High-end talent background investigation method for full-cycle employment risk management and control

Abstract

A high-end talent background investigation method for full-cycle employment risk management and control comprises the following steps: step 1, verifying identities, educational background and legal dispute record information of applicants by searching network public information, an educational information network and a legal document network; 2, directional crawling is carried out on news websites, social media, forums, blogs and the like through a web crawler tool Scrapy, public opinion information of applicants in past work is searched for through a public opinion monitoring system, then TF-IDF feature extraction is carried out on texts, and the TF-IDF features of the texts are extracted; and carrying out positive public opinion classification, negative public opinion classification and neutral public opinion classification on the extracted data by adopting a support vector machine algorithm so as to evaluate positive or negative influences of applicants on past enterprises. According to the invention, background investigation can be comprehensively carried out on the applicants in multiple dimensions, so that enterprises can truly and objectively know the applicants, and the recruitment efficiency of the enterprises is improved.

Description

Technical Field

[0001] This invention relates to the field of information technology in human resource surveys, specifically to a method for background investigation of high-end talent with full-cycle employment risk management. Background Technology

[0002] During the recruitment process, some applicants for mid-to-senior level positions submit falsified resumes. Employers often struggle to identify these falsified resumes. Thoroughly verifying the authenticity of a resume requires significant time and effort from the employer. Employers typically only check an applicant's education and identity information, failing to investigate past criminal records or thoroughly understand their performance in previous positions. This lack of comprehensive understanding leads to discrepancies between the applicant's actual abilities and their resume description, resulting in the company failing to recruit candidates who truly meet the job requirements and ultimately impacting its development.

[0003] In addition, traditional background check processes only focus on investigating the applicant's past work experience. Once the applicant has completed onboarding, the investigation process ends, without investigating the applicant's performance during the new job or after leaving the previous job. Therefore, the service content of traditional background check processes is relatively one-sided and fails to achieve full-cycle management of hiring risks. Summary of the Invention

[0004] The purpose of this invention is to provide a high-end talent background investigation method for full-cycle employment risk management. This method can conduct a comprehensive and multi-dimensional background investigation of applicants, helping companies to understand applicants truthfully and objectively, and improve the efficiency of recruitment.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for background investigation of high-end talent for full-cycle employment risk management, comprising the following steps: Step 1: Verify the applicant's identity, academic qualifications, and legal dispute records by searching publicly available information online, the China Higher Education Student Information System (CHESICC), and legal document databases. Step 2: Using the web crawler Scrapy, targeted data is crawled from news websites, social media, forums, blogs, etc., and a public opinion monitoring system is used to search for public opinion information about the applicant's past work. Then, TF-IDF features are extracted from the text, and the support vector machine algorithm is used to classify the extracted data into positive, negative, and neutral public opinion, thereby assessing the positive or negative impact that the applicant has had on the past companies. Step 3: Combining social security records and individual income tax records, conduct interviews with HR or the applicant's superiors and subordinates to create interview comparison texts. Use the BERT / RoBERTa model to compare the interview comparison texts, social security records, individual income tax records, and the applicant's resume description to verify the applicant's work unit, start and end dates, position and reporting relationship, salary, reasons for leaving the previous job, major violations of rules and regulations, labor arbitration records, and non-compete agreement information. Step 4: Through interviews with the applicant's former direct supervisor and other colleagues, multiple interview scoring samples are generated. The TF-IDF algorithm is used to count the core evaluation words in each interview scoring sample for dimensions such as the applicant's work efficiency, work completion quality, professional ability, learning ability, management ability, interpersonal relationships, workload, work pressure, and stress resistance. A random forest model is trained to assign corresponding scores to different core evaluation words. The random forest model is used to score each dimension of each interview scoring sample to form the applicant's comprehensive quality matrix. Step 5: Input the applicant's comprehensive quality matrix into the comprehensive evaluation model XGBoost, and use the comprehensive evaluation model XGBoost to predict the suitability of the applicant and the applied position.

[0006] Step Six: Input the applicant's comprehensive quality matrix into the Sentence-BERT model. The Sentence-BERT model compares the applicant's comprehensive quality matrix with the applicant's resume to evaluate the authenticity of the applicant's resume. Step 7: After the applicant joins the company, the company scans the job application information across the entire platform and uses a turnover intention model to obtain the frequency of the applicant refreshing job application information on job search websites and the frequency of the applicant taking leave, thereby identifying the applicant's turnover intention. Step 8: Follow up on the applicant's integration status. Through colleague interviews, track the applicant's work status in real time after they start work. Use the TF-IDF algorithm to count the core evaluation words in each interview rating sample for the applicant's work efficiency, work engagement, workload, and work results. Train a random forest model to assign corresponding scores to different core evaluation words. Based on the corresponding scores, use Tableau to generate a dynamic profile of the talent status. Step Nine: After the applicant leaves the company, the community discovery algorithm will be used to continuously follow up on relevant information, analyze the path and key nodes of information dissemination, and understand whether the applicant has taken away the client of the company he left. In combination with the non-compete agreement signed by the applicant, we will analyze whether the applicant has violated the provisions of the non-compete agreement. Step 10: Maintain long-term telephone contact with job applicants after they leave the company to preserve their good relationship with the former employer and increase the likelihood of them rejoining the company.

[0007] Specifically, step one includes checking and verifying the applicant's personal litigation records, records of being restricted from high-level consumption, and civil and commercial judgment documents.

[0008] Specifically, in step one, the applicant's negative tax information and administrative violation records are checked and verified.

[0009] Specifically, in step two, the applicant's negative social information / concern information is queried and verified.

[0010] Specifically, in step one, the applicant's current financial credit status and online lending risks are checked and verified.

[0011] Specifically, in step one, the applicant's educational background and identity information are queried and verified.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: Once a candidate joins the company, their job search information is scanned across the entire platform. A turnover intention model is used to obtain information on how frequently the candidate refreshes their job search information on job search websites and how often they request leave. This helps identify the candidate's turnover intention. Once a turnover intention is identified, human resources personnel are immediately dispatched to communicate with the candidate to fully understand their dissatisfaction with the company and strive to retain them.

[0013] We continuously monitor applicants' integration status, track their work performance in real time through colleague interviews, and use the TF-IDF algorithm to analyze the core evaluation terms for each interview sample, including work efficiency, work engagement, workload, and work results. We then train a random forest model to assign corresponding scores to different core evaluation terms. Based on these scores, we use Tableau to generate a dynamic profile of the talent status, enabling us to monitor applicants' work performance in real time and assess their suitability for the position.

[0014] After a candidate leaves a company, the community discovery algorithm continuously tracks relevant information, analyzes the information dissemination path and key nodes to understand whether the candidate took away the company's clients. Combined with the candidate's signed non-compete agreement, it analyzes whether the candidate violated the provisions of the non-compete agreement. This step allows for real-time monitoring of the candidate's post-departure activities, forming a closed-loop monitoring system covering the entire lifecycle of a candidate—before joining, after joining, and after leaving. This ensures risk control throughout the entire recruitment process, minimizing the risks for the company in recruitment and staffing, safeguarding the company's development, improving recruitment efficiency, and contributing to cost reduction and efficiency improvement. Detailed Implementation

[0015] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0016] A method for background investigation of high-end talent for full-cycle employment risk management, comprising the following steps: Step 1: Verify the applicant's identity, academic qualifications, and legal dispute records by searching publicly available information online, the China Higher Education Student Information System (CHESICC), and legal document databases. Step 2: Using the web crawler Scrapy, targeted data is crawled from news websites, social media, forums, blogs, etc., and a public opinion monitoring system is used to search for public opinion information about the applicant's past work. Then, TF-IDF features are extracted from the text, and the support vector machine algorithm is used to classify the extracted data into positive, negative, and neutral public opinion, thereby assessing the positive or negative impact that the applicant has had on the past companies. Step 3: Combining social security records and individual income tax records, conduct interviews with HR or the applicant's superiors and subordinates to create interview comparison texts. Use the BERT / RoBERTa model to compare the interview comparison texts, social security records, individual income tax records, and the applicant's resume description to verify the applicant's work unit, start and end dates, position and reporting relationship, salary, reasons for leaving the previous job, major violations of rules and regulations, labor arbitration records, and non-compete agreement information. Step 4: Through interviews with the applicant's former direct supervisor and other colleagues, multiple interview scoring samples are generated. The TF-IDF algorithm is used to count the core evaluation words in each interview scoring sample for dimensions such as the applicant's work efficiency, work completion quality, professional ability, learning ability, management ability, interpersonal relationships, workload, work pressure, and stress resistance. A random forest model is trained to assign corresponding scores to different core evaluation words. The random forest model is used to score each dimension of each interview scoring sample to form the applicant's comprehensive quality matrix. Step 5: Input the applicant's comprehensive quality matrix into the comprehensive evaluation model XGBoost, and use the comprehensive evaluation model XGBoost to predict the suitability of the applicant and the applied position.

[0017] Step Six: Input the applicant's comprehensive quality matrix into the Sentence-BERT model. The Sentence-BERT model compares the applicant's comprehensive quality matrix with the applicant's resume to evaluate the authenticity of the applicant's resume. Step 7: After the applicant joins the company, the company scans the job application information across the entire platform and uses a turnover intention model to obtain the frequency of the applicant refreshing job application information on job search websites and the frequency of the applicant taking leave, thereby identifying the applicant's turnover intention. Step 8: Continuously follow up on the applicant's integration status. Through colleague interviews, track the applicant's work status after joining the company in real time. Use the TF-IDF algorithm to count the core evaluation words of the applicant's work efficiency, work engagement, workload, and work results in each interview rating sample. Train a random forest model to assign corresponding scores to different core evaluation words. Based on the corresponding scores, use Tableau tool to generate a dynamic profile of the talent status. Step Nine: After the applicant leaves the company, the community discovery algorithm will be used to continuously follow up on relevant information, analyze the path and key nodes of information dissemination, and understand whether the applicant has taken away the client of the company he left. In combination with the non-compete agreement signed by the applicant, we will analyze whether the applicant has violated the provisions of the non-compete agreement. Step 10: Maintain long-term telephone contact with job applicants after they leave the company to preserve their good relationship with the former employer and increase the likelihood of them rejoining the company.

[0018] Specifically, step one includes checking and verifying the applicant's personal litigation records, records of being restricted from high-level consumption, and civil and commercial judgment documents.

[0019] Specifically, in step one, the applicant's negative tax information and administrative violation records are checked and verified.

[0020] Specifically, in step two, the applicant's negative social information / concern information is queried and verified.

[0021] Specifically, in step one, the applicant's current financial credit status and online lending risks are checked and verified.

[0022] Specifically, in step one, the applicant's educational background and identity information are queried and verified.

[0023] The beneficial effects of this invention are as follows: By using the web crawler Scrapy to target news websites, social media, forums, blogs, etc., relevant data can be obtained in a targeted and accurate manner, ensuring accurate identification. A public opinion monitoring system is used to search for public opinion information related to applicants' past work, thereby obtaining massive amounts of public opinion data in real time, ensuring comprehensive identification. TF-IDF feature extraction is performed on the text, and a support vector machine algorithm is used to classify the extracted data into positive, negative, and neutral public opinion, enabling accurate classification of public opinion information and achieving scientific and efficient public opinion analysis.

[0024] By combining social security records and individual income tax records, and through interviews with HR personnel or the applicant's superiors and subordinates, comparative interview transcripts are generated. The BERT / RoBERTa model is used to compare these transcripts with the applicant's social security records, individual income tax records, and resume descriptions. By using interview transcripts from the applicant's former colleagues as data sources, and then using the BERT / RoBERTa model to transform the interview data into comparable data, this data is compared with the applicant's social security records, individual income tax records, and resume descriptions. This allows for efficient and accurate verification of the applicant's work unit, start and end dates, position and reporting relationship, salary, reasons for leaving previous jobs, major violations of rules and regulations, labor arbitration records, and non-compete agreements.

[0025] Interviews with the applicant's former direct supervisor and other colleagues generated multiple interview rating samples. The TF-IDF algorithm was used to statistically analyze the core evaluation terms for each interview rating sample across dimensions such as work efficiency, work quality, professional skills, learning ability, management skills, interpersonal relationships, workload, work pressure, and stress resistance. A random forest model was trained to assign corresponding scores to different core evaluation terms. This random forest model was then used to score each dimension of each interview rating sample, forming a comprehensive applicant competency matrix. This matrix was then input into the XGBoost comprehensive evaluation model to predict the suitability of the applicant for the position. This process is efficient, accurate, and objective, reducing the influence of subjectivity on recruitment decisions. Finally, the comprehensive applicant competency matrix was input into the Sentence-BERT model. The Sentence-BERT model compared the matrix with the applicant's resume to evaluate the authenticity of the resume, thus providing early warning of fraudulent resumes.

[0026] Once a candidate joins the company, their job search information is scanned across the entire platform. A turnover intention model is used to obtain information on how frequently the candidate refreshes their job search information on job search websites and how often they request leave. This helps identify the candidate's turnover intention. Once a turnover intention is identified, human resources personnel are immediately dispatched to communicate with the candidate to fully understand their dissatisfaction with the company and strive to retain them.

[0027] We continuously monitor applicants' integration status, track their work performance in real time through colleague interviews, and use the TF-IDF algorithm to analyze the core evaluation terms for each interview sample, including work efficiency, work engagement, workload, and work results. We then train a random forest model to assign corresponding scores to different core evaluation terms. Based on these scores, we use Tableau to generate a dynamic profile of the talent status, enabling us to monitor applicants' work performance in real time and assess their suitability for the position.

[0028] After a candidate leaves a company, the community discovery algorithm continuously tracks relevant information, analyzes the information dissemination path and key nodes to understand whether the candidate took away the company's clients. Combined with the candidate's signed non-compete agreement, it analyzes whether the candidate violated the provisions of the non-compete agreement. This step allows for real-time monitoring of the candidate's post-departure activities, forming a closed-loop monitoring system covering the entire lifecycle of a candidate—before joining, after joining, and after leaving. This ensures risk control throughout the entire recruitment process, minimizing the risks for the company in recruitment and staffing, safeguarding the company's development, improving recruitment efficiency, and contributing to cost reduction and efficiency improvement.

[0029] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A high-end talent background investigation method for whole-cycle employment risk management, characterized in that, Includes the following steps: Step 1: Verify the applicant's identity, academic qualifications, and legal dispute records by searching publicly available information online, the China Higher Education Student Information System (CHESICC), and legal document databases. Step 2: Using the web crawler Scrapy, targeted data is crawled from news websites, social media, forums, blogs, etc., and a public opinion monitoring system is used to search for public opinion information about the applicant's past work. Then, TF-IDF features are extracted from the text, and the support vector machine algorithm is used to classify the extracted data into positive, negative, and neutral public opinion, thereby assessing the positive or negative impact that the applicant has had on the past companies. Step 3: Combining social security records and individual income tax records, conduct interviews with HR or the applicant's superiors and subordinates to create interview comparison texts. Use the BERT / RoBERTa model to compare the interview comparison texts, social security records, individual income tax records, and the applicant's resume description to verify the applicant's work unit, start and end dates, position and reporting relationship, salary, reasons for leaving the previous job, major violations of rules and regulations, labor arbitration records, and non-compete agreement information. Step 4: Through interviews with the applicant's former direct supervisor and other colleagues, multiple interview scoring samples are generated. The TF-IDF algorithm is used to count the core evaluation words in each interview scoring sample for dimensions such as the applicant's work efficiency, work completion quality, professional ability, learning ability, management ability, interpersonal relationships, workload, work pressure, and stress resistance. A random forest model is trained to assign corresponding scores to different core evaluation words. The random forest model is used to score each dimension of each interview scoring sample to form the applicant's comprehensive quality matrix. Step 5: Input the applicant's comprehensive quality matrix into the comprehensive evaluation model XGBoost, and use the comprehensive evaluation model XGBoost to predict the suitability of the applicant and the applied position; Step Six: Input the applicant's comprehensive quality matrix into the Sentence-BERT model. The Sentence-BERT model compares the applicant's comprehensive quality matrix with the applicant's resume to evaluate the authenticity of the applicant's resume. Step 7: After the applicant joins the company, the company scans the job application information across the entire platform and uses a turnover intention model to obtain the frequency of the applicant refreshing job application information on job search websites and the frequency of the applicant taking leave, thereby identifying the applicant's turnover intention. Step 8: Follow up on the applicant's integration status. Through colleague interviews, track the applicant's work status in real time after they start work. Use the TF-IDF algorithm to count the core evaluation words in each interview rating sample for the applicant's work efficiency, work engagement, workload, and work results. Train a random forest model to assign corresponding scores to different core evaluation words. Based on the corresponding scores, use Tableau to generate a dynamic profile of the talent status. Step Nine: After the applicant leaves the company, the community discovery algorithm will be used to continuously follow up on relevant information, analyze the path and key nodes of information dissemination, and understand whether the applicant took away the client of the company he left. In combination with the non-compete agreement signed by the applicant, we will analyze whether the applicant violated the provisions of the non-compete agreement. Step 10: Maintain long-term telephone contact with job applicants after they leave the company to preserve their good relationship with the former employer and increase the likelihood of them rejoining the company.

2. The high-end talent background investigation method of full-cycle employment risk management according to claim 1, characterized in that: Step one includes checking and verifying the applicant's personal litigation records, records of being restricted from high-level consumption, and civil and commercial judgment documents.

3. The high-end talent background investigation method of full-cycle employment risk management according to claim 1, characterized in that: In step one, the applicant's negative tax information and administrative violation records are checked and verified.

4. The method for background investigation of high-end talent for full-cycle employment risk management according to claim 1, characterized in that: In step two, information regarding the applicant's negative social impact or areas of concern is investigated and verified.

5. The high-end talent background investigation method of full-cycle employment risk management according to claim 1, characterized in that: In step one, the applicant's current financial credit status and online lending risks are checked and verified.

6. The high-end talent background investigation method of full-cycle employment risk management according to claim 1, characterized in that: In step one, the applicant's educational background and identity information are queried and verified.