Position recommendation method based on collaborative filtering algorithm

Abstract

The invention discloses a position recommendation method based on a collaborative filtering algorithm, and relates to the technical field of data analysis. The method comprises the steps of 1, collecting and preprocessing data, 2, constructing a user-position scoring matrix: constructing the user-position scoring matrix according to historical behavior data of users, the rows of the user-position scoring matrix representing users and the columns of the user-position scoring matrix representing positions, elements in the matrix representing association scores of the users to the positions, and elements in the matrix representing association scores of the users to the positions; the method comprises the following steps of 1, selecting a total score of the positions, 2, performing collaborative filtering on the job seekers and collaborative filtering on the positions on the basis of a collaborative filtering algorithm, 4, sequencing the positions according to the total score of the positions, generating a position list and recommending the position list to a target user, and 5, collecting feedback of the user on a position recommendation result for a subsequent optimization recommendation process.

Description

Technical Field

[0001] The invention discloses a job recommendation method based on a collaborative filtering algorithm, and relates to the technical field of data analysis. Background Art

[0002] With the development of Internet technology, there are more and more online recruitment platforms, and a large amount of job information and job seeker information has flooded into the network. Users are faced with a large amount of job information when looking for jobs, making it difficult for job seekers to quickly find a suitable job, and it is also difficult for recruiters to screen out suitable candidates from a large number of job seekers. Existing job recommendation methods are often based on keyword matching, ignoring the similarity between users and the correlation between jobs, and cannot accurately match user needs. Summary of the invention

[0003] In view of the problems in the prior art, the present invention provides a job recommendation method based on a collaborative filtering algorithm. The method obtains historical behavior data of users, uses a collaborative filtering algorithm to analyze the similarities between users, and mines the potential associations between users and jobs to achieve personalized and accurate job recommendations.

[0004] The specific scheme proposed by the present invention is:

[0005] The present invention provides a job recommendation method based on a collaborative filtering algorithm, comprising:

[0006] Step 1: Collect and preprocess data:

[0007] Step 11: Collect job data, job seeker resume data and user behavior data,

[0008] Step 12: Clean and format the collected data;

[0009] Step 2: Construct a user-position rating matrix: Based on the user's historical behavior data, construct a user-position rating matrix, where rows represent users and columns represent positions. Elements in the matrix represent the user's association score with the position.

[0010] Step 3: Perform collaborative filtering of job seekers and positions based on collaborative filtering algorithms:

[0011] Step 31: Calculate user similarity: Use the Jaccard similarity coefficient calculation method to calculate user similarity.

[0012] Step 32: Calculate job similarity: Form a job feature vector based on the job field content, and calculate job similarity based on the job feature vector.

[0013] Step 33: Select neighbor users: According to user similarity, select K users who are most similar to the target user as neighbor users.

[0014] Step 34: Select neighboring positions: Based on the position similarity, compare the target position with other positions in terms of user preferences, and find the K positions that are most similar to the target position.

[0015] Step 35: Predicting job scores: Multiply the neighbor user's association score by the user similarity as a weight to get the predicted score, and then sort the unvisited jobs by the predicted score. According to the sorting, the total job score is calculated by multiplying the assessment score by the job similarity.

[0016] Step 4: Sort the positions according to their total scores, generate a list of positions and recommend them to target users.

[0017] Step 5: Collect user feedback on job recommendation results for subsequent optimization of the recommendation process.

[0018] Furthermore, the job data collected in step 1 of the job recommendation method based on collaborative filtering algorithm includes: job title, industry category, work experience requirements, salary, education requirements, nature of work, number of people required, work location, job description, skill requirements, and job highlight data.

[0019] The collected resume data of job seekers include: basic information, job search intention, work experience, project experience, education experience, language ability, professional skills, training status, and certification status.

[0020] The user behavior data collected includes: the positions that users have viewed and the length of time they have viewed them, the positions that they have searched for, the positions that they have submitted their resumes to, the positions that they have collected, the job recommendations that they have clicked on, and the evaluation and ratings of the positions.

[0021] Furthermore, in step 2 of the job recommendation method based on collaborative filtering algorithm, the user's relevance score for the job is calculated by a relevance score model, and the calculation formula is:

[0022] Preference(u,j)=α*(Views(u,j)+Applications(u,j)+Favorites(u,j))+β*Rating(u,j)

[0023] in:

[0024] Preference(u,j) indicates the preference of user u for position j.

[0025] Views(u,j) indicates the number of times user u viewed job j.

[0026] Applications(u,j) indicates the number of times user u submitted a resume to position j.

[0027] Favorites(u,j) indicates the number of times user u has favorited job j.

[0028] Rating(u,j) represents the evaluation score of user u on position j.

[0029] α and β represent weight coefficients, which are used to adjust the relative importance of the number of views, number of submissions, number of favorites, and evaluation scores in the calculation of preference.

[0030] Configure according to the data situation to meet the condition of α+β=1.

[0031] Furthermore, in step 4 of the job recommendation method based on a collaborative filtering algorithm, the jobs are sorted according to the predicted scores, a job list is generated, jobs with higher scores are recommended first, and the generated job list is displayed to users through the recommendation column of the recruitment platform, email notifications, and SMS reminders.

[0032] Furthermore, in step 5 of the job recommendation method based on collaborative filtering algorithm, collecting user feedback on the job recommendation results includes: collecting user behavior data on clicks, submissions, and collections of recommended jobs, as well as user evaluations of recommendation quality;

[0033] At the same time, data on market changes and changes in user needs are collected for subsequent optimization of the recommendation process.

[0034] The present invention also provides a job recommendation device based on a collaborative filtering algorithm, comprising a collection and preprocessing module, a scoring management module, a filtering module, a recommendation module and a feedback module.

[0035] The collection and preprocessing module collects and preprocesses data:

[0036] Step 11: Collect job data, job seeker resume data and user behavior data,

[0037] Step 12: Clean and format the collected data;

[0038] The scoring management module constructs a user-position scoring matrix: Based on the user's historical behavior data, a user-position scoring matrix is ​​constructed. The rows of the user-position scoring matrix represent users, the columns represent positions, and the elements in the matrix represent the user's association score with the position.

[0039] The filtering module performs collaborative filtering of job seekers and positions based on the collaborative filtering algorithm:

[0040] Step 31: Calculate user similarity: Use the Jaccard similarity coefficient calculation method to calculate user similarity.

[0041] Step 32: Calculate job similarity: Form a job feature vector based on the job field content, and calculate job similarity based on the job feature vector.

[0042] Step 33: Select neighbor users: According to user similarity, select K users who are most similar to the target user as neighbor users.

[0043] Step 34: Select neighboring positions: Based on the position similarity, compare the target position with other positions in terms of user preferences, and find the K positions that are most similar to the target position.

[0044] Step 35: Predicting job scores: Multiply the neighbor user's association score by the user similarity as a weight to get the predicted score, and then sort the unvisited jobs by the predicted score. According to the sorting, the total job score is calculated by multiplying the assessment score by the job similarity.

[0045] The recommendation module sorts the positions according to their total scores, generates a list of positions and recommends them to target users.

[0046] The feedback module collects user feedback on job recommendation results for subsequent optimization of the recommendation process.

[0047] Furthermore, the job data collected by the collection and preprocessing module of the job recommendation device based on the collaborative filtering algorithm includes: job title, industry category, work experience requirements, salary, education requirements, job nature, number of people required, work location, job description, skill requirements, and job highlight data.

[0048] The resume data collected by the collection and preprocessing module includes: basic information, job search intention, work experience, project experience, education experience, language ability, professional skills, training status, and certificate status.

[0049] The user behavior data collected by the collection and preprocessing module include: the positions that users have browsed and the browsing time, the positions they have searched for, the positions they have submitted resumes to, the positions they have collected, the job recommendations they have clicked on, and the evaluation and scoring of the positions.

[0050] Furthermore, the score management module of the job recommendation device based on the collaborative filtering algorithm calculates the user's relevance score for the job through a relevance score model, and the calculation formula is:

[0051] Preference(u,j)=α*(Views(u,j)+Applications(u,j)+Favorites(u,j))+β*Rating(u,j)

[0052] in:

[0053] Preference(u,j) indicates the preference of user u for position j.

[0054] Views(u,j) indicates the number of times user u viewed job j.

[0055] Applications(u,j) indicates the number of times user u submitted a resume to position j.

[0056] Favorites(u,j) indicates the number of times user u has favorited job j.

[0057] Rating(u,j) represents the evaluation score of user u on position j.

[0058] α and β represent weight coefficients, which are used to adjust the relative importance of the number of views, number of submissions, number of favorites, and evaluation scores in the calculation of preference.

[0059] Configure according to the data situation to meet the condition of α+β=1.

[0060] Furthermore, the recommendation module of the job recommendation device based on the collaborative filtering algorithm sorts the jobs according to the predicted scores, generates a job list, gives priority to recommending jobs with higher scores, and displays the generated job list to users through the recommendation column of the recruitment platform, email notifications, and SMS reminders.

[0061] Furthermore, the feedback module of the job recommendation device based on the collaborative filtering algorithm collects user feedback on the job recommendation results, including: collecting user behavior data on clicks, submissions, and collections of recommended jobs, as well as user evaluations of recommendation quality;

[0062] At the same time, data on market changes and changes in user needs are collected for subsequent optimization of the recommendation process.

[0063] The benefits of the present invention are:

[0064] Improve the accuracy of job recommendations: By deeply analyzing job seekers' historical behavior data and user similarities, the present invention can more accurately grasp the real needs of job seekers, provide job recommendations that meet job seekers' expectations, and reduce the time job seekers spend on screening jobs.

[0065] Enhance user experience and satisfaction: The present invention provides personalized job recommendations based on the unique interests and needs of each job seeker. This customized service can significantly improve the job seeker's experience, reduce the trouble caused by information overload to job seekers, and enable job seekers to find their desired positions more easily.

[0066] Promote the commercial development of recruitment platforms: Accurate job recommendations can improve user experience and satisfaction, thereby increasing user dependence and loyalty to recruitment platforms. It can also provide recruitment platforms with opportunities for value-added services. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a diagram of the job seeker-position relationship.

[0068] Figure 2 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0069] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0070] Example 1

[0071] The present invention provides a job recommendation method based on a collaborative filtering algorithm, comprising:

[0072] Step 1: Collect and preprocess data:

[0073] Step 11: Collect job data, job seeker resume data and user behavior data. Job data includes: job title, industry category, work experience requirements, salary, education requirements, nature of work, number of people required, work location, job description, skill requirements, and job highlights. Job seeker resume data includes: basic information, job search intention, work experience, project experience, education experience, language ability, professional skills, training status, and certificate status. User historical behavior data includes: the positions that users have browsed and the browsing time, the positions that have been searched, the positions that have submitted resumes, the positions that have been collected, the job recommendations that have been clicked, and the evaluation and scoring of the positions.

[0074] Step 12: Clean and format the collected data. When cleaning and formatting the collected data, remove invalid, erroneous or redundant data, standardize the user's rating and browsing time numerical data into a unified data format, and provide high-quality data input for subsequent algorithm execution.

[0075] Step 2: Construct a user-position rating matrix: Construct a user-position rating matrix based on the user's historical behavior data. The user-position rating matrix is ​​a two-dimensional array, where rows represent users and columns represent positions. The elements in the matrix represent the user's association score with the position.

[0076] The correlation score can be calculated based on the correlation score model, and the calculation formula is:

[0077] Preference(u,j)=α*(Views(u,j)+Applications(u,j)+Favorites(u,j))+β*Rating(u,j)

[0078] in:

[0079] Preference(u,j): user u’s preference for position j.

[0080] Views(u,j): The number of times user u viewed job j.

[0081] Applications(u,j): The number of times user u submitted a resume to position j.

[0082] Favorites(u,j): The number of times user u has favorited job j.

[0083] Rating(u,j): user u's rating score for position j (if any, it can be five points, ten points, etc., depending on the platform settings).

[0084] α and β: weight coefficients used to adjust the relative importance of the number of views, number of submissions, number of favorites, and evaluation scores in the calculation of preference.

[0085] Configure according to the data situation to meet the condition of α+β=1.

[0086] Step 3: Perform collaborative filtering of job seekers and positions based on collaborative filtering algorithms:

[0087] Step 31: Calculate user similarity: Use the Jaccard similarity coefficient calculation method to calculate user similarity. Calculate the similarity between users to form a user similarity matrix. The higher the similarity, the more similar the interests and needs between the two users.

[0088] Step 32: Calculate job similarity: Form a job feature vector based on the job field content, and calculate job similarity based on the job feature vector.

[0089] Step 33: Select neighbor users: According to user similarity, select K users who are most similar to the target user as neighbor users.

[0090] Step 34: Select neighboring positions: Based on the position similarity, compare the target position with other positions in terms of user preferences, and find the K positions that are most similar to the target position.

[0091] Step 35: Predict job score: Multiply the neighbor user's association score by the user similarity as the weight to get the predicted score, and then sort the unvisited jobs by the predicted score. According to the sorting, calculate the total job score by multiplying the assessment score by the job similarity. The higher the total score, the greater the target user's interest and demand for the job.

[0092] Step 4: Sort the positions according to their total scores, generate a job list and recommend it to the target users. The positions can be sorted according to the predicted scores to generate a job list, with higher-scoring positions being recommended first. The generated job list will be displayed to users through the recommendation column of the recruitment platform, email notifications, and SMS reminders.

[0093] Step 5: Collect user feedback on job recommendation results for subsequent optimization of the recommendation process.

[0094] Collecting user feedback on job recommendation results may include: collecting user click, submission, and favorite behavior data on recommended jobs, as well as user evaluation of recommendation quality;

[0095] At the same time, data on market changes and changes in user needs are collected for subsequent optimization of the recommendation process.

[0096] Example 2

[0097] The present invention also provides a job recommendation device based on a collaborative filtering algorithm, comprising a collection and preprocessing module, a scoring management module, a filtering module, a recommendation module and a feedback module.

[0098] The collection and preprocessing module collects and preprocesses data:

[0099] Step 11: Collect job data, job seeker resume data and user behavior data,

[0100] Step 12: Clean and format the collected data;

[0101] The scoring management module constructs a user-position scoring matrix: Based on the user's historical behavior data, a user-position scoring matrix is ​​constructed. The rows of the user-position scoring matrix represent users, the columns represent positions, and the elements in the matrix represent the user's association score with the position.

[0102] The filtering module performs collaborative filtering of job seekers and positions based on the collaborative filtering algorithm:

[0103] Step 31: Calculate user similarity: Use the Jaccard similarity coefficient calculation method to calculate user similarity.

[0104] Step 32: Calculate job similarity: Based on the job field content, form a job feature vector, calculate job similarity based on the job feature vector, and then associate it through user behavior

[0105] Step 33: Select neighbor users: According to user similarity, select K users who are most similar to the target user as neighbor users.

[0106] Step 34: Select neighboring positions: Based on the position similarity, compare the target position with other positions in terms of user preferences, and find the K positions that are most similar to the target position.

[0107] Step 35: Predicting job scores: Multiply the neighbor user's association score by the user similarity as a weight to get the predicted score, and then sort the unvisited jobs by the predicted score. According to the sorting, the total job score is calculated by multiplying the assessment score by the job similarity.

[0108] The recommendation module sorts the positions according to their total scores, generates a list of positions and recommends them to target users.

[0109] The feedback module collects user feedback on job recommendation results for subsequent optimization of the recommendation process.

[0110] As the information interaction and execution process between the modules in the above-mentioned device are based on the same concept as the embodiment of the method of the present invention, the specific contents can be found in the description of the embodiment of the method of the present invention and will not be repeated here.

[0111] Likewise, the benefits of the device of the present invention are:

[0112] Improve the accuracy of job recommendations: By deeply analyzing job seekers' historical behavior data and user similarities, the present invention can more accurately grasp the real needs of job seekers, provide job recommendations that meet job seekers' expectations, and reduce the time job seekers spend on screening jobs.

[0113] Enhanced user experience and satisfaction: The present invention provides personalized job recommendations based on the unique interests and needs of each job seeker. This customized service can significantly improve the job seeker's experience, reduce the trouble caused by information overload to job seekers, and enable job seekers to find their desired positions more easily.

[0114] Promote the commercial development of recruitment platforms: Accurate job recommendations can improve user experience and satisfaction, thereby increasing user dependence and loyalty to recruitment platforms. It can also provide recruitment platforms with opportunities for value-added services.

[0115] It should be noted that not all steps and modules in the above-mentioned processes and device structures are necessary, and some steps or modules can be ignored according to actual needs. The execution order of each step is not fixed and can be adjusted as needed. The system structure described in the above-mentioned embodiments can be a physical structure or a logical structure, that is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities, or some components in multiple independent devices may be implemented together.

[0116] The above-described embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or changes made by those skilled in the art based on the present invention are within the protection scope of the present invention. The protection scope of the present invention shall be subject to the claims.

Claims

1. A job recommendation method based on collaborative filtering algorithm, characterized by include: Step 1: Collect and preprocess data: Step 11: Collect job data, job seeker resume data and user behavior data, Step 12: Clean and format the collected data; Step 2: Construct a user-position rating matrix: Based on the user's historical behavior data, construct a user-position rating matrix, where rows represent users and columns represent positions. Elements in the matrix represent the user's association score with the position. Step 3: Perform collaborative filtering of job seekers and positions based on collaborative filtering algorithms: Step 31: Calculate user similarity: Use the Jaccard similarity coefficient calculation method to calculate user similarity. Step 32: Calculate job similarity: Based on the job field content, form a job feature vector, calculate job similarity based on the job feature vector, and then associate it through user behavior Step 33: Select neighbor users: According to user similarity, select K users who are most similar to the target user as neighbor users. Step 34: Select neighboring positions: Based on the position similarity, compare the target position with other positions in terms of user preferences, and find the K positions that are most similar to the target position. Step 35: Predicting job scores: Multiply the neighbor user's association score by the user similarity as a weight to get the predicted score, and then sort the unvisited jobs by the predicted score. According to the sorting, the total job score is calculated by multiplying the assessment score by the job similarity. Step 4: Sort the positions according to their total scores, generate a list of positions and recommend them to target users. Step 5: Collect user feedback on job recommendation results for subsequent optimization of the recommendation process.

2. The method for job recommendation based on collaborative filtering algorithm according to claim 1, characterized in that The job data collected in step 1 include: job title, industry category, work experience requirements, salary, education requirements, nature of work, number of people required, work location, job description, skill requirements, and job highlights. The collected resume data of job seekers include: basic information, job search intention, work experience, project experience, education experience, language ability, professional skills, training status, and certification status. The user behavior data collected includes: the positions that users have viewed and the length of time they have viewed them, the positions that they have searched for, the positions that they have submitted their resumes to, the positions that they have collected, the job recommendations that they have clicked on, and the evaluation and ratings of the positions.

3. The method for job recommendation based on collaborative filtering algorithm according to claim 1, characterized in that In step 2, the user's relevance score for the position is calculated using the relevance score model, and the calculation formula is: Preference(u,j)=α*(Views(u,j)+Applications(u,j)+Favorites(u,j))+β*Rating(u,j) in: Preference(u,j) indicates the preference of user u for position j. Views(u,j) indicates the number of times user u viewed job j. Applications(u,j) indicates the number of times user u submitted a resume to position j. Favorites(u,j) indicates the number of times user u has favorited job j. Rating(u,j) represents the evaluation score of user u on position j. α and β represent weight coefficients, which are used to adjust the relative importance of the number of views, number of submissions, number of favorites, and evaluation scores in the calculation of preference. Configure according to the data situation to meet the condition of α+β=1.

4. The method for job recommendation based on collaborative filtering algorithm according to claim 1, characterized in that In step 4, the positions are sorted according to the predicted scores to generate a position list, with positions with higher scores being recommended first. The generated position list is then displayed to users via the recommendation column of the recruitment platform, email notifications, and SMS reminders.

5. The method for job recommendation based on collaborative filtering algorithm according to claim 1, characterized in that In step 5, user feedback on the job recommendation results is collected, including: collecting user behavior data on clicks, submissions, and favorites of recommended jobs, as well as user evaluations of the quality of recommendations; At the same time, data on market changes and changes in user needs are collected for subsequent optimization of the recommendation process.

6. A job recommendation device based on collaborative filtering algorithm, characterized in that It includes collection and preprocessing module, scoring management module, filtering module, recommendation module and feedback module. The collection and preprocessing module collects and preprocesses data: Step 11: Collect job data, job seeker resume data and user behavior data, Step 12: Clean and format the collected data; The scoring management module constructs a user-position scoring matrix: Based on the user's historical behavior data, a user-position scoring matrix is ​​constructed. The rows of the user-position scoring matrix represent users, the columns represent positions, and the elements in the matrix represent the user's association score with the position. The filtering module performs collaborative filtering of job seekers and positions based on the collaborative filtering algorithm: Step 31: Calculate user similarity: Use the Jaccard similarity coefficient calculation method to calculate user similarity. Step 32: Calculate job similarity: Based on the job field content, form a job feature vector, calculate job similarity based on the job feature vector, and then associate it through user behavior Step 33: Select neighbor users: According to user similarity, select K users who are most similar to the target user as neighbor users. Step 34: Select neighboring positions: Based on the position similarity, compare the target position with other positions in terms of user preferences, and find the K positions that are most similar to the target position. Step 35: Predicting job scores: Multiply the neighbor user's association score by the user similarity as a weight to get the predicted score, and then sort the unvisited jobs by the predicted score. According to the sorting, the total job score is calculated by multiplying the assessment score by the job similarity. The recommendation module sorts the positions according to their total scores, generates a list of positions and recommends them to target users. The feedback module collects user feedback on job recommendation results for subsequent optimization of the recommendation process.

7. The job recommendation device based on collaborative filtering algorithm according to claim 6, characterized in that The job data collected by the collection and preprocessing module include: job title, industry category, work experience requirements, salary, education requirements, nature of work, number of people required, work location, job description, skill requirements, and job highlights. The resume data collected by the collection and preprocessing module includes: basic information, job search intention, work experience, project experience, education experience, language ability, professional skills, training status, and certificate status. The user behavior data collected by the collection and preprocessing module include: the positions that users have browsed and the browsing time, the positions they have searched for, the positions they have submitted resumes to, the positions they have collected, the job recommendations they have clicked on, and the evaluation and scoring of the positions.

8. The job recommendation device based on collaborative filtering algorithm according to claim 6, characterized in that The scoring management module calculates the user's relevance score for the position through the relevance score model. The calculation formula is: Preference(u,j)=α*(Views(u,j)+Applications(u,j)+Favorites(u,j))+β*Rating(u,j) in: Preference(u,j) indicates the preference of user u for position j. Views(u,j) indicates the number of times user u viewed job j. Applications(u,j) indicates the number of times user u submitted a resume to position j. Favorites(u,j) indicates the number of times user u has favorited job j. Rating(u,j) represents the evaluation score of user u on position j. α and β represent weight coefficients, which are used to adjust the relative importance of the number of views, number of submissions, number of favorites, and evaluation scores in the calculation of preference. Configure according to the data situation to meet the condition of α+β=1.

9. The job recommendation device based on collaborative filtering algorithm according to claim 6, characterized in that The recommendation module sorts the positions according to the predicted scores, generates a position list, gives priority to positions with higher scores, and displays the generated position list to users through the recommendation column of the recruitment platform, email notifications, and SMS reminders.

10. The job recommendation device based on collaborative filtering algorithm according to claim 6, characterized in that The feedback module collects user feedback on job recommendation results, including: collecting user behavior data on clicks, submissions, and collections of recommended jobs, as well as user evaluations of the quality of recommendations; it also collects market change data and user demand change data for subsequent optimization of the recommendation process.

Images