regression analysis: and adding a layer of linear regression unit on the outer layer of the stacked noise reduction self-encoder, taking the low-dimensional high-level features as input, taking the implicit features of the tasks on the training set as output, and constructing a linear regression model.
5. The developer recommendation method for cold start task according to claim 4, wherein the linear regression unit obtains the predicted value of the implicit feature of task i (i-1, …, n) by weighted summation of the high-level features after dimensionality reduction
Where l is the dimension of the high level feature, wjIs the weight of the j-th dimension feature, w0Is an offset; the linear regression unit correlation weight coefficients are then determined by solving the following optimization problem by gradient descent:
wherein v isiFor the implicit characteristic value of the task i,
the implicit characteristic value of the task i is predicted by the depth regression model, and n is the number of the existing tasks.
6. The developer recommendation method for cold start task according to claim 4, wherein in the deep regression model, the stacked noise reduction self-encoder is initialized by using pre-trained weights, and the outer linear regression unit is initialized randomly; all weights are fine-tuned using a back-propagation algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010538552.3A CN112529350B (en) | 2020-06-13 | 2020-06-13 | Developer recommendation method for cold start task |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010538552.3A CN112529350B (en) | 2020-06-13 | 2020-06-13 | Developer recommendation method for cold start task |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112529350A true CN112529350A (en) | 2021-03-19 |
| CN112529350B CN112529350B (en) | 2022-10-18 |
Family
ID=74978702
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010538552.3A Active CN112529350B (en) | 2020-06-13 | 2020-06-13 | Developer recommendation method for cold start task |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112529350B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114139052A (en) * | 2021-11-19 | 2022-03-04 | 北京百度网讯科技有限公司 | Ranking model training method for intelligent recommendation, intelligent recommendation method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3179434A1 (en) * | 2015-12-10 | 2017-06-14 | Deutsche Telekom AG | Designing context-aware recommendation systems, based on latent contexts |
| CN107256508A (en) * | 2017-05-27 | 2017-10-17 | 上海交通大学 | Commercial product recommending system and its method based on Novel Temporal Scenario |
| CN108960293A (en) * | 2018-06-12 | 2018-12-07 | 玩咖欢聚文化传媒（北京）有限公司 | CTR predictor method and system based on FM algorithm |
| CN110348968A (en) * | 2019-07-15 | 2019-10-18 | 辽宁工程技术大学 | A kind of recommender system and method analyzed based on user and project coupled relation |
| 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 |
-
2020
- 2020-06-13 CN CN202010538552.3A patent/CN112529350B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3179434A1 (en) * | 2015-12-10 | 2017-06-14 | Deutsche Telekom AG | Designing context-aware recommendation systems, based on latent contexts |
| CN107256508A (en) * | 2017-05-27 | 2017-10-17 | 上海交通大学 | Commercial product recommending system and its method based on Novel Temporal Scenario |
| CN108960293A (en) * | 2018-06-12 | 2018-12-07 | 玩咖欢聚文化传媒（北京）有限公司 | CTR predictor method and system based on FM algorithm |
| CN110348968A (en) * | 2019-07-15 | 2019-10-18 | 辽宁工程技术大学 | A kind of recommender system and method analyzed based on user and project coupled relation |
| 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 |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114139052A (en) * | 2021-11-19 | 2022-03-04 | 北京百度网讯科技有限公司 | Ranking model training method for intelligent recommendation, intelligent recommendation method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112529350B (en) | 2022-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111428147B (en) | Social recommendation method of heterogeneous graph volume network combining social and interest information | |
| CN111611472B (en) | Binding recommendation method and system based on graph convolution neural network | |
| CN111222054B (en) | Session social contact recommendation method based on context neighbor relation modeling | |
| Hassan et al. | Genetic algorithm approaches for improving prediction accuracy of multi-criteria recommender systems | |
| CN108876044B (en) | Online content popularity prediction method based on knowledge-enhanced neural network | |
| CN111563770A (en) | Click rate estimation method based on feature differentiation learning | |
| CN113343119A (en) | Group recommendation method based on hierarchical attention mechanism | |
| CN112819523B (en) | Marketing prediction method combining inner/outer product feature interaction and Bayesian neural network | |
| CN114817663A (en) | Service modeling and recommendation method based on class perception graph neural network | |
| CN113239897A (en) | Human body action evaluation method based on space-time feature combination regression | |
| CN113255908A (en) | Method, neural network model and device for service prediction based on event sequence | |
| CN112529350B (en) | Developer recommendation method for cold start task | |
| CN114817773A (en) | Time sequence prediction system and method based on multi-stage decomposition and fusion | |
| Chen et al. | Combining fuzzy iteration model with dynamic programming to solve multiobjective multistage decision making problems | |
| CN114238765A (en) | Block chain-based position attention recommendation method | |
| CN113821724A (en) | Graph neural network recommendation method based on time interval enhancement | |
| CN115935067A (en) | Article recommendation method integrating semantics and structural view for socialized recommendation | |
| CN115438842A (en) | Load prediction method based on adaptive improved dayflies and BP neural network | |
| CN114363671A (en) | Multimedia resource pushing method, model training method, device and storage medium | |
| CN113010774A (en) | Click rate prediction method based on dynamic deep attention model | |
| CN112487799A (en) | Crowdsourcing task recommendation algorithm using extrinsic product attention | |
| Chaveesuk et al. | Economic valuation of capital projects using neural network metamodels | |
| CN111709275A (en) | Deep network construction method for Affordance reasoning | |
| CN116485501B (en) | Graph neural network session recommendation method based on graph embedding and attention mechanism | |
| CN111612573B (en) | Recommendation system scoring recommendation prediction method based on full Bayesian method |
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 |