CN113486240B - Position recommendation method based on SWPEM routing algorithm - Google Patents
Position recommendation method based on SWPEM routing algorithm Download PDFInfo
- Publication number
- CN113486240B CN113486240B CN202110770959.3A CN202110770959A CN113486240B CN 113486240 B CN113486240 B CN 113486240B CN 202110770959 A CN202110770959 A CN 202110770959A CN 113486240 B CN113486240 B CN 113486240B
- Authority
- CN
- China
- Prior art keywords
- user
- attention
- job
- dimensional
- capsules
- 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 18
- 239000013598 vector Substances 0.000 claims abstract description 37
- 230000007246 mechanism Effects 0.000 claims abstract description 27
- 239000002775 capsule Substances 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims description 3
- 238000011524 similarity measure Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims 1
- 238000000605 extraction Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000007115 recruitment Effects 0.000 description 2
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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
- G06F16/906—Clustering; Classification
-
- 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
- G06Q10/1053—Employment or hiring
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a job recommendation method based on SWP EM routing algorithm, comprising the following steps: coding the high-dimensional sparse features corresponding to the user information and the position browsing history data, and converting the high-dimensional sparse features into low-dimensional dense vectors; inputting the low-dimensional dense vector into a feature clustering layer, and clustering position features in the low-dimensional dense vector by using a GMM to obtain a plurality of clustering clusters; calculating the similarity between the cluster vector and the position vector, and distributing attention weights to different candidate positions according to the similarity; according to the attention weight of the user to each position, the capsules expected by different positions of the user are arranged in a reverse order, and the attention mechanism is utilized to extract the capsules expected by different positions of the user; and outputting actual recommendation positions converted by the plurality of expected capsules by using a softmax function, and returning the recommendation position combination corresponding to the maximum value of the attention weights of the plurality of clusters sequenced by the user to the user. By applying the embodiment of the invention, the problem of weight distribution of a disturbed attention mechanism of an overlong position browsing history sequence is solved.
Description
Technical Field
The invention relates to the technical field of deep learning, in particular to a job recommendation method based on a SWP EM routing algorithm.
Background
The rapid development of the internet has led people to enter an era of complex information, such as thousands of recruitment information in a human resource recruitment requirement scene, and the requirements of interviewees on position information become diversified and personalized, that is, interviewees may not be able to search for valuable positions from massive position information. In order to better provide services for users, the personalized recommendation technology is generated to assist the users to find positions which truly meet the needs of the users.
For research of recommendation technology, early recommendation systems are based on collaborative filtering, later develop into machine learning, and finally deep learning formally enter the field of recommendation systems. Currently, the mainstream recommendation method is to use an attention mechanism to extract recommendations for different positions of the user that are desired for position combinations or sequences. The positive effects of the attention mechanism on recommended tasks have been well documented. However, as the tasks that the recommendation model needs to handle become increasingly diverse and personalized, traditional attention mechanisms have failed to meet the needs of the recommendation model.
Currently, the recommendation model fusing traditional attention mechanisms mainly faces several inherent problems:
1. lack of multi-target extraction: extraction of the user's different job expectations is critical to reasonable recommendations, for example, in resume delivery, an interviewer may interview multiple companies in order to increase the probability of taking a recording notice by himself, and may deliver multiple relevant posts of his own desired job, and the expression capability of a single job embedding vector is insufficient, often giving unreasonable recommendations, and thus multiple vectors are needed to represent the user's different job expectations.
2. Distraction under very long behavioural sequences: there is a relationship between a plurality of job combinations, and we generally consider that the longer the user job browsing history sequence is, the more clearly the job expectations of the user can be described. However, an excessively long post browsing history sequence may disturb the weight allocation of the attention mechanism, possibly giving a higher attention weight to the undesired post, affecting the accuracy of post recommendation.
3. Feature confusion in high dimensions: the use of high-dimensional vectors to abstract represent the user's job expectations makes it difficult to capture similarities between job positions, resulting in job position feature expression that is too generalized and thus loses focus.
Disclosure of Invention
Compared with the prior art, the position recommending method based on the SWP EM routing algorithm can respectively recommend different expected positions, and under the condition of multi-target extraction, the position recommending method mechanism combined with the SWP EM routing algorithm improves recommending accuracy and multiple elements, can solve the problem of overlarge user quantity, and improves user experience.
In order to achieve the above object, the present invention provides a job recommendation method based on SWP EM routing algorithm, including:
encoding the high-dimensional sparse features corresponding to the user information and the position browsing history data to convert the high-dimensional sparse features into low-dimensional dense vectors;
inputting the low-dimensional dense vector into a feature clustering layer, and clustering position features in the low-dimensional dense vector by using a GMM to obtain a plurality of clustering clusters;
parallelly calculating the similarity between the cluster vectors and the position vectors through a SWP EM routing algorithm, and distributing attention weights to different candidate positions according to the similarity;
according to the attention weight of the user to each position, the capsules expected by different positions of the user are arranged in an inverted order and extracted by an attention mechanism;
and outputting actual recommendation positions converted by the plurality of expected capsules by using a softmax function, and returning the recommendation position combination corresponding to the maximum value of the attention weights of the plurality of clusters sequenced by the user to the user.
Optionally, the step of encoding the high-dimensional sparse feature corresponding to the user information and the position browsing history data to be converted into the low-dimensional dense vector includes:
the high-dimensional sparse features are preliminarily encoded by the Doc2Vec algorithm to be converted into low-dimensional dense vectors by the aid of user information at least comprising gender, age and occupation and a job browsing history sequence at least comprising job types, month salaries, academic requirements and corporate properties.
In one implementation, the step of arranging in reverse order according to the attention weight of the user to each position and extracting capsules expected by different positions of the user by using an attention mechanism includes:
the capsules are arranged in a reverse order according to the attention weight of the user to each position, and the attention mechanism is utilized to extract capsules expected by different positions of the user;
if the desired feature is not included in the training data, the score is further parameterized by an attention mechanism and weighted summed to obtain the final desired capsule.
By applying the position recommendation method based on the SWP EM routing algorithm, which is provided by the embodiment of the invention, the high-dimensional characteristics of the user browsing history sequence can be fully mined, and an improved attention mechanism is integrated to separate a plurality of position expectations of the user; the similarity among positions can be captured, the problem of weight distribution of a disturbed attention mechanism of an overlong position browsing history sequence can be solved, and the higher attention weight is given to an unexpected position in a balanced manner.
Drawings
Fig. 1 is a schematic flow chart of a job recommendation method based on a SWP EM routing algorithm according to an embodiment of the present invention.
Detailed Description
Other advantages and effects of the present invention will become apparent to those skilled in the art from the following disclosure, which describes the embodiments of the present invention with reference to specific examples. The invention may be practiced or carried out in other embodiments that depart from the specific details, and the details of the present description may be modified or varied from the spirit and scope of the present invention.
The invention provides a job recommendation method based on SWP EM routing algorithm, which comprises the following steps:
s110, coding high-dimensional sparse features corresponding to the user information and the position browsing history data to convert the high-dimensional sparse features into low-dimensional dense vectors;
it should be noted that, 12 items of user information such as gender, age, occupation, etc. and 18 items of job browsing history sequences such as job type, moon salary, academic requirement, company property, etc. are preliminarily encoded in the feature encoding layer by the Doc2Vec algorithm, so that the feature encoding layer is converted into a computable and structured low-dimensional dense vector to be used as the input of the feature clustering layer.
S120, inputting the low-dimensional dense vector into a feature clustering layer, and clustering position features in the low-dimensional dense vector by using a GMM to obtain a plurality of cluster clusters;
it can be appreciated that after the obtained low-dimensional dense vector is input to the feature clustering layer, different job features in the GMM cluster coding information are used and grouped into a plurality of cluster clusters.
S130, parallelly calculating the similarity between the cluster vectors and the position vectors through a SWP EM routing algorithm, and distributing attention weights to different candidate positions according to the similarity;
it should be noted that, the similarity in the present invention refers to an attention value, and calculating the similarity is an important process of obtaining an attention weight. And calculating the similarity between the cluster vector and the position vector, namely calculating the attention value, and carrying out normalization processing on the calculated original score to obtain the attention weight. The attention mechanism is a weight parameter allocation mechanism, and aims to assist the model in capturing important information. The similarity measure is used to determine the degree of equality between the two job vectors, and computing the similarity better understands the degree of closeness between the job. And attention should be paid to similar positions, which are worth recommending.
Specifically, according to the obtained clusters, similarity between cluster vectors and position vectors is calculated in parallel through a SWP EM routing algorithm, attention weights are obtained after normalization processing is carried out on the calculated attention values, then the attention weights are distributed to different candidate positions, the SWP EM routing algorithm is delivered to a Spark distributed computing platform to accelerate computing, for the recommended service updated once a day, a recommendation result is calculated in advance through Spark, and the recommendation result is stored for front-end service calling. The attention mechanism of a matrix sharing version (all parameters of a sharing weight matrix) is added into an original EM routing algorithm to replace the attention mechanism of an original full-connection version, so that the position browsing history sequence with an indefinite length of a user is processed, and all high-level capsules are mapped to the feature space with the same dimension.
S140, arranging the capsules in a reverse order according to the attention weight of the user to each position, and extracting the capsules expected by different positions of the user by using an attention mechanism;
it can be understood that the attention weights of the users to each position are arranged in an inverted order in the induction evaluation layer, and the attention mechanisms are utilized to extract the expected capsules of different positions of the users; for the desired features that never appear in the training data, the attention score for their interaction cannot be estimated, the invention uses the attention mechanism to further parameterize the score, and then performs a weighted summation to arrive at the final desired capsule.
The invention processes the history sequence of position browsing of the user with indefinite length and maps all the high-level capsules to the feature space problem of the same dimension, and utilizes the information entropy to express the significance degree of the position feature, thereby fully explaining the interrelationship between the behavior capsule and the expected capsule.
And S150, outputting actual recommendation positions converted by the plurality of expected capsules by using a softmax function, and returning the recommendation position combination corresponding to the maximum value of the attention weights of the plurality of clusters sequenced by the user to the user.
The method and the device output the actual recommendation positions converted by the plurality of expected capsules by using the softmax function, and return the recommendation position combination corresponding to the maximum value of the attention weights of the N clustering clusters of the user to the user, so that the position is recommended to the user, the accuracy and the multiple of the recommendation are improved, the problem of overlarge user quantity can be solved, and the user experience is improved.
The above embodiments are merely illustrative of the principles of the present invention and its effectiveness, and are not intended to limit the invention. Modifications and variations may be made to the above-described embodiments by those skilled in the art without departing from the spirit and scope of the invention. Accordingly, it is intended that all equivalent modifications and variations of the invention be covered by the claims, which are within the ordinary skill of the art, be within the spirit and scope of the present disclosure.
Claims (3)
1. The utility recommendation method based on the SWP EM routing algorithm is characterized by comprising the following steps of:
encoding the high-dimensional sparse features corresponding to the user information and the position browsing history data to convert the high-dimensional sparse features into low-dimensional dense vectors;
inputting the low-dimensional dense vector into a feature clustering layer, and clustering position features in the low-dimensional dense vector by using a GMM to obtain a plurality of clustering clusters;
parallelly calculating the similarity between the cluster vectors and the position vectors through a SWP EM routing algorithm, and distributing attention weights to different candidate positions according to the similarity;
the method comprises the steps of calculating similarity between cluster vectors and position vectors in parallel through a SWP EM routing algorithm, normalizing the calculated attention value to obtain attention weight, and then distributing the attention weight to different candidate positions, wherein the SWP EM routing algorithm is delivered to a Spark distributed computing platform to accelerate calculation; the attention mechanism is an allocation mechanism of weight parameters, and the similarity measure is used for determining the equality degree between two position vectors;
according to the attention weight of the user to each position, the capsules expected by different positions of the user are arranged in an inverted order and extracted by an attention mechanism;
and outputting actual recommendation positions converted by the plurality of expected capsules by using a softmax function, and returning the recommendation position combination corresponding to the maximum value of the attention weights of the plurality of clusters sequenced by the user to the user.
2. The job recommendation method based on SWP EM routing algorithm of claim 1, wherein the step of encoding the high-dimensional sparse features corresponding to the user information and the job browsing history data to be converted into a low-dimensional dense vector comprises:
the high-dimensional sparse features are preliminarily encoded by the Doc2Vec algorithm to be converted into low-dimensional dense vectors by the aid of user information at least comprising gender, age and occupation and a job browsing history sequence at least comprising job types, month salaries, academic requirements and corporate properties.
3. The job recommendation method based on SWP EM routing algorithm according to claim 1, wherein the step of arranging in reverse order according to the attention weight of the user to each job and extracting the capsules desired by the different job of the user by using the attention mechanism comprises the steps of:
the capsules are arranged in a reverse order according to the attention weight of the user to each position, and the attention mechanism is utilized to extract capsules expected by different positions of the user;
if the desired feature is not included in the training data, the score is further parameterized by an attention mechanism and weighted summed to obtain the final desired capsule.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110770959.3A CN113486240B (en) | 2021-07-08 | 2021-07-08 | Position recommendation method based on SWPEM routing algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110770959.3A CN113486240B (en) | 2021-07-08 | 2021-07-08 | Position recommendation method based on SWPEM routing algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113486240A CN113486240A (en) | 2021-10-08 |
| CN113486240B true CN113486240B (en) | 2024-02-02 |
Family
ID=77937446
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110770959.3A Active CN113486240B (en) | 2021-07-08 | 2021-07-08 | Position recommendation method based on SWPEM routing algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113486240B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102184364A (en) * | 2011-05-26 | 2011-09-14 | 南京财经大学 | Semi-supervised learning-based recommendation system shilling attack detection method |
| US8364559B1 (en) * | 2010-01-07 | 2013-01-29 | Amazon Technologies, Inc. | Method, medium, and system of recommending a substitute item |
| CN112364236A (en) * | 2019-07-24 | 2021-02-12 | 阿里巴巴集团控股有限公司 | Target object recommendation system, method and device, and data processing method and device |
| CN112699215A (en) * | 2020-12-24 | 2021-04-23 | 齐鲁工业大学 | Grading prediction method and system based on capsule network and interactive attention mechanism |
| CN112765461A (en) * | 2021-01-12 | 2021-05-07 | 中国计量大学 | Session recommendation method based on multi-interest capsule network |
-
2021
- 2021-07-08 CN CN202110770959.3A patent/CN113486240B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8364559B1 (en) * | 2010-01-07 | 2013-01-29 | Amazon Technologies, Inc. | Method, medium, and system of recommending a substitute item |
| CN102184364A (en) * | 2011-05-26 | 2011-09-14 | 南京财经大学 | Semi-supervised learning-based recommendation system shilling attack detection method |
| CN112364236A (en) * | 2019-07-24 | 2021-02-12 | 阿里巴巴集团控股有限公司 | Target object recommendation system, method and device, and data processing method and device |
| CN112699215A (en) * | 2020-12-24 | 2021-04-23 | 齐鲁工业大学 | Grading prediction method and system based on capsule network and interactive attention mechanism |
| CN112765461A (en) * | 2021-01-12 | 2021-05-07 | 中国计量大学 | Session recommendation method based on multi-interest capsule network |
Non-Patent Citations (3)
| Title |
|---|
| Twitter Social Networking Graph using Hypernym based Semantic Similarity Detection;Sashank Sridhar 等;《2020 International Conference on Smart Electronics and Communication (ICOSEC)》;28-35 * |
| 一种基于EM路由算法的推荐模型;常思维;《中国优秀硕士学位论文全文数据库信息科技辑》(第4期);I138-1320 * |
| 基于深度学习的中文网络招聘文本中的技能词抽取方法;文益民 等;《桂林电子科技大学学报》;第40卷(第4期);338-348 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113486240A (en) | 2021-10-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112785397B (en) | Product recommendation method, device and storage medium | |
| CN113254803B (en) | Social recommendation method based on multi-feature heterogeneous graph neural network | |
| Khan et al. | Multi-gcn: Graph convolutional networks for multi-view networks, with applications to global poverty | |
| Wang et al. | Kvt: k-nn attention for boosting vision transformers | |
| Tsang et al. | Feature interaction interpretability: A case for explaining ad-recommendation systems via neural interaction detection | |
| CN107066616A (en) | Method, device and electronic equipment for account processing | |
| CN108009148A (en) | Text emotion classification method for expressing based on deep learning | |
| Ma et al. | Spatial pyramid attention for deep convolutional neural networks | |
| CN114817663A (en) | Service modeling and recommendation method based on class perception graph neural network | |
| Shamsolmoali et al. | High-dimensional multimedia classification using deep CNN and extended residual units | |
| Ma et al. | Error correcting input and output hashing | |
| CN111538846A (en) | Third-party library recommendation method based on mixed collaborative filtering | |
| CN114491084B (en) | Self-encoder-based relation network information mining method, device and equipment | |
| Wang et al. | An enhanced multi-modal recommendation based on alternate training with knowledge graph representation | |
| CN113486240B (en) | Position recommendation method based on SWPEM routing algorithm | |
| CN117557331A (en) | Product recommendation method and device, computer equipment and storage medium | |
| Kapoor | Data mining: Past, present and future scenario | |
| CN116340516A (en) | Entity relation cluster extraction method, device, equipment and storage medium | |
| CN115983982A (en) | Credit risk identification method, credit risk identification device, credit risk identification equipment and computer readable storage medium | |
| CN112734519B (en) | Commodity recommendation method based on convolution self-encoder network | |
| Zhang et al. | Improved image retrieval algorithm of GoogLeNet neural network | |
| Zin et al. | Deep learning model for integration of clustering with ranking in social networks | |
| Jain et al. | Telecom churn prediction using CNN with variational autoencoder | |
| Guillermo et al. | Content-based Fashion Recommender System Using Unsupervised Learning | |
| Ranjith et al. | An Effective Content Based Image Retrieval System Using Deep Learning Based Inception Model |
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 |