An employment recommendation method based on a cause-effect association ontology model

Abstract

The application discloses an employment recommendation method based on a cause-effect association ontology model and belongs to the field of employment recommendation.The application analyzes the cause-effect association relationship of graduate information on employment selection and constructs a cause-effect association ontology model, which is used as prior knowledge to improve a collaborative filtering algorithm, referred to as an OWCF algorithm.The process of the OWCF algorithm for employment recommendation is as follows: through matching of the graduate information and the cause-effect association ontology model prior knowledge base, corresponding characteristic weights are obtained, and the characteristic weights are added to corresponding characteristics to obtain weighted characteristic values of the graduates for a certain type of post; the weighted characteristic values are subjected to similarity calculation with post characteristics to be recommended, meanwhile, the cause-effect association ontology model is used to calculate the similarity of the graduates, through setting of a similarity threshold value, employment units with high similarity and the graduates are obtained, and the obtained units and the employment units of the graduates with high similarity are fused to form a recommendation list.

Description

Technical Field

[0001] This invention relates to an employment recommendation method based on a causal correlation ontology model, belonging to the field of employment recommendation. Background Technology

[0002] In recent years, with the development of information technology, people are faced with an overload of information, making it difficult to choose, and products often fail to reach potential consumers, leading to a supply-demand imbalance in the commodity trading market. This has spurred research into recommendation algorithms, with many scholars dedicating themselves to this field. Meanwhile, with the expansion of university enrollment and the continuous increase in the number of graduates, the job market also faces a serious supply-demand imbalance, a major challenge for current university employment efforts. To address this supply-demand problem in the job market, effectively filtering and utilizing the vast amount of redundant information is a pressing issue. Compared to other recommendation fields, scholars researching job recommendation are relatively few. Common problems in job recommendation include cold start, interpretability of recommendation results, personalized recommendations, and supply-demand matching. In addition to these issues, job recommendation, compared to other product recommendations, suffers from an inherent flaw: data sparsity. Regardless of whether the graduate is a current or past graduate, job information is essentially only available once, leading to...

[0003] Facing the aforementioned problems of cold start, interpretability of recommendation results, personalized recommendation, and supply-demand matching, most job recommendation efforts address the cold start problem by calculating similarity between information on current and past graduates or by recommending trending content. To solve the personalized recommendation problem, many scholars utilize implicit operations by graduates to uncover their potential interests and recommend jobs. To address the supply-demand matching problem, many scholars employ bilateral matching algorithms. However, research on the interpretability of recommendation results is scarce, yet interpretability is an indispensable aspect of job recommendation. Given this external environment and the inherent data sparsity problem in job recommendation, there is an urgent need to propose a practical and feasible method for job recommendation. Summary of the Invention

[0004] This invention addresses the shortcomings of existing job recommendation technologies by providing a job recommendation method based on a causal association rule ontology model, which solves problems such as cold start, interpretability of recommendation results, personalized recommendation, and data sparsity in the job recommendation process.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: an employment recommendation method based on a causal correlation ontology model, comprising the following steps:

[0006] (1) Obtain personal characteristic information and employment data of previous graduates and personal characteristic information of current graduates, perform data cleaning, data completion, data standardization and uniform standardization of data expression processing, and obtain the characteristic set of influencing factors of previous graduates W, the characteristic set of employment element information of graduates WJ and the characteristic set of influencing factors of current graduates U;

[0007] W{A1, A2, A3, ..., A n}, WJ{B1, B2, B3, ..., B n}, U{E1, E2, E3, ..., E n}

[0008] (2) Using the FP-Growth association rule algorithm, association rules are extracted from the features in the feature set W of previous graduates' influencing factors and all features in the feature set WJ of graduates' employment element information. The formulas for extracting the support and confidence of graduates' influencing factors and employment element information are as follows:

[0009] Support(A n →B)=(A n ∩B) / N

[0010] Confidence (A) n →B)=(Support(A n →B)) / (Support(A n ))

[0011] In the formula: Support(A→B) represents the support level; Confidence(A→B) represents the confidence level; A n B represents a single feature in the feature set W of influencing factors of previous graduates; B represents all elements in the feature set WJ of graduate employment factor information; N represents A n The quantity.

[0012] (3) Assess the credibility of the corresponding impact results of the influencing factors, perform information fusion, take the confidence level as the degree of influence of the graduate influencing factors on employment factors, analyze the causal relationship, and construct a causal relationship ontology model.

[0013] (4) Transform the causal association ontology model into feature weights of different categories, classify the influence results of the factors affecting graduates in the causal association ontology, and assign corresponding weights to the feature values ​​F of each category to obtain the feature weight set F of the causal association ontology.

[0014] F={F1:w1,F2:w2,F3:w3,...,F n :w n}

[0015] In the formula: F n :w n Represents the nth feature F n Multiple corresponding weight values ​​w n .

[0016] (5) After information fusion through causal association ontology, job recommendations are made based on the influencing factors of recent graduates, the feature values ​​of the recommended employers, and the causal association ontology. By comparing the influencing factors of recent graduates with the corresponding information in the causal association ontology model, corresponding weights are assigned to the recent graduates. Using the resulting feature weight information of recent graduates, the feature weights are weighted with the corresponding original feature values ​​to obtain the weighted feature values ​​of the recent graduates. These weighted feature values ​​are then matched bilaterally with the feature information of the recommended employers to obtain a high-similarity candidate recommendation list. The specific steps are as follows:

[0017] (51) Calculate the distance between the feature set U of the influencing factors of fresh graduates and the feature weight set F of the causal relationship ontology. The distance calculation formula is as follows:

[0018]

[0019] In the formula: U i F represents the i-th feature in the feature set U of influencing factors for recent graduates; i The i-th feature in the feature weight set F of the causal association ontology.

[0020] (52) Based on the similarity calculation results of step (51), the feature weights in the causal association ontology with high similarity are assigned to the recent graduates.

[0021] (53) The feature weights w obtained from step (52) for recent graduates n The original values ​​{E1, E2, E3, ..., E...} of the corresponding features in the feature set U of influencing factors for recent graduates. n The weighted eigenvalues ​​are obtained by weighting the eigenvalues, and are called the weighted eigenvalues ​​U. f The calculation formula is as follows:

[0022] U f =E n w n

[0023] In the formula: E n This represents the original value of the nth feature in the feature set U of factors influencing recent graduates; w n The feature weight set F representing the causal relationship ontology and E n Features F with high similarity n The corresponding weights.

[0024] (54) Obtain the characteristic information of the employer to be recommended for employment, which is called the characteristic value J. f The weighted eigenvalues ​​U of recent graduates f The characteristic value J of the unit to be recommended for employment f The similarity calculation is performed using the following formula:

[0025]

[0026] In the formula: P(U f J f ) represents U f and J f Similarity; E represents expectation.

[0027] (55) Based on the similarity calculation result P(U) obtained in step (54), f J f ), sort by similarity, and form a list of the n most similar employers to be recommended.

[0028] J n =Rank n (P(U f J f ))

[0029] In the formula: J n Rank represents the list of the top n employers with high similarity to the target employer; n Indicates U f and J f The obtained similarity calculation result P(U) f J f The top n employers with the highest similarity after sorting are selected as recommended employers.

[0030] (6) Based on different feature classifications, repeat steps (4) and (5) to merge and deduplicate the recommended employment units obtained from different features, and sort the recommended employment units according to feature importance.

[0031] (7) Based on the influencing factors of current graduates and previous graduates, as well as the causal relationship ontology, job recommendations are made using a collaborative filtering algorithm. The specific steps are as follows:

[0032] (71) Calculate the similarity between the feature set U of the influencing factors of current graduates and the feature set W of the influencing factors of previous graduates according to the feature categories in step (4). The similarity calculation formula is as follows:

[0033]

[0034] In the formula: C(U, W) represents the similarity between the feature set U of influencing factors for current graduates and the feature set W of influencing factors for previous graduates; UW represents the inner product of the feature vectors of U and W; num(U∩W) represents the number of identical features in the feature vectors of U and W; ||U|| represents the norm of U, and ||W|| represents the norm of W; num U num represents the number of features in U; W This represents the number of features in W.

[0035] (72) The similarity obtained in step (71) is classified according to the features. The similarity of the influencing factors of each category is weighted and fused to obtain the weighted feature similarity and SIM. The corresponding weights are the feature weights in step (4). The weighted fusion formula is as follows:

[0036]

[0037] In the formula: SIM represents the sum of weighted feature similarities; w i F represents the feature weight set F of the causal association ontology corresponding to the features calculated in C(U,W). n The corresponding weights.

[0038] (73) Based on the weighted feature similarity and SIM obtained in step (72), the top N graduates with the weighted feature similarity and SIM are set as a group; the employers of the top N graduates with the weighted feature similarity and SIM are used as a candidate recommendation list.

[0039] (8) Weighted merge of the recommendation lists from step (6) and step (7);

[0040] (9) Recommend the list of recommendations obtained in step (8) to the graduating students.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] I. The method of using an ontology-based collaborative filtering model based on causal relationships can effectively express the fusion of information and the influence path of causal relationship rules. It mainly considers the causal relationship between factors influencing employment and employment choices. An ontology model is constructed to represent the influence rules of these factors, allowing for the visualization of their influence mechanisms.

[0043] Second, recommendation based on feature analysis of graduate information can alleviate the cold start problem; analyzing the impact of graduate information on employment choices through causal association rules can systematize employment choices, and the weights of these choices can be determined by the confidence of the association rules. Integrating the employment recommendation method based on ontology collaborative filtering model with causal association rules into the collaborative filtering algorithm can effectively alleviate the interpretability problem of recommendation results.

[0044] Third, using this method for recommendations makes it easier to interpret the results and makes the source of the recommendations traceable. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the process of the present invention;

[0046] Figure 2 This is a schematic diagram of the framework of the present invention;

[0047] Figure 3 This is a schematic diagram of the main body of the present invention. Detailed Implementation

[0048] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments.

[0049] Example 1: As Figure 1 As shown, graduate employment data is obtained, and association rules are extracted using the FP-Growth association rule algorithm. Causal relationships are analyzed, and influencing features are categorized, with confidence levels used as classification feature weights. Based on feature distance calculations, corresponding category feature weights are assigned to recent graduates. The corresponding category features of recent graduates are weighted, and the weighted feature values ​​are compared with the feature values ​​of the recommended employers. The top N similarity units are selected to form a set, and this step is repeated for different feature categories. The similarity between recent graduates and previous graduates is calculated according to their corresponding category feature information. After all categories are calculated, the similarity is weighted and fused based on the corresponding category feature weights in the causal association ontology database. The top N recent graduates with the highest weighted feature similarity are selected to form a set of recommended employers. The recommended employer sets are then merged to form a recommendation list. Based on the above method, the following results are obtained: Figure 2 and Figure 3 The diagram shown is as follows:

[0050] A job recommendation method based on a causal ontology model includes the following steps:

[0051] (1) Obtain personal characteristic information and employment data of previous graduates and personal characteristic information of current graduates, perform data cleaning, data completion, data standardization and uniform standardization of data expression processing, and obtain the characteristic set of influencing factors of previous graduates W, the characteristic set of employment element information of graduates WJ and the characteristic set of influencing factors of current graduates U;

[0052] W{A1, A2, A3, ..., A n}, WJ{B1, B2, B3, ..., B n}, U{E1, E2, E3, ..., E n}

[0053] (2) Using the FP-Growth association rule algorithm, association rules are extracted from the features in the feature set W of previous graduates' influencing factors and all features in the feature set WJ of graduates' employment element information. The formulas for extracting the support and confidence of graduates' influencing factors and employment element information are as follows:

[0054] Support(A n →B)=(A n ∩B) / N

[0055] Confidence (A) n →B)=(Support(A n →B)) / (Support(A n ))

[0056] In the formula: Support(A→B) represents the support level; Confidence(A→B) represents the confidence level; A n B represents a single feature in the feature set W of influencing factors of previous graduates; B represents all elements in the feature set WJ of graduate employment factor information; N represents A n The quantity.

[0057] (3) Assess the credibility of the corresponding impact results of the influencing factors, perform information fusion, take the confidence level as the degree of influence of the graduate influencing factors on employment factors, analyze the causal relationship, and construct a causal relationship ontology model.

[0058] (4) Transform the causal association ontology model into feature weights of different categories, classify the influence results of the factors affecting graduates in the causal association ontology, and assign corresponding weights to the feature values ​​F of each category to obtain the feature weight set F of the causal association ontology.

[0059] F={F1:w1,F2:w2,F3:w3,...,F n :w n}

[0060] In the formula: F n :w n Represents the nth feature F n Multiple corresponding weight values ​​w n .

[0061] (5) After information fusion through causal association ontology, job recommendations are made based on the influencing factors of recent graduates, the feature values ​​of the recommended employers, and the causal association ontology. By comparing the influencing factors of recent graduates with the corresponding information in the causal association ontology model, corresponding weights are assigned to the recent graduates. Using the resulting feature weight information of recent graduates, the feature weights are weighted with the corresponding original feature values ​​to obtain the weighted feature values ​​of the recent graduates. These weighted feature values ​​are then matched bilaterally with the feature information of the recommended employers to obtain a high-similarity candidate recommendation list. The specific steps are as follows:

[0062] (51) Calculate the distance between the feature set U of the influencing factors of fresh graduates and the feature weight set F of the causal relationship ontology. The distance calculation formula is as follows:

[0063]

[0064] In the formula: U i F represents the i-th feature in the feature set U of influencing factors for recent graduates; i The i-th feature in the feature weight set F of the causal association ontology.

[0065] (52) Based on the similarity calculation results of step (51), the feature weights in the causal association ontology with high similarity are assigned to the recent graduates.

[0066] (53) The feature weights w obtained from step (52) for recent graduates n The original values ​​{E1, E2, E3, ..., E...} of the corresponding features in the feature set U of influencing factors for recent graduates. n The weighted eigenvalues ​​are obtained by weighting the eigenvalues, and are called the weighted eigenvalues ​​U. f The calculation formula is as follows:

[0067] U f =E n w n

[0068] In the formula: E n This represents the original value of the nth feature in the feature set U of factors influencing recent graduates; w n The feature weight set F representing the causal relationship ontology and E n Features F with high similarity n The corresponding weights.

[0069] (54) Obtain the characteristic information of the employer to be recommended for employment, which is called the characteristic value J. f The weighted eigenvalues ​​U of recent graduates f The characteristic value J of the unit to be recommended for employment f The similarity calculation is performed using the following formula:

[0070]

[0071] In the formula: P(U f J f ) represents U f and J f The similarity; E represents the expectation;

[0072] (55) Based on the similarity calculation result P(U) obtained in step (54), f J f ), sort by similarity, and form a list of the n most similar employers to be recommended.

[0073] J n =Rank n (P(U f J f ))

[0074] In the formula: J n Rank represents the list of the top n employers with high similarity to the target employer; n Indicates U f and J f The obtained similarity calculation result P(U) f J f The top n employers with the highest similarity after sorting are selected as recommended employers.

[0075] (6) Based on different feature classifications, repeat steps (4) and (5) to merge and deduplicate the recommended employment units obtained from different features, and sort the recommended employment units according to feature importance.

[0076] (7) Based on the influencing factors of current graduates and previous graduates, as well as the causal relationship ontology, job recommendations are made using a collaborative filtering algorithm. The specific steps are as follows:

[0077] (71) Calculate the similarity between the feature set U of the influencing factors of current graduates and the feature set W of the influencing factors of previous graduates according to the feature categories in step (4). The similarity calculation formula is as follows:

[0078]

[0079] In the formula: C(U, W) represents the similarity between the feature set U of influencing factors for current graduates and the feature set W of influencing factors for previous graduates; UW represents the inner product of the feature vectors of U and W; num(U∩W) represents the number of identical features in the feature vectors of U and W; ||U|| represents the norm of U, and ||W|| represents the norm of W; num Unum represents the number of features in U; W This represents the number of features in W.

[0080] (72) The similarity obtained in step (71) is classified according to the features. The similarity of the influencing factors of each category is weighted and fused to obtain the weighted feature similarity and SIM. The corresponding weights are the feature weights in step (4). The weighted fusion formula is as follows:

[0081]

[0082] In the formula: SIM represents the sum of weighted feature similarities; w i F represents the feature weight set F of the causal association ontology corresponding to the features calculated in C(U,W). n The corresponding weights.

[0083] (73) Based on the weighted feature similarity and SIM obtained in step (72), the top N graduates with the weighted feature similarity and SIM are set as a group; the employers of the top N graduates with the weighted feature similarity and SIM are used as a candidate recommendation list.

[0084] (8) Weighted merge of the recommendation lists from step (6) and step (7);

[0085] (9) Recommend the list of recommendations obtained in step (8) to the graduating students.

[0086] This invention mines the correlation between graduates' personal characteristics and employment choices, analyzes the causal relationship rules between these characteristics and choices, constructs a causal relationship ontology model, and integrates this model into employment recommendation. It belongs to the fields of causal relationship rule recommendation, knowledge base, and data mining. The causal relationship ontology model in this invention is defined as: a set consisting of graduates as individuals, different categories of characteristics and attributes within each individual, and the influence of these different categories of characteristics and attributes on employment choices. The causal relationship ontology definition in this invention is based on information fusion through the causal relationship relationships mined from association rules, abstracting concepts and entity models from these relationships.

[0087] The present invention has been described through the above examples; however, it should be understood that the above examples are for illustrative purposes only and are not intended to limit the invention to the described embodiments. Furthermore, those skilled in the art will understand that the present invention is not limited to the above embodiments, and many variations and modifications can be made based on the teachings of the present invention, all of which fall within the scope of protection claimed by the present invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A job recommendation method based on a causal ontology model, characterized in that: Includes the following steps: (1) Obtain personal characteristic information and employment data of previous graduates and personal characteristic information of current graduates, perform data cleaning, data completion, and data standardization, and process the data in a unified and standardized manner to obtain the characteristic set of influencing factors of previous graduates W and the characteristic set of employment element information of graduates. and the characteristic set of influencing factors of recent graduates (U); ； This represents the original value of the nth feature in the feature set U of factors influencing recent graduates; (2) Using the FP-Growth association rule algorithm to extract association rules, the features in the feature set W of the influencing factors of previous graduates and the feature set of graduate employment factor information are respectively... All features are used to extract association rules, and the support and confidence scores of graduate influencing factors and employment elements are extracted. (3) Assess the credibility of the corresponding impact results of the influencing factors, perform information fusion, take the confidence level as the degree of influence of the graduate influencing factors on employment factors, analyze the causal relationship, and construct a causal relationship ontology model; (4) The causal association ontology model is transformed into feature weights of different categories. The influence results of the factors affecting graduates in the causal association ontology are classified, and the feature values ​​of each category are assigned corresponding weights to obtain the feature weight set F of the causal association ontology: ； In the formula: Represents the nth feature Multiple corresponding weight values ; (5) After information fusion through causal association ontology, job recommendations are made based on the characteristics of the influencing factors of fresh graduates and the feature values ​​of the employers to be recommended and the causal association ontology. After comparing the influencing factors of fresh graduates with the corresponding information of the causal association ontology model, the corresponding weights are assigned to the fresh graduates. Through the formed feature weight information of fresh graduates, the feature weights are weighted and calculated with the corresponding original feature values. The fresh graduates obtain the weighted feature values ​​and perform bilateral matching with the feature information of the employers to be recommended to obtain a high similarity candidate recommendation list. (6) Based on different feature classifications, repeat steps (4) and (5) to merge and deduplicate the employers to be recommended based on different features, and sort the employers to be recommended based on feature importance; (7) Based on the characteristics of influencing factors of recent graduates and previous graduates, and the causal relationship ontology, job recommendations are made using a collaborative filtering algorithm: (8) Weighted merge of the recommendation lists from step (6) and step (7); (9) Recommend the list of recommendations obtained in step (8) to the graduating students.

2. The job recommendation method based on a causal ontology model according to claim 1, characterized in that: The specific steps of step (5) are as follows: (51) Calculate the distance between the feature set U of the influencing factors of fresh graduates and the feature weight set F of the causal relationship ontology. The distance calculation formula is as follows: ； In the formula: This represents the first characteristic in the characteristic set U of factors influencing recent graduates. i One feature; The first feature weight set F representing the causal relationship ontology is... i One feature; （ 52) Based on the similarity calculation results of step (51), assign the feature weights of the causal association ontology with high similarity to the recent graduates; (53) Feature weights obtained from step (52) for recent graduates The original values ​​of the corresponding features in the feature set U of the influencing factors of recent graduates The weighted eigenvalues ​​are obtained by weighting, and are called weighted eigenvalues. The calculation formula is as follows: ； In the formula: The feature weight set F representing the causal relationship ontology is related to... Features with high similarity The corresponding weights; (54) Obtain the characteristic information of the employer to be recommended for employment, which is called the characteristic value. The weighted eigenvalues ​​of recent graduates The characteristic values ​​of the employers to be recommended The similarity calculation is performed using the following formula: ； In the formula: express and The similarity; E represents the expectation; (55) Based on the similarity calculation results obtained in step (54) Sort the companies by similarity and create a list of the top m most similar employers. ； In the formula: This represents a list of the top m employers with high similarity to the target companies for job recommendations. Indicates to and The obtained similarity calculation results The top m employers with the highest similarity after sorting are selected for job recommendations.

3. The employment recommendation method based on a causal ontology model according to claim 1, characterized in that: The specific steps of step (7) are as follows: (71) Calculate the similarity between the feature set U of the influencing factors of current graduates and the feature set W of the influencing factors of previous graduates according to the feature categories in step (4). The similarity calculation formula is as follows: ； In the formula: This represents the similarity between the characteristic set U of influencing factors for current graduates and the characteristic set W of influencing factors for previous graduates; UW Represents the inner product of the feature set vectors of U and the feature set vectors of W; This represents the number of identical features in the feature set vectors of U and W; Denotes the norm of U. Denotes the norm of W; Indicates the number of features in U; This represents the number of features in W; (72) Based on the feature classification, the similarity obtained in step (71) is weighted and fused according to the similarity of the influencing factors of each category to obtain the weighted feature similarity and The corresponding weights are the feature weights in step (4), and the weighted fusion formula is as follows: ； In the formula: The sum represents the weighted feature similarity. The feature weight set F representing the causal relationship ontology is related to... The features calculated in The corresponding weights; (73) Based on the weighted feature similarity obtained in step (72) and The set of previous graduates with weighted feature similarity and the top N graduates is used as a group; the list of employers of the previous graduates with weighted feature similarity and the top N graduates is used as a candidate recommendation list.

Images