High-quality occupational recommendation system based on big data analysis

Abstract

The invention relates to the technical field of big data analysis, and provides a high-quality occupational recommendation system based on big data analysis, which collects data of social media, an online learning platform and the like from multiple sources on the data level, and broadens the information dimension; in analysis, real-time stream processing and an innovation algorithm accurately capture user interests and market demand changes; in the modeling aspect, deep learning and sentiment analysis are combined to construct a model which is more practical; during recommendation, a unique algorithm and an exploration mechanism balance personalization and diversification; the feedback module can quickly optimize according to user feedback; on the whole, the accuracy, timeliness and diversity of occupational recommendation are greatly improved, the problem that a traditional system is difficult to adapt to dynamic changes is effectively solved, and higher job hunting and recruitment efficiency is brought to users and enterprises.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis, and specifically to a high-quality career recommendation system based on big data analysis. Background Art

[0002] With the wide application of big data technology in the field of career recommendation, traditional career recommendation systems provide career recommendation services for users by leveraging user behavior data and job information.

[0003] However, there are many key problems in the existing technology that need to be solved urgently. First, the data source is single, mainly relying on the limited data of job hunting platforms, making it difficult to comprehensively understand users' career interests and not fully reflecting market demands. Second, it is unable to capture the dynamic changes of users' career interests and market demands in real time. Users' interests can change at any time due to learning new skills, participating in industry activities, etc., and market demands also change continuously with industry development, policy adjustments, etc. However, the existing systems are difficult to respond quickly, resulting in lagging and inaccurate recommendation results, seriously affecting users' job hunting experience and efficiency. In addition, there are limitations in the recommendation algorithm. Either it is overly personalized, narrowing the recommendation scope, or it lacks an effective exploration mechanism and cannot explore potential suitable careers for users.

[0004] Therefore, there is an urgent need to invent a high-quality career recommendation system based on big data analysis, aiming to solve the above-mentioned problems in the existing technology through innovative technical means, achieve accurate, timely, and diversified career recommendations, and improve user satisfaction and the matching efficiency of the employment market. Summary of the Invention

[0005] Technical Problems to be Solved

[0006] In view of the deficiencies of the existing technology, the present invention provides a high-quality career recommendation system based on big data analysis.

[0007] Technical Solutions

[0008] To achieve the above-mentioned solution objectives, the present invention provides the following technical solutions: A high-quality career recommendation system based on big data analysis, characterized by including the following modules:

[0009] S1: Data collection module: Responsible for collecting data related to users' career interests and market job demands from multiple sources, providing a comprehensive and real-time data basis for subsequent analysis and modeling, and ensuring that the system can promptly grasp the latest dynamics.

[0010] S2: Real-time data analysis module: This module performs real-time cleaning and analysis on the collected data, mines the changing trends of users' interest preferences and market job demands, and provides accurate and timely data for subsequent modeling.

[0011] S3: User Interest Modeling Module: Based on the real-time data analysis results, use deep learning algorithms to dynamically model user interests, update and adjust the model according to the user behavior data, and provide a basis for personalized recommendations.

[0012] S4: Market Demand Modeling Module: Use natural language processing technology to perform semantic analysis on job descriptions, construct a market demand model, monitor demand changes through cluster analysis, and provide market demand references for career recommendations.

[0013] S5: Dynamic Recommendation Engine Module: Combine the results of user interest and market demand modeling, use personalized recommendation algorithms to generate a recommendation list, adjust the recommendation results in real time, and introduce an exploration mechanism to avoid over-personalization.

[0014] S6: User Feedback Module: Collect user feedback information on the recommendation results, optimize the user interest model and recommendation algorithms, form a closed-loop optimization mechanism, and improve the recommendation accuracy and user satisfaction.

[0015] Preferably, the multi-source data acquisition module not only collects the traditional behavior data of users on the job search platform, but also obtains relevant user behavior data through cooperation with social media platforms and online learning platforms, and uses web crawler technology to frequently collect job information on recruitment websites, and the collection period does not exceed 15 minutes.

[0016] Preferably, the real-time data analysis module uses real-time stream processing technology to clean the collected data, establishes a data quality monitoring index system to monitor the data quality in real time, and starts the repair process when the data error rate exceeds the set threshold; uses the sliding window technology to analyze the user behavior data to calculate the interest index, and adopts the autoregressive moving average model to predict the changes in market job demands.

[0017] Preferably, the user interest modeling module, based on the real-time data analysis results, uses a long short-term memory network to build a user interest model, processes the user real-time behavior data sequence through the structures of forget gate, input gate, output gate, candidate memory unit and memory unit to update the model; introduces sentiment analysis technology to perform sentiment analysis on the user's social media career-related content, and integrates the sentiment tendency score into the interest model.

[0018] Preferably, the market demand modeling module uses natural language processing technology to perform semantic analysis on job descriptions, extracts key information through word segmentation, word embedding, and text classification to construct a market demand model, uses the K-means clustering algorithm to perform cluster analysis on the job demand feature vectors, and judges the market demand changes and updates the model by monitoring the changes in the cluster center vector and the number of samples within the cluster.

[0019] Preferably, the dynamic recommendation engine module combines the results of user interest modeling and market demand modeling, implements a recommendation algorithm that combines matrix factorization and deep learning, solves the user feature matrix and the occupation feature matrix by minimizing the loss function to calculate the matching degree score to generate a recommendation list, and introduces an exploration factor to control the proportion of personalized recommendations and exploratory recommendations.

[0020] Preferably, the user feedback module collects in real time the behavioral data of users' clicks, favorites, and resume submissions on the recommendation results, as well as text evaluation feedback. By calculating the matching degree between the recommended occupation and the user interest model, if the matching degree is lower than expected, the reasons are analyzed to adjust the parameters of the user interest model and the recommendation algorithm.

[0021] Preferably, the multi-source data acquisition module obtains data from the social media platform through the OAuth authorization mechanism. The obtained data includes the set of user career dynamic information and the set of industry topic discussion content; it cooperates with the online learning platform to obtain the set of user learning course names, learning durations, and course completion progress.

[0022] Preferably, in the real-time data analysis module, the autoregressive moving average model is where D(t) is the sequence of job demand quantities at time t, p and q are the orders, is the autoregressive coefficient, θ j is the moving average coefficient, ò t is the white noise sequence.

[0023] Preferably, the loss function in the dynamic recommendation engine module is where Ω is the set of user-occupation pairs with known ratings, r ui is the actual rating of user u for occupation i, is the predicted rating, λ is the regularization parameter, U is the user feature matrix, and V is the occupation feature matrix.

[0024] Beneficial effects

[0025] Compared with the prior art, the present invention provides a high-quality career recommendation system based on big data analysis, having the following beneficial effects:

[0026] 1. The high-quality career recommendation system based on big data analysis, at the data level, collects data from multiple sources such as social media and online learning platforms to broaden the information dimension; in terms of analysis, real-time stream processing and innovative algorithms accurately capture changes in user interests and market demands; in terms of modeling, combines deep learning and sentiment analysis to build a more practical model; when making recommendations, a unique algorithm and exploration mechanism balance personalization and diversification; the feedback module can quickly optimize according to user feedback; overall, it greatly improves the accuracy, timeliness and diversity of career recommendations, effectively solves the problem that traditional systems are difficult to adapt to dynamic changes, and brings higher job search and recruitment efficiency to users and enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic diagram of the system framework of the present invention;

[0028] Figure 2 It is a schematic diagram of the system flow of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0030] Please refer to Figures 1 to 2 , the present invention proposes a high-quality career recommendation system based on big data analysis, including the following:

[0031] S1: Data collection module

[0032] The data collection module is responsible for collecting data related to users' career interests and market job demands from multiple sources, providing a comprehensive and real-time data basis for subsequent analysis and modeling, and ensuring that the system can promptly grasp the latest dynamics.

[0033] 1. Social media data collection

[0034] Sign a cooperation agreement with social media platforms (such as LinkedIn, Weibo, etc.) and obtain data according to the OAuth authorization mechanism. Let the set of user career dynamic information obtained be D social , and the set of industry topic discussion content be T social . These data can intuitively reflect the career interests shown by users on social media.

[0035] 2. Online learning platform data collection

[0036] Cooperate with online learning platforms to obtain data. The set of user learning course names is C name, the set C of learning durations length , the set C of course completion progress progress , which can be used to infer the direction of users' vocational skill improvement and their effort level.

[0037] 3. Recruitment website data collection

[0038] Using web crawler technology, collect recruitment website data at a 15-minute interval. Collect the job posting time t post , the set S of skill requirements = {s1, s2, …, s n}, salary w and other information, providing data support for analyzing market job demands.

[0039] S2: Real-time data analysis module

[0040] The real-time data analysis module is like the "brain" of the system, which performs real-time cleaning and in-depth analysis on the massive, complex and dynamically changing data collected, and extracts the subtle change trends of users' interest preferences and the significant dynamic changes of market job demands, providing accurate, timely and high-quality data support for user interest modeling and market demand modeling, ensuring that the subsequent recommendation model can make scientific and reasonable recommendation decisions based on accurate information.

[0041] 1. Data cleaning

[0042] Build a high-performance cluster using the industry-leading real-time stream processing technology Apache Flink. In the cluster environment, carefully write the code logic for data cleaning, and remove noise data, duplicate data and missing values through a series of strict data processing rules. To effectively monitor data quality, establish a comprehensive and detailed data quality monitoring index system, and define the data error rate as a key indicator, where N total represents the total amount of data, and N error represents the amount of error data. When ò exceeds the pre-set threshold (such as 5%), the system will automatically trigger the data repair process, and repair the data through technical means such as data backtracking, intelligent completion and error correction to ensure the accuracy and integrity of the subsequent analysis data.

[0043] 2. User behavior data analysis

[0044] Use the sliding window technology to perform fine-grained real-time statistical analysis on user behavior data. Set the sliding window time length to T, and this time length can be flexibly adjusted according to actual business needs and data characteristics. For example, when analyzing short-term user interest changes, T can be set to 1 hour. Deeply analyze the set C of job categories browsed by users within the window = {c1, c2, …, c m}, including various types of positions browsed by users, such as software development, marketing, financial analysis, etc. The browsing frequency f reflects the number of times a user browses different types of positions per unit time, and the stay time t stay reflects the average time spent by the user when browsing each position. By calculating the user's interest index for different types of positions where f c is the browsing frequency of the user for positions in category c, and t stay,c is the average stay time for positions in category c. For example, if a user frequently browses software development positions in a day and spends a long time on each browse, the calculated interest index for software development positions will be significantly higher than other categories, thus accurately judging the shift in the user's interest preference towards the software development field.

[0045] 3. Market Position Data Analysis

[0046] For market position data, a professional time series model is constructed to accurately predict changes in market demand. Let the sequence of the number of position demands at time t be D(t), and the autoregressive moving average model ARMA(p,q) is used: where p and q are the orders of autoregression and moving average respectively, and their values need to be determined through experiments and algorithm optimization according to the characteristics of historical data and the model fitting effect. is the autoregressive coefficient, and the magnitude and sign of its value reflect the degree and direction of the impact of the number of position demands in the past i periods on the current demand quantity. For example indicates that the number of position demands in the previous period has a strong positive impact on the current. θ j is the moving average coefficient, reflecting the effect of the white noise ò t in the past j periods on the current demand quantity. The white noise ò t is a random sequence with a mean of 0 and a constant variance, representing unpredictable random interference factors, such as sudden industry policy adjustments, short-term impacts of global events on the employment market, etc. Through this model, the trends and random fluctuations of historical data can be comprehensively considered to predict the short-term and long-term change trends of market demand. For example, accurately predicting whether the number of position demands in a certain popular industry will increase or decrease within the next month, as well as the range of increase or decrease.

[0047] S3: User Interest Modeling Module

[0048] The User Interest Modeling module uses cutting-edge deep learning algorithms to dynamically and accurately model user interests based on real-time data analysis. As user behavior data continuously updates, the module can promptly adjust model parameters, deeply explore users' potential career interests, and accurately portray their career interests. This provides the core basis for personalized career recommendations, ensuring that recommendations closely align with users' evolving interests and needs.

[0049] 1. Construction of interest model based on LSTM

[0050] Build a user interest model based on the long short-term memory network (LSTM) in a mature deep learning framework such as TensorFlow or PyTorch. The input of the LSTM unit is the user's real-time behavior data sequence X = {x1, x2, ..., x T}, where x t is the input vector at time t, which may contain multi-dimensional information such as the job categories browsed by the user at that time, the course information learned, and career-related activities on social media. Let the forget gate F t =σ(W f x t +U f h t-1 +b f ), where σ is the sigmoid activation function, and its mathematical expression is It maps the input value to a range between 0 and 1, and controls the degree of retention of the memory unit information at the previous moment through the value output by this function. For example, when the output value of σ is close to 1, it means that more memory unit information at the previous moment is retained. f is the input x t The corresponding weight matrix, whose element value determines the degree of influence of each dimension of the input data on the forget gate; U f is the hidden state h at the previous moment t-1 The corresponding weight matrix reflects the effect of the hidden state on the forget gate at the previous moment; b f Is the bias vector, used to adjust the output threshold of the forget gate. Input gate I t =σ(W i x t +U i h t-1 +b i ), its function is to control the degree to which the current input information enters the memory unit, and its principle is similar to that of the forget gate. Output gate O t =σ(W o x t +U o h t-1 +b o ), determines the information that the memory unit outputs to the hidden state. Candidate memory unit C t= tanh(W c x t + U c h t-1 + b c ), where tanh is the hyperbolic tangent activation function, and its expression is It maps the input value to the range between -1 and 1, which is used to generate candidate memory content and enrich the information in the memory unit. The memory unit C t = F t ⊙ C t-1 + I t ⊙ C t , where ⊙ is element-wise multiplication. Through the control of the forget gate and the input gate, the memory unit is updated to retain useful information and remove outdated information. The hidden state h t = O t ⊙ tanh(C t ), and the final output hidden state is used to represent the user's interest characteristics at this moment. As the user behavior data is continuously updated, the model parameters (i.e., various weight matrices and bias vectors) are continuously adjusted through optimization techniques such as the backpropagation algorithm. For example, when the user frequently browses job positions related to artificial intelligence, the weights corresponding to the features related to the artificial intelligence field in the model will gradually increase during the training process, thereby increasing the weight of the occupations in this field in the recommendation.

[0051] 2. Sentiment Analysis Aided Modeling

[0052] Perform sentiment analysis on the user's social media career-related content. Use natural language processing tools to convert the text into word vectors Judge the sentiment tendency through the sentiment classification model. Let the sentiment tendency score E ∈ [-1, 1], where -1 represents negative and 1 represents positive. Incorporate the score into the interest model to comprehensively depict the user's interests.

[0053] S4: Market Demand Modeling Module

[0054] The market demand modeling module uses advanced natural language processing techniques to perform in-depth semantic analysis on job descriptions, accurately extracts key information to construct a comprehensive and dynamic market demand model, and monitors the changing trend of market demand in real time through clustering analysis, providing a reliable reference for market demand in career recommendation to ensure that the recommended occupations are closely aligned with the actual market demand.

[0055] 1. Semantic Analysis of Job Descriptions

[0056] Use natural language processing (NLP) techniques to systematically and meticulously process job descriptions. First, tokenize the job description text. Through professional tokenization algorithms such as the tokenization method combining dictionary and statistical models, obtain the word sequence W = {w1, w2, …, w N}. Then, each word is mapped to a low - dimensional vector representation through a word embedding model (such as Word2Vec). Through training on a large - scale text corpus, this model makes words with similar semantics close in the vector space. For example, the position - related words "data analyst" and "data analysis specialist" are close in the vector space. Use a text classification algorithm (such as Support Vector Machine - SVM) to classify job descriptions. SVM determines the industry category I and job type T to which a position belongs by finding an optimal hyperplane to separate text data of different categories. For example, accurately classifying a job description into the investment analysis job type in the financial industry. At the same time, extract information such as key skills, knowledge areas, and job responsibilities from the job description through keyword matching and semantic understanding techniques. Let the set of key skills be S = {s1, s2, …, s n}, for example, extract key skills such as Java programming and algorithm design from a software development job description. These key skills are the core reflection of the market demand for this position and provide key data support for subsequent market demand modeling.

[0057] 2. Cluster analysis to monitor demand changes

[0058] Conduct a comprehensive and in - depth cluster analysis on the demand characteristics of different industries and different types of positions. Let the job demand feature vector be where v i is the i - th eigenvalue, and these eigenvalues can be quantified values such as skill requirements (e.g., the proficiency in mastering a certain programming language is quantified as a value from 0 - 10), educational requirements (1 for undergraduate, 2 for master, etc.), work experience requirements (in years), etc. Use the K - means clustering algorithm. The core idea of this algorithm is to divide data points into K clusters. Through continuous iterative optimization, the data points within each cluster have a high degree of similarity, while the data points between different clusters have a low degree of similarity. During the clustering process, calculate the center vector of the cluster where n k is the number of samples in the k - th cluster, and C k is the set of samples in the k - th cluster. Real - time monitor the changes in the center vector of the cluster and the changes in the number of samples within the cluster. When it is found that the characteristics of a certain cluster change significantly (such as the Euclidean distance of the cluster center vector changes exceeding a set threshold, and the Euclidean distance calculation formula is used to measure the distance between two vectors) or the number of samples in a certain cluster increases or decreases sharply, it is judged that the market demand has changed, and the market demand model is updated in a timely manner. For example, if the number of samples in a cluster related to positions in a certain emerging industry increases significantly in a short period of time, it indicates that the market demand for this industry is growing rapidly, and the system needs to adjust the market demand model in a timely manner to reflect this change, such as increasing the recommendation weight for positions related to this industry.

[0059] S5: Dynamic Recommendation Engine Module

[0060] As the core execution unit of the system, the Dynamic Recommendation Engine Module skillfully combines the results of user interest modeling and market demand modeling, and uses innovative personalized recommendation algorithms to generate highly personalized career recommendation lists in real time, and can adjust the recommendation results in real time according to the dynamic changes of user interests and market demands. At the same time, a unique exploration mechanism is introduced to avoid over-personalization, broaden the user's career horizons, and provide users with diversified career recommendations that not only fit their current interests but also have development potential.

[0061] 1. Implementation of Personalized Recommendation Algorithm

[0062] Implement an innovative recommendation algorithm that combines matrix factorization and deep learning in the recommendation engine. Let the user-career rating matrix be R, and the element r in the matrix ui represents the rating of user u for career i (which can be converted into a rating through the user's click, favorite, resume submission and other behaviors, such as click is 1 point, favorite is 2 points, resume submission is 3 points, etc.), and it is decomposed into a user feature matrix U and a career feature matrix V, that is, R≈UV T . Solve U and V by minimizing the loss function . Among them, Ω is the set of user-career pairs with known ratings, that is, the set of career pairs for which the system has obtained clear behavioral feedback from users; r ui is the actual rating of user u for career i, is the predicted rating, and the predicted rating of the user for the unrated career is obtained by multiplying the user feature matrix and the career feature matrix obtained by matrix factorization; λ is the regularization parameter, which is used to prevent overfitting, and its value is optimized and selected within a certain range through methods such as cross-validation. Combine the deep learning model (such as the output of the user interest model and the market demand model) with the matrix factorization result. Let the user interest feature vector be The career demand feature vector is Calculate the matching degree score Sort the careers according to the score. The higher the score, the higher the matching degree between the user and the career, and the higher its position in the recommendation list. For example, if the dimension value related to data analysis in the user interest feature vector is relatively high, and the dimension value of the requirement for data analysis skills in a certain career demand feature vector is also relatively high, then their dot product S will be relatively large, and this career will be preferentially displayed in the recommendation list.

[0063] 2. Introduction of Exploration Mechanism

[0064] To avoid the recommendation results being overly concentrated in the fields where the user has already shown interest, this system introduces an innovative exploration mechanism. Set the exploration factor α ∈ [0, 1] of the recommendation strategy parameter. This factor is used to flexibly control the proportion of personalized recommendations and exploratory recommendations in the recommendation results. When generating the recommendation list, randomly select some occupations that are somewhat related to the user's current interests but are also innovative with probability α and add them to the recommendation list. Let the current personalized recommendation list be L p , and the occupations in this list are generated based on the user's existing interest model and the matching degree with market demands, and can accurately meet the user's current clear interest preferences. The exploratory recommendation list is L e , and the occupation selection comes from the excavation of various emerging occupations in the market and occupations with good potential development prospects but not yet concerned by the user. The final recommendation list L = (1 - α)L p + αL e . For example, when α = 0.2, it means that 20% of the occupations in the final recommendation list come from the exploratory recommendation list L e . By adjusting the value of α, the system can dynamically balance the proportion of personalized recommendations and exploring new occupations according to the actual situation. When the user has a relatively clear understanding of their career interests and the market demand is relatively stable, the value of α can be appropriately reduced to focus on providing highly personalized recommendations; while when the market is in a period of rapid change or the user shows a potential willingness to explore new fields, increase the value of α to recommend more new occupations with development potential to the user, broaden the user's career vision, and help the user discover more potential career development opportunities.

[0065] S6: User feedback module

[0066] Collect the feedback information of the user on the recommendation results, optimize the user interest model and the recommendation algorithm, form a closed-loop optimization mechanism, and improve the recommendation accuracy and user satisfaction.

[0067] 1. Feedback information collection

[0068] Design a feedback entry at the front end of the system, and record the click behavior b click,i (click is 1, not click is 0), favorite behavior b favorite,i , resume submission behavior b apply,i of the user for the recommended occupation i and the text evaluation.

[0069] 2. Feedback information analysis and model optimization

[0070] If the user frequently clicks on a certain occupation but does not submit a resume, calculate the matching degree between this occupation and the user interest model as the user interest feature vector, This is the occupational feature vector. If M is lower than expected, analyze the reasons, adjust the parameters of the user interest model, optimize the recommendation algorithm, and improve the recommendation accuracy.

[0071] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A high-quality career recommendation system based on big data analysis, characterized in that: It includes the following modules: S1: Data Acquisition Module: Responsible for collecting data related to users' career interests and market job demands from multiple sources, providing a comprehensive and real-time data foundation for subsequent analysis and modeling, and ensuring that the system can promptly grasp the latest trends; S2: Real-time Data Analysis Module: This module conducts real-time cleaning and analysis of the collected data, mines the changing trends of users' interest preferences and market job demands, and provides accurate and timely data for subsequent modeling; S3: User Interest Modeling Module: Based on the results of real-time data analysis, uses deep learning algorithms to dynamically model users' interests, updates and adjusts the model with the user's behavior data, and provides a basis for personalized recommendations; S4: Market Demand Modeling Module: Applies natural language processing technology to conduct semantic analysis on job descriptions, constructs a market demand model, monitors demand changes through cluster analysis, and provides market demand references for career recommendations; S5: Dynamic Recommendation Engine Module: Combines the results of user interest and market demand modeling, uses personalized recommendation algorithms to generate a recommendation list, adjusts the recommendation results in real time, and introduces an exploration mechanism to avoid over-personalization; S6: User Feedback Module: Collects users' feedback information on the recommendation results, optimizes the user interest model and recommendation algorithms, forms a closed-loop optimization mechanism, and improves the recommendation accuracy and user satisfaction.

2. The high-quality career recommendation system based on big data analysis according to claim 1, wherein: The multi-source data acquisition module not only collects the traditional behavior data of users on the job hunting platform, but also obtains relevant user behavior data through cooperation with social media platforms and online learning platforms, and uses web crawler technology to frequently collect job information on recruitment websites, with a collection cycle of no more than 15 minutes.

3. A high-quality career recommendation system based on big data analysis according to claim 1, characterized in that: The real-time data analysis module uses real-time stream processing technology to clean the collected data, monitors the data quality in real time by establishing a data quality monitoring index system, and starts the repair process when the data error rate exceeds the set threshold; Uses sliding window technology to analyze user behavior data to calculate the interest index, and adopts an autoregressive moving average model to predict the changes in market job demands.

4. A high-quality career recommendation system based on big data analysis according to claim 1, characterized in that: The user interest modeling module is based on the results of real-time data analysis, uses a long short-term memory network to build a user interest model, processes the user's real-time behavior data sequence through the structures of forget gate, input gate, output gate, candidate memory unit and memory unit to update the model; introduces sentiment analysis technology to conduct sentiment analysis on the user's social media career-related content, and integrates the sentiment tendency score into the interest model.

5. A high-quality career recommendation system based on big data analysis according to claim 1, characterized in that: The market demand modeling module applies natural language processing technology to conduct semantic analysis on job descriptions, extracts key information through word segmentation, word embedding, and text classification to construct a market demand model, uses the K-means clustering algorithm to conduct cluster analysis on the job demand feature vectors, and judges and updates the model of market demand changes by monitoring the changes in the cluster center vector and the number of samples within the cluster.

6. A high-quality career recommendation system based on big data analysis according to claim 1, characterized in that: The dynamic recommendation engine module combines the results of user interest modeling and market demand modeling, implements a recommendation algorithm that combines matrix factorization and deep learning, calculates the matching degree score to generate a recommendation list by minimizing the loss function to solve the user feature matrix and the career feature matrix, and introduces an exploration factor to control the proportion of personalized recommendations and exploratory recommendations.

7. A high-quality career recommendation system based on big data analysis according to claim 1, characterized in that: The user feedback module collects in real time the behavioral data of users' clicks, collections, and resume submissions on the recommended results, as well as text evaluation feedback. By calculating the matching degree between the recommended occupations and the user interest model, if the matching degree is lower than expected, the reasons are analyzed to adjust the parameters of the user interest model and the recommendation algorithm.

8. The high-quality career recommendation system based on big data analysis according to claim 2, characterized in that: The multi-source data acquisition module obtains data from the social media platform through the OAuth authorization mechanism. The acquired data includes the set of user career dynamic information and the set of industry topic discussion content. It cooperates with the online learning platform to obtain the set of user learning course names, learning durations, and course completion progress.

9. The high-quality career recommendation system based on big data analysis according to claim 3, wherein: In the real-time data analysis module, the autoregressive moving average model is where D(t) is the sequence of job demand quantities at time t, p and q are the orders, is the autoregressive coefficient, θ j is the moving average coefficient, ò t is the white noise sequence.

10. A high-quality career recommendation system based on big data analysis according to claim 6, characterized in that: The loss function in the dynamic recommendation engine module is where Ω is the set of user-occupation pairs with known ratings, r ui is the actual rating of user u for occupation i, is the predicted rating, λ is the regularization parameter, U is the user feature matrix, and V is the occupation feature matrix.

Images