Part-time job recommendation system based on big data analysis

Abstract

The present invention discloses a part-time job recommendation system for personnel based on big data analysis, which belongs to the technical field of big data analysis and comprises a data acquisition module, a data analysis module, a multi-dimensional feature fusion module, a portrait construction module, a situational awareness module, a trend prediction module, an intelligent recommendation module and a real-time feedback adjustment module. The present invention can collect user feedback on recommendation results in real time through the real-time feedback adjustment module, and dynamically adjust the feature weight vector, so as to quickly respond to market changes and user needs and improve dynamic adaptability; the intelligent recommendation module combines user portraits, part-time job portraits, real-time situational awareness and future trend prediction to provide users with personalized part-time job recommendations, which not only takes into account the user's current needs and preferences, but also predicts future market trends and changes in user needs, thereby improving the accuracy of recommendations.

Description

Technical Field

[0001] The present invention belongs to the technical field of big data analysis, and specifically refers to a personnel part-time job recommendation system based on big data analysis. Background Art

[0002] With the rapid development of big data technology, enterprises and individuals have an increasing demand for accurate and efficient part-time job recommendation systems. Traditional part-time job recommendation methods often rely on manual screening or simple matching algorithms, which are inefficient and difficult to meet personalized needs.

[0003] However, the existing part-time job recommendation system still has certain defects. The existing part-time job recommendation system cannot quickly respond to market changes and users' immediate needs, and adjust the recommendation results accordingly. It lacks real-time perception of users' current status, resulting in a mismatch between the recommended positions and users' current actual conditions. It fails to fully consider users' multi-dimensional characteristics and the specific requirements of part-time jobs, resulting in a large deviation between the recommendation results and users' actual needs. It is difficult for users to actively participate in the recommendation process, express their needs and preferences, and provide timely feedback on their satisfaction with the recommendation results. Therefore, a part-time job recommendation system based on big data analysis is proposed. Summary of the invention

[0004] The purpose of the present invention is to provide a part-time job recommendation system based on big data analysis to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a part-time job recommendation system for personnel based on big data analysis, comprising a data collection module, a data analysis module, a multi-dimensional feature fusion module, a portrait construction module, a context perception module, a trend prediction module, an intelligent recommendation module and a real-time feedback adjustment module;

[0006] Wherein, the data acquisition module is wirelessly connected to the data analysis module, and the data analysis module is used to pre-process the data collected by the data acquisition module and extract key features;

[0007] The data analysis module is wirelessly connected to the multi-dimensional feature fusion module, and the multi-dimensional feature fusion module fuses the key features extracted by the data analysis module to form a more comprehensive multi-dimensional feature;

[0008] The multi-dimensional feature fusion module is wirelessly connected to the portrait construction module and the trend prediction module. The portrait construction module constructs user portraits and part-time job portraits according to the multi-dimensional feature vectors; the trend prediction module predicts the future development trend of the part-time job market according to the multi-dimensional features.

[0009] The data analysis module is wirelessly connected to the situation perception module, and the situation perception module perceives the real-time status and situation of the personnel according to the real-time data of the data analysis module;

[0010] The intelligent recommendation module is wirelessly connected to the situational perception module, the portrait construction module and the trend prediction module. The intelligent recommendation module performs part-time job matching according to the situational perception results, user portraits, part-time job portraits and trend prediction results.

[0011] The multi-dimensional feature fusion module performs fusion according to the key features extracted by the data analysis module. Let X = [X1, X2, ..., X i ] is the feature matrix after data analysis module processing and key feature extraction, X i represents the eigenvector, let W = [w1,w2,...,w i ] is the feature weight vector, w i Represents the weight of the i-th feature. Each feature vector is weighted by the feature weight vector. The weighted feature vector is X wd =w i ·X i , feature vector fusion is performed according to the weighted feature vector, and the multi-dimensional feature vector formula is:

[0012]

[0013] In the formula, X fd represents the fused feature vector, X wd,i represents the weighted ith feature vector, and n represents the total number of features. The feature weight vector dynamically adjusts the weight according to real-time feedback, and updates the weight through gradient descent. The implementation formula is:

[0014]

[0015] w i =w i +Δw i ,

[0016] In the formula, η represents the learning rate, which controls the weight update speed, and L represents the loss function.

[0017] The portrait construction module constructs user portraits and part-time job portraits according to the multidimensional feature vectors. Suppose m feature vectors are selected from the multidimensional feature vectors to construct user portraits. i Get an m-dimensional user feature vector X ui =[x 1i ,x 2i ,...,x mi ], the user feature vector is standardized and the formula is: μ j and σ j Respectively represent the mean and standard deviation of the jth feature among all users;

[0018] Map each feature j to a label t j , and assign a weight w to each label j , user u i The label vector T ui Represented as T ui =[t 1i w1,t 2i w2,...,t ji w j ];

[0019] Suppose m feature vectors are selected from the multidimensional feature vector to construct a part-time job profile. k Get an m-dimensional part-time job feature vector X pk =[x 1k ,x 2k ,...,x mk ], after normalizing the user feature vector, Then assign the label t j ′ and calculate the weight w j ′, part-time job p k The label vector T pk Represented as T pk =[t1′ k w1′,t′ 2k w′2,...,t′ jk w′ j ].

[0020] Among them, similarity matching is performed based on the constructed user portrait and part-time job portrait, and the matching degree between the user and the part-time job is calculated by cosine similarity. The implementation formula is:

[0021]

[0022] In the formula, T ui ·T pk The dot product of the label vectors representing users and part-time jobs, ||T ui || represents the length of the user's label vector, ||T pk || represents the length of the label vector for part-time positions.

[0023] Through the multi-dimensional feature fusion module, the multi-dimensional features of users and part-time positions are fused to form a more comprehensive multi-dimensional feature, making the portrait construction more accurate, thereby improving the accuracy of recommendations.

[0024] The trend prediction module performs trend prediction based on the multidimensional feature vector of the multidimensional feature fusion module. Assume a basic time series prediction model as f bs (Y t )=α·Y t +(1-α)·f bs (Y t-1 ), α represents the smoothing coefficient, (Y t-1 ) represents the time series value of the previous moment. The prediction model is enhanced by the multidimensional feature vector, and the multidimensional feature vector is combined with the time series prediction model to achieve:

[0025] f ed (Y t .X fd ) = f bs (Y t )+β·g(X fd ),

[0026] In the formula, g(X fd ) represents the conversion of the multidimensional feature vector into a numerical representation, reflecting the influence of the multidimensional feature vector on the time series prediction, β represents the weight factor, g(X fd ) is expressed as:

[0027]

[0028] In the formula, N represents the number of features, w i Represents the weight corresponding to each feature, x fd,i Represents the value of the i-th feature.

[0029] The time series prediction model evaluates the prediction error of the time series prediction model through the loss function, and the implementation formula is:

[0030]

[0031] In the formula, T represents the number of time steps in the training data, and Y t+1 Represents the actual value of the next moment, and the optimal parameters α, β, w are obtained by minimizing the loss function through gradient descent i After obtaining the optimal parameters, the prediction is performed through the time series prediction model. The prediction formula is:

[0032] The context perception module acquires data streams in real time according to the data analysis module and extracts context features. Suppose the context features include position p, time t and user behavior b, the context feature vector is x=(p, t, b), and the weight added to each context feature vector is w=(w p ,w t ,w b), context recognition is performed through the inner product of the weighted feature vector and the context feature weight vector, and the implementation formula is:

[0033] s(x)=w T x=w p ·p+w t ·t+w b b,

[0034] In the formula, s(x) represents the situational score, and the sorting is performed according to the situational score.

[0035] The context-aware module obtains data streams in real time and extracts context features, such as location, time, and user behavior, to perform context recognition, making the recommendation results more in line with the user's current actual context and needs, thereby improving the pertinence and practicality of the recommendations.

[0036] The intelligent recommendation module makes intelligent recommendations based on the user portrait, part-time job portrait, user, part-time job matching, prediction results of the trend prediction module and situational scores. The final recommendation score is:

[0037] R(ui,pk)=w1·S(ui,pk)+w2·Y t+1 +w3·s(x)+w4·F ui ,

[0038] In the formula, R(ui,pk) represents the final recommendation score list, F ui Represents the user feedback score.

[0039] The intelligent recommendation module comprehensively considers multiple factors such as user portrait, part-time job portrait, matching degree, trend prediction results and situational score to make intelligent recommendations, making the recommendation results more comprehensive and accurate.

[0040] The real-time feedback adjustment module collects user feedback on the recommendation results in real time, and transmits the user feedback result data to the data analysis module to adjust the feature weight vector, thereby updating the recommendation results in real time.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. In order to solve the problems of real-time, inaccurate and real-time dynamic adjustment of part-time job recommendations, and to improve the real-time, accuracy and real-time dynamic adjustment of part-time job recommendations, the present invention provides users with personalized part-time job recommendations through an intelligent recommendation module combined with user portraits, part-time job portraits, real-time situation perception and future trend prediction, which not only considers the current needs and preferences of users, but also predicts future market trends and changes in user needs, thereby improving the accuracy of recommendations;

[0043] 2. The present invention fuses the multi-dimensional features of users and part-time jobs through a multi-dimensional feature fusion module to form a more comprehensive feature vector. The portrait construction module constructs user portraits and part-time job portraits based on these feature vectors, making the recommendation more personalized and able to more accurately match user needs and job requirements;

[0044] 3. The present invention obtains the user's status and situation in real time through the situation perception module, and collects the user's feedback on the recommendation results through the real-time feedback adjustment module, so that the recommendation process has real-time and dynamic characteristics. The trend prediction module enhances the time series prediction model according to the multi-dimensional feature vector, which improves the prediction accuracy of the future part-time market development trend;

[0045] 4. The present invention can collect user feedback on recommendation results in real time through a real-time feedback adjustment module, and dynamically adjust the feature weight vector, so as to quickly respond to market changes and user needs and improve the dynamic adaptability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a structural diagram of a part-time job recommendation system for personnel based on big data analysis of the present invention;

[0047] Figure 2 This is an operation flow chart of a portrait construction module of a part-time job recommendation system for personnel based on big data analysis of the present invention;

[0048] Figure 3 This is an operation flow chart of a trend prediction module of a part-time job recommendation system for personnel based on big data analysis of the present invention;

[0049] Figure 4 This is an operation flow chart of the context awareness module of the part-time job recommendation system based on big data analysis of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Example

[0052] See also Figure 1-Figure 4 As shown, the present invention provides a technical solution: including a data acquisition module, a data analysis module, a multi-dimensional feature fusion module, a portrait construction module, a situational awareness module, a trend prediction module, an intelligent recommendation module and a real-time feedback adjustment module;

[0053] Wherein, the data acquisition module is wirelessly connected to the data analysis module, and the data analysis module is used to pre-process the data collected by the data acquisition module and extract key features;

[0054] The data analysis module is wirelessly connected to the multi-dimensional feature fusion module, and the multi-dimensional feature fusion module fuses the key features extracted by the data analysis module to form a more comprehensive multi-dimensional feature;

[0055] The multi-dimensional feature fusion module is wirelessly connected to the portrait construction module and the trend prediction module. The portrait construction module constructs user portraits and part-time job portraits according to the multi-dimensional feature vectors; the trend prediction module predicts the future development trend of the part-time job market according to the multi-dimensional features.

[0056] The data analysis module is wirelessly connected to the situation perception module, and the situation perception module perceives the real-time status and situation of the personnel according to the real-time data of the data analysis module;

[0057] The intelligent recommendation module is wirelessly connected to the situational perception module, the portrait construction module and the trend prediction module. The intelligent recommendation module performs part-time job matching according to the situational perception results, user portraits, part-time job portraits and trend prediction results.

[0058] The multi-dimensional feature fusion module performs fusion according to the key features extracted by the data analysis module. Let X = [X1, X2, ..., X i ] is the feature matrix after data analysis module processing and key feature extraction, X i represents the eigenvector, let W = [w1,w2,...,w i ] is the feature weight vector, w i Represents the weight of the i-th feature. Each feature vector is weighted by the feature weight vector. The weighted feature vector is X wd =w i ·X i , feature vector fusion is performed according to the weighted feature vector, and the multi-dimensional feature vector formula is:

[0059]

[0060] In the formula, X fd represents the fused feature vector, X wd,i represents the weighted ith feature vector, and n represents the total number of features. The feature weight vector dynamically adjusts the weight according to real-time feedback, and updates the weight through gradient descent. The implementation formula is:

[0061]

[0062] w i=w i +Δw i ,

[0063] In the formula, η represents the learning rate, which controls the weight update speed, and L represents the loss function.

[0064] The portrait construction module constructs user portraits and part-time job portraits according to the multidimensional feature vectors. Suppose m feature vectors are selected from the multidimensional feature vectors to construct user portraits. i Get an m-dimensional user feature vector X ui =[x 1i ,x 2i ,...,x mi ], the user feature vector is standardized and the formula is: μ j and σ j Respectively represent the mean and standard deviation of the jth feature among all users;

[0065] Map each feature j to a label t j , and assign a weight w to each label j , user u i The label vector T ui Represented as T ui =[t 1i w1,t 2i w2,...,t ji w j ];

[0066] Suppose m feature vectors are selected from the multidimensional feature vector to construct a part-time job profile. k Get an m-dimensional part-time job feature vector X pk =[x 1k ,x 2k ,...,x mk ], after normalizing the user feature vector, Then assign the label t j ′ and calculate the weight w j ′, part-time job p k The label vector T pk Represented as T pk =[t1′ k w1′,t′ 2k w′2,...,t′ jk w′ j ].

[0067] Among them, similarity matching is performed based on the constructed user portrait and part-time job portrait, and the matching degree between the user and the part-time job is calculated by cosine similarity. The implementation formula is:

[0068]

[0069] In the formula, T ui ·T pk The dot product of the label vectors representing users and part-time jobs, ||T ui || represents the length of the user's label vector, ||T pk || represents the length of the label vector for part-time positions.

[0070] Through the multi-dimensional feature fusion module, the multi-dimensional features of users and part-time positions are fused to form a more comprehensive multi-dimensional feature, making the portrait construction more accurate, thereby improving the accuracy of recommendations.

[0071] The trend prediction module performs trend prediction based on the multidimensional feature vector of the multidimensional feature fusion module. Assume a basic time series prediction model as f bs (Y t )=α·Y t +(1-α)·f bs (Y t-1 ), α represents the smoothing coefficient, (Y t-1 ) represents the time series value of the previous moment. The prediction model is enhanced by the multidimensional feature vector, and the multidimensional feature vector is combined with the time series prediction model to achieve:

[0072] f ed (Y t .X fd ) = f bs (Y t )+β·g(X fd ),

[0073] In the formula, g(X fd ) represents the conversion of the multidimensional feature vector into a numerical representation, reflecting the influence of the multidimensional feature vector on the time series prediction, β represents the weight factor, g(X fd ) is expressed as:

[0074]

[0075] In the formula, N represents the number of features, w i Represents the weight corresponding to each feature, x fd,i Represents the value of the i-th feature.

[0076] The time series prediction model evaluates the prediction error of the time series prediction model through the loss function, and the implementation formula is:

[0077]

[0078] In the formula, T represents the number of time steps in the training data, and Y t+1 Represents the actual value of the next moment, and the optimal parameters α, β, w are obtained by minimizing the loss function through gradient descent i After obtaining the optimal parameters, the prediction is performed through the time series prediction model. The prediction formula is:

[0079] The context perception module acquires data streams in real time according to the data analysis module and extracts context features. Suppose the context features include position p, time t and user behavior b, the context feature vector is x=(p, t, b), and the weight added to each context feature vector is w=(w p ,w t ,w b ), context recognition is performed through the inner product of the weighted feature vector and the context feature weight vector, and the implementation formula is:

[0080] s(x)=w T x=w p ·p+w t ·t+w b b,

[0081] In the formula, s(x) represents the situational score, and the sorting is performed according to the situational score.

[0082] The context-aware module obtains data streams in real time and extracts context features, such as location, time, and user behavior, to perform context recognition, making the recommendation results more in line with the user's current actual context and needs, thereby improving the pertinence and practicality of the recommendations.

[0083] The intelligent recommendation module makes intelligent recommendations based on the user portrait, part-time job portrait, user, part-time job matching, prediction results of the trend prediction module and situational scores. The final recommendation score is:

[0084] R(ui,pk)=w1·S(ui,pk)+w2·Y t+1 +w3·s(x)+w4·F ui ,

[0085] In the formula, R(ui,pk) represents the final recommendation score list, F ui Represents the user feedback score.

[0086] The intelligent recommendation module comprehensively considers multiple factors such as user portrait, part-time job portrait, matching degree, trend prediction results and situational score to make intelligent recommendations, making the recommendation results more comprehensive and accurate.

[0087] The real-time feedback adjustment module collects user feedback on the recommendation results in real time, and transmits the user feedback result data to the data analysis module to adjust the feature weight vector, thereby updating the recommendation results in real time.

[0088] In this example, specifically: the portrait construction module constructs user portraits and part-time job portraits according to the multidimensional feature vectors. Suppose m feature vectors are selected from the multidimensional feature vectors to construct user portraits. i Get an m-dimensional user feature vector X ui =[x 1i ,x 2i ,...,x mi ], the user feature vector is standardized and the formula is: μ j and σ j Respectively represent the mean and standard deviation of the jth feature among all users;

[0089] Map each feature j to a label t j , and assign a weight w to each label j , user u i The label vector T ui Represented as T ui =[t 1i w1,t 2i w2,...,t ji w j ];

[0090] Suppose m feature vectors are selected from the multidimensional feature vector to construct a part-time job profile. k Get an m-dimensional part-time job feature vector X pk =[x 1k ,x 2k ,...,x mk ], after normalizing the user feature vector, Then assign the label t j ′ and calculate the weight w j ′, part-time job p k The label vector T pk Represented as T pk =[t1′ k w1′,t′ 2k w′2,...,t′ jk w′ j ].

[0091] Among them, similarity matching is performed based on the constructed user portrait and part-time job portrait, and the matching degree between the user and the part-time job is calculated by cosine similarity. The implementation formula is:

[0092]

[0093] In the formula, T ui ·T pk The dot product of the label vectors representing users and part-time jobs, ||T ui || represents the length of the user's label vector, ||T pk || represents the length of the label vector for part-time positions.

[0094] In this example, specifically: the trend prediction module performs trend prediction based on the multidimensional feature vector of the multidimensional feature fusion module, assuming a basic time series prediction model as f bs (Y t )=α·Y t +(1-α)·f bs (Y t-1 ), α represents the smoothing coefficient, (Y t-1 ) represents the time series value of the previous moment. The prediction model is enhanced by the multidimensional feature vector, and the multidimensional feature vector is combined with the time series prediction model to achieve:

[0095] f ed (Y t .X fd ) = f bs (Y t )+β·g(X fd ),

[0096] In the formula, g(X fd ) represents the conversion of the multidimensional feature vector into a numerical representation, reflecting the influence of the multidimensional feature vector on the time series prediction, β represents the weight factor, g(X fd ) is expressed as:

[0097]

[0098] In the formula, N represents the number of features, w i Represents the weight corresponding to each feature, x fd,i Represents the value of the i-th feature.

[0099] The time series prediction model evaluates the prediction error of the time series prediction model through the loss function, and the implementation formula is:

[0100]

[0101] In the formula, T represents the number of time steps in the training data, and Y t+1 Represents the actual value of the next moment, and the optimal parameters α, β, w are obtained by minimizing the loss function through gradient descent iAfter obtaining the optimal parameters, the prediction is performed through the time series prediction model. The prediction formula is:

[0102] In this example, specifically: the context perception module obtains the data stream in real time according to the data analysis module and extracts context features. Suppose the context features include position p, time t and user behavior b, the context feature vector is x=(p, t, b), and the weight added to each context feature vector is w=(w p ,w t ,w b ), context recognition is performed through the inner product of the weighted feature vector and the context feature weight vector, and the implementation formula is:

[0103] s(x)=w T x=w p ·p+w t ·t+w b b,

[0104] In the formula, s(x) represents the situational score, and the sorting is performed according to the situational score.

[0105] In this example, specifically: the intelligent recommendation module makes intelligent recommendations based on the user portrait, part-time job portrait, user, part-time job matching, the prediction results of the trend prediction module and the situational score, and the final recommendation score is:

[0106] R(ui,pk)=w1·S(ui,pk)+w2·Y t+1 +w3·s(x)+w4·F ui ,

[0107] In the formula, R(ui,pk) represents the final recommendation score list, F ui Represents the user feedback score.

[0108] Working principle: The data collection module collects user information and part-time job information from big data. The data analysis module needs to perform pre-processing such as cleaning, deduplication, and formatting on the collected raw data, and conduct in-depth analysis on the pre-processed data to extract key features.

[0109] The key features extracted by the data analysis module are fused through the multidimensional feature fusion module, and each feature vector is weighted by the feature weight vector and fused into a more comprehensive multidimensional feature vector. The feature weight vector will be dynamically adjusted according to real-time feedback. The portrait construction module constructs user portraits and part-time job portraits through the feature vectors output by the multidimensional feature fusion module. The user portrait and part-time job portrait are generated through standardization, feature mapping to labels, and weight allocation steps respectively; the trend prediction module predicts the future development trend of the part-time market based on the multidimensional feature vectors of the multidimensional feature fusion module, and dynamically tracks and predicts the demand of the part-time market by constructing a time series prediction model and enhancing the model in combination with the multidimensional feature vector. The situational awareness module obtains the user's data stream in real time, extracts situational features, and weights and identifies the situational features; the intelligent recommendation module conducts a comprehensive evaluation based on the user portrait, part-time job portrait, the matching degree between the user and the job, the trend prediction results, and the situational score to generate the final recommendation list. The real-time feedback adjustment module is responsible for collecting user feedback on the recommendation results and transmitting the feedback data back to the data analysis module. The data analysis module dynamically adjusts the feature weight vector based on user feedback and updates the recommendation results in real time.

[0110] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0111] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A part-time job recommendation system based on big data analysis, characterized by: It includes data collection module, data analysis module, multi-dimensional feature fusion module, portrait construction module, situational awareness module, trend prediction module, intelligent recommendation module and real-time feedback adjustment module; Wherein, the data acquisition module is wirelessly connected to the data analysis module, and the data analysis module is used to pre-process the data collected by the data acquisition module and extract key features; The data analysis module is wirelessly connected to the multi-dimensional feature fusion module, and the multi-dimensional feature fusion module fuses the key features extracted by the data analysis module to form a more comprehensive multi-dimensional feature; The multi-dimensional feature fusion module is wirelessly connected to the portrait construction module and the trend prediction module. The portrait construction module constructs user portraits and part-time job portraits according to the multi-dimensional feature vectors; the trend prediction module predicts the future development trend of the part-time job market according to the multi-dimensional features. The data analysis module is wirelessly connected to the situation perception module, and the situation perception module perceives the real-time status and situation of the personnel according to the real-time data of the data analysis module; The intelligent recommendation module is wirelessly connected to the situational awareness module, the portrait construction module and the trend prediction module, and the intelligent recommendation module performs part-time job matching based on the situational awareness results, the user portrait, the part-time job portrait and the trend prediction results; The portrait construction module constructs user portraits and part-time job portraits according to the multidimensional feature vectors. Suppose m feature vectors are selected from the multidimensional feature vectors to construct user portraits. Get an m-dimensional user feature vector The user feature vector is standardized and the formula is: , and Respectively represent the mean and standard deviation of the jth feature among all users; Map each feature j to a label , and assign a weight to each label ,user The label vector Expressed as ; Suppose m feature vectors are selected from the multidimensional feature vector to construct a part-time job profile. Get an m-dimensional feature vector of part-time jobs , after normalizing the user feature vector, , then each feature Map to a label And calculate the weight , part-time positions The label vector Expressed as ; Similarity matching is performed based on the constructed user portrait and part-time job portrait, and the matching degree between the user and the part-time job is calculated by cosine similarity. The implementation formula is: ， In the formula, The dot product of the label vectors representing users and part-time jobs, Represents the length of the user's label vector, The length of the label vector representing part-time positions; The trend prediction module performs trend prediction based on the multidimensional feature vector of the multidimensional feature fusion module. Suppose a basic time series prediction model is , α represents the smoothing coefficient, Represents the time series value of the previous moment. The prediction model is enhanced by the multidimensional feature vector. The multidimensional feature vector is combined with the time series prediction model to achieve: ， In the formula, It means converting the multidimensional feature vector into a numerical representation, reflecting the influence of the multidimensional feature vector on time series prediction. β represents the weight factor. It is expressed as: ， In the formula, N represents the number of features, Represents the weight corresponding to each feature, represents the value of the i-th feature; The time series prediction model evaluates the prediction error of the time series prediction model through the loss function, and the implementation formula is: ， In the formula, T represents the number of time steps in the training data, Represents the actual value of the next moment, and the optimal parameter is obtained by minimizing the loss function through gradient descent After obtaining the optimal parameters, the prediction is performed through the time series prediction model. The prediction formula is: ; The context perception module acquires data streams in real time according to the data analysis module and extracts context features. Suppose the context features include position p, time t and user behavior b, and the context feature vector is x=(p, t, b). The weight added to each context feature vector is , context recognition is performed by the inner product of the weighted feature vector and the context feature weight vector, and the implementation formula is: ， In the formula, s(x) represents the situational score, and the sorting is performed according to the situational score.

2. The part-time job recommendation system based on big data analysis according to claim 1 is characterized by: The multi-dimensional feature fusion module fuses the key features extracted by the data analysis module. It is the feature matrix after data analysis module processing and key feature extraction. represents the feature vector, let is the feature weight vector, Represents the weight of the i-th feature. Each feature vector is weighted by the feature weight vector. The weighted feature vector is , feature vector fusion is performed according to the weighted feature vector, and the multi-dimensional feature vector formula is: ， In the formula, represents the fused feature vector, represents the weighted ith eigenvector, and n represents the total number of features; feature The weight vector dynamically adjusts the weight according to real-time feedback and updates the weight through gradient descent. The implementation formula is: ， ， In the formula, η represents the learning rate, which controls the weight update speed, and L represents the loss function.

3. The part-time job recommendation system based on big data analysis according to claim 1 is characterized by: The intelligent recommendation module makes intelligent recommendations based on user portraits, part-time job portraits, user, part-time job matching, prediction results of the trend prediction module and situational scores. The final recommendation score is: ， In the formula, Represents the final recommendation score list, Displays user feedback scores.

4. The part-time job recommendation system based on big data analysis according to claim 1 is characterized by: The real-time feedback adjustment module collects user feedback on the recommendation results in real time, and transmits the user feedback result data to the data analysis module to adjust the feature weight vector, thereby updating the recommendation results in real time.

Images