User intelligent matching method and system based on data collision algorithm

Abstract

The invention relates to the technical field of artificial intelligence big data processing, in particular to a user intelligent matching method and system based on a data collision algorithm. The establishment of the user tag is the basis of intelligent matching, the basic features of the feature information are correspondingly associated with the user tag, the behavior features are specific behaviors of the user, the achievement features are results generated by the specific behaviors, and the behavior features and the achievement features are associated with the user tag through the basic features. By using a data collision algorithm, matching points can be quickly found in large-scale and multi-dimensional data. Compared with a traditional matching algorithm, the data collision algorithm can process a complex data set more efficiently, and the matching precision and efficiency are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence big data processing technology, and in particular to a user intelligent matching method based on a data collision algorithm and a system thereof. Background Art

[0002] Collision Detection Algorithms: These algorithms are mainly used to detect collisions in the game, such as collisions between users and the environment or other objects. In smart matching, these algorithms can be used to detect potential matching relationships between users.

[0003] Data mining and machine learning: used to analyze and understand user behavior patterns and preferences. Through data mining technology, match developers can extract valuable information from users' match history, interactions, and preferences to make matches more accurately.

[0004] Real-time data processing and analysis: In order to support real-time matching and feedback, competitions usually need to process a large amount of real-time data. Background technology may involve real-time data stream processing systems and big data analysis platforms.

[0005] Algorithm optimization and performance tuning: To ensure the efficiency and accuracy of matching, algorithm optimization and performance tuning are required. This includes selecting appropriate data structures, algorithms, and computing resource configurations, as well as optimization for large-scale concurrent processing.

[0006] Personalization and dynamic adjustment: Based on user feedback and changing competition environments, the intelligent matching system may need to dynamically adjust parameters and strategies. This may involve dynamic adjustment algorithms based on real-time data or personalized matching models based on machine learning.

[0007] Natural Language Processing (NLP) technology is used to analyze and understand user text data, such as chat records and comments. NLP can extract user emotions, interests, and behavioral characteristics, providing more dimensional data support for the matching system.

[0008] Data Privacy and Security technology ensures the security and privacy of user data. The user matching system needs to process a large amount of personal data, so it must comply with privacy protection laws and regulations and use technical means to protect data security. Summary of the invention

[0009] The creative purposes of the user intelligent matching method based on data collision algorithm mainly include the following aspects: 1. Improve matching accuracy and efficiency Creative purpose: By using the data collision algorithm, it is possible to quickly find matching points in large-scale, multi-dimensional data. Compared with traditional matching algorithms, the data collision algorithm can process complex data sets more efficiently and significantly improve the accuracy and efficiency of matching.

[0010] 2. Realize multi-dimensional comprehensive matching Creative purpose: Comprehensively consider and encode the multi-dimensional characteristics of users (such as geographic location, behavior patterns, preferences, etc.) to generate a comprehensive score. This multi-dimensional matching method can provide more comprehensive and accurate matching results, overcoming the limitations of traditional single-dimensional matching algorithms.

[0011] 3. Dynamic adjustment and real-time response Creative purpose: Introduce a dynamic adjustment mechanism to adjust the parameters and weights of the matching algorithm according to the changes in real-time data. This mechanism enables the matching algorithm to respond to user needs and environmental changes in real time, providing more flexible and adaptive matching services.

[0012] 4. Coping with high concurrency requirements Creative purpose: Design an efficient matching algorithm that can maintain fast response capabilities in a high-concurrency environment. Through the efficient calculation and fast comparison of the data collision algorithm, the performance bottleneck problem of the traditional matching algorithm in high-concurrency situations is solved.

[0013] 5. Data Privacy and Security Protection Creative purpose: In the matching process, encryption and anonymization technology are used to protect user data privacy. Ensure that user data is not leaked or abused during transmission and processing, and improve user trust in the matching system.

[0014] 6. Scalability and portability Creative purpose: To design a matching algorithm with good scalability and portability, which can be adjusted and optimized according to different application scenarios and needs. Whether in agriculture, big data, government affairs or other fields, this matching method can be flexibly applied to provide personalized and customized matching services.

[0015] 7. Support diverse application scenarios Creative purpose: Through the application of data collision algorithms, matching problems in different fields can be solved and a wide range of application scenarios can be covered. From agricultural product supply and demand matching to data recommendation in big data, from government service matching to risk management and prediction, data collision algorithms can provide effective solutions. The creative purpose of the user intelligent matching method based on the data collision algorithm is to improve the accuracy and efficiency of matching through efficient, multi-dimensional, dynamically adjusted and secure matching technology, meet high concurrency requirements, and support a variety of application scenarios. Through these innovative purposes, this method can provide smarter and more flexible matching services in various fields, and promote the intelligent and information development of the industry.

[0016] The following technical solutions are used to solve the above technical problems: A user intelligent matching method based on data collision algorithm comprises the following steps: Step 1: Data collection and preprocessing Collect user data and annotated data from various data sources, perform data cleaning and standardization, and encrypt sensitive information; Step 2: User tag generation Generate user tags based on user data, including age, gender, geographic location, education level, years of experience, technical level, and historical behavior; Step 3: Feature Extraction Feature extraction is to extract feature information from the labeled data for use in the model. Feature information is divided into basic features, behavioral features, and achievement features. All features are extracted from the labeled data and standardized. Step 4: Model training Use the random forest model for training. The random forest model processes a large number of features and samples, divides the data set into training set, validation set, and test set, and uses the random forest for training; Decision tree training: For each tree t, the selection of the optimal split point s*, (D, х j , s) is the splitting point of the labeled data of the sample set; Information gain: Information gain IG is defined as: D is the sample set of the current node, D1 and D2 are the two subsets after splitting, and H(D) is the entropy of the node; Entropy H is defined as: Among them, C is the number of categories, the total number of categories in the feature information, and pi is the probability that category i accounts for the total number of categories; c. Prediction of random forest: Let T be the number of decision trees, is the prediction result of the tth tree, then the final prediction of the random forest is: Step 5: Data Collision Data collision is to match the newly input user needs with the existing user data to find the user who best meets the needs; The data collision algorithm is used: Euclidean distance, where xi and yi are two feature vectors respectively; Step 6: Model optimization algorithm: mean square error MSE, where N is the number of samples, is the true value, is the predicted value; Root mean square error RMSE formula: The data is used to extract features and retrain the model for optimization, and the hyperparameters are adjusted through the validation set so that the MSE of the model is continuously reduced and the model is continuously updated with new data.

[0017] In step 2, the user tags include personnel attribute tags, behavior tags, production information tags, land environment information tags, social and economic tags, special identity tags, policy subsidy tags, and time tags; the feature information is divided into basic features, behavior features, and achievement features; the basic features of the feature information are associated with the user tags, the behavior features are the specific behaviors of the users, and the achievement features are the results of the specific behaviors. The behavior features and achievement features are associated with the user tags through the basic features. User tags can be customized to increase or decrease according to different industries and knowledge bases, and the feature information is extracted based on data collection.

[0018] Beneficial effects of the present invention 1. Improve matching accuracy and fairness: Multi-dimensional feature encoding enables the matching algorithm to comprehensively consider all aspects of the user's characteristics, ensuring that the matching results are more reasonable and fair.

[0019] 2. Improve matching speed: The efficient computing power of the data collision algorithm enables the matching process to be completed quickly, reducing user waiting time and improving the game experience.

[0020] 3. Enhance matching flexibility and adaptability: Dynamic adjustment mechanism and real-time data collection and analysis enable the algorithm to be optimized and adjusted according to actual conditions to always maintain the best matching effect.

[0021] 4. Ensure data security and privacy: Security and privacy protection technologies ensure the security of user data during the matching process and enhance user trust. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the matching process of the present invention; Figure 2 A schematic diagram of the connection between the system memory and the processor; Figure 3 A schematic diagram showing the connections for the system's electronic equipment components. DETAILED DESCRIPTION

[0023] refer to Figure 1 , a user intelligent matching method based on a data collision algorithm is implemented by the following steps.

[0024] 1. Establish user tags. The establishment of user tags is the basis of intelligent matching. Each user's tags include but are not limited to the following aspects: personnel attribute tags, behavior tags, production information tags, land environment information tags, social and economic tags, special identity tags, policy subsidy tags, time tags, etc.

[0025] 2. User preparation: The user preparation stage includes the user going online, joining the match and entering the waiting matching thread pool. The specific steps are as follows: 2.1 User enters the matching system.

[0026] 2.2 Users select the matching mode they want to participate in (such as agricultural technology, government services, etc.).

[0027] 2.3 The user enters the matching thread pool and waits for matching.

[0028] 3. Enter the matching requirements for matching. The user or system can enter specific matching requirements. For example: enter the opponent matching information through voice or input interface: "Match a male high-level and qualified senior agricultural user as a teammate to guide the solution of the problem of viral diseases in the cultivation and production of Sunshine Rose." Matching rules: matching by points, matching by role, matching by gender, matching by professional level, matching by daily behavior, etc.

[0029] 4. Intelligent matching large model engine The intelligent matching large model engine performs intelligent screening based on the input matching requirements. The engine uses deep learning and data collision algorithms, combined with the following steps to achieve accurate matching: 4.1.1 Data collection and preprocessing Data collection: Collect basic information, technical level, historical behavior, environmental information, etc. of agricultural users. Collect data related to the cultivation of Sunshine Rose, including geographical location, soil type, climate conditions, etc. Collect technical information related to the prevention and control of viral diseases.

[0030] Data preprocessing: Data cleaning: remove noisy data and incomplete data.

[0031] Data standardization: Process different types of data in a unified format.

[0032] Data encryption: Encrypt users' sensitive information to ensure data privacy.

[0033] 4.1.2 User Tag Generation Tag generation: Based on the collected data, detailed labels are generated for each user, including basic information labels, technology application labels, historical behavior labels, etc.

[0034] Specially mark out user tags that meet the target conditions, such as male, high level, high quality, and growing Sunshine Roses.

[0035] 4.1.3 User screening and matching Filter rule settings: Determine the screening criteria: gender is male, technical level is senior or above, quality evaluation is high, and has experience in growing Sunshine Roses.

[0036] Initial screening: Qualified users are screened out from the user database. During the screening process, multi-party secure computing (MPC) technology is used to protect user privacy, ensure that data is not leaked during the calculation process, and that all parties involved can only see their own data.

[0037] 4.1.4 Matching and Recommendation Matching algorithm: Use a matching engine based on a data collision algorithm to perform refined matching on the initially screened users, taking into account multi-dimensional factors such as the user's technical level, historical behavior, and geographic location.

[0038] Recommendation generation: Generate matching results and recommend the best matching advanced farming users as teammates.

[0039] 4.1.5 Real-time verification and feedback Verification mechanism: Real-time verification of the availability and willingness of the recommended user, and confirmation of the user's willingness to participate in the guidance through instant messaging tools. If the user refuses, the next eligible user will be automatically re-matched.

[0040] Feedback collection: Collect feedback and comments from recommended users to optimize matching algorithms and recommendation strategies.

[0041] 4.2. Specific implementation steps Step 1: Data collection and preprocessing Collect user data and environmental information from various data sources (such as agricultural technology promotion platforms, agricultural cooperative databases, etc.), clean and standardize data, and encrypt sensitive information.

[0042] Step 2: User tag generation Generate detailed labels based on user data, including age, gender, geographic location, education level, years of experience, technical level, historical behavior, etc.

[0043] Step 3: Feature Extraction Feature extraction is to extract useful information from labeled data for use by the model. Features can be divided into basic features, behavioral features, and achievement features.

[0044] Feature extraction method: extract all features from the labeled data and perform standardization.

[0045] Step 4: Model training Use the random forest model for training. Random forest can effectively handle a large number of features and samples and has strong generalization ability.

[0046] Model training method: Divide the data set into training set, validation set and test set, and use random forest for training.

[0047] a. Decision tree training: For each tree t, the selection of the optimal split point s* b. Information Gain: Information gain IG is defined as: where D is the sample set of the current node, D1 and D2 are the two subsets after the split, and H(D) is the entropy of the node.

[0048] Entropy H is defined as: Where C is the number of categories and pi is the probability of category i.

[0049] c. Prediction of random forest: Let T be the number of decision trees, is the prediction result of the tth tree, then the final prediction of the random forest is: Step 5: Data Collision Data collision is to match newly input user needs with existing user data to find the users who best meet the needs.

[0050] Data collision algorithm: Euclidean distance, where x and y are two feature vectors.

[0051] Step 6: Model optimization algorithm: Mean Squared Error (MSE), where N is the number of samples, is the true value, is the predicted value.

[0052] Root Mean Square Error (RMSE): The data after the competition / contest / project is further used for feature extraction and retraining of the model. The optimization process is a cycle, and the hyperparameters are adjusted through the validation set to minimize the MSE of the model, and the model is continuously updated with new data.

[0053] 5. Play games or form teams to tackle problems according to the matching rules After the match is successful, the system will pull the matched users into the team room. The specific competition rules vary depending on the type of competition, such as: Rules of the quiz contest: Users take turns answering questions, points are awarded for correct answers, and points are deducted for incorrect answers.

[0054] Rules for tackling the topic: Provide a tackling plan based on the topic, and users who publish the topic will score and evaluate it.

[0055] 6. Data processing after the game After the game, the intelligent matching model engine processes the data generated during the game, including: Data collection: Collect all data generated during the game / contest / project research process, including user operation records, scores, behavioral habits, etc.

[0056] Feature extraction: Extract features from the collected data and update user tag information.

[0057] Model training: Use new data to perform data collision and model training to optimize the intelligent matching algorithm.

[0058] Model optimization: Based on the newly trained data model, further optimize the matching accuracy and improve the matching effect.

[0059] Specific implementation cases: In this case, an online agricultural knowledge quiz competition is designed. The system needs to intelligently match N users to improve the competition experience. Assuming N=5, we will show the whole process from data labeling, feature extraction, model training, data collision to model optimization.

[0060] Assume that there are the following five users' initial data: User 1: Male, physics teacher, 25 years old, active for 10 hours per week, win rate 0.6, aggressive, chat frequency 5, number of friends 3, points 1200, level 12 User 2: Female, history teacher, 30 years old, active for 15 hours per week, win rate 0.75, conservative, chat frequency 10, number of friends 5, points 1500, level 15 User 3: Male, physics teacher, 22 years old, active for 8 hours per week, win rate 0.5, aggressive, chat frequency 2, number of friends 1, points 900, level 9 User 4: Female, chemistry teacher, 28 years old, active for 12 hours per week, win rate 0.65, conservative, chat frequency 6, number of friends 4, points 1300, level 13 User 5: Male, history teacher, 26 years old, active for 9 hours per week, win rate 0.7, aggressive, chat frequency 4, number of friends 2, points 1100, level 11 Smart matching process: 1. Data annotation: User data is categorized and labeled, and discrete features such as gender, role, and playing habits are processed through One-Hot encoding. For example, genders 'M' and 'F' are encoded as 0 and 1 respectively, and roles 'Physics Teacher', 'History Teacher', and 'Chemistry Teacher' are encoded as different binary values.

[0061] 2. Feature extraction: Extract and standardize various user features (such as game duration, winning rate, chat frequency, etc.) to form feature vectors to facilitate subsequent model training and matching calculations.

[0062] User 1: [1, 0, 1, 0, 0, 0, 25, 10, 0.6, 5, 3, 1200, 12] User 2: [0, 1, 0, 1, 0, 0, 30, 15, 0.75, 10, 5, 1500, 15] User 3: [1, 0, 1, 0, 0, 0, 22, 8, 0.5, 2, 1, 900, 9] User 4: [0, 1, 0, 0, 1, 0, 28, 12, 0.65, 6, 4, 1300, 13] User 5: [1, 0, 0, 1, 0, 1, 26, 9, 0.7, 4, 2, 1100, 11] 3. Model training: Use the random forest model for training. Divide the user data set into training set, validation set, and test set. Use the training set to train the model, adjust the model parameters through the validation set, and finally evaluate the model performance on the test set.

[0063] 4. Data collision: Input matching requirements (such as "matching a male, high-level, radical history teacher"), calculate the feature vectors of all users, and use methods such as Euclidean distance to find the user who best meets the requirements.

[0064] Matching requirement vector: [1, 0, 0, 1, 0, 1, age, game duration, win rate, chat frequency, number of friends, points, level] Assume that other continuous features are user averages: Age = 26.2, Game Duration = 10.8, Win Rate = 0.64, Chat Frequency = 5.4, Number of Friends = 3, Points = 1200, Level = 12.

[0065] The matching demand vector is: [1, 0, 0, 1, 0, 1, 26.2, 10.8, 0.64, 5.4, 3, 1200, 12] The matching degree calculation formula is: 5. Matching results: According to the matching degree, find the user who best meets the requirements for the match. For example, the matching requirements are male, high-level, radical history teachers, and the results are calculated based on the Euclidean distance: User 1: Match score = 1.5 User 2: Match score = 2.8 User 3: Match score = 1.9 User 4: Match score = 3.1 User 5: Match score = 0.8 Then the system will match user 5.

[0066] User 5 is matched with the user demand side to participate in a quiz competition. After the competition, the competition data is recorded, such as the competition duration, winning rate, score change, etc.

[0067] 6. Going further, use the new data generated by the competition to optimize the model: 6.1 Model Optimization Use new data generated by the competition to optimize the model: 6.2 Feature Extraction: Extract features from the new data and update the user's feature vector. For example, the winning rate, points, level, and other data of user 5 may change.

[0068] 6.3 Data Fusion: The new data is merged with the original data to form a new training set, and then re-standardized.

[0069] 6.4 Model training: Retrain the random forest model and adjust the hyperparameters to optimize model performance.

[0070] 6.5 Model Evaluation: Evaluate the model using the validation set and test set, and calculate the mean square error (MSE) and root mean square error (RMSE): The present application also provides an electronic device, see Figure 2 and 3 As shown, an application system provided by an embodiment of the present application is specifically as follows: The electronic device includes: a memory 100 for storing a computer program; a processor 200, which can implement the steps of the business processing method disclosed in any of the aforementioned embodiments when executing the computer program. The memory 100 includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer-readable instructions, and the internal memory provides an environment for the operation of the operating system and computer-readable instructions in the non-volatile storage medium. In some embodiments, the processor 200 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor or other data processing chip, which provides computing and control capabilities for the electronic device, and when executing the computer program stored in the memory 100, the steps of the business processing method disclosed in any of the aforementioned embodiments can be implemented.

[0071] Based on the above embodiments, as a preferred embodiment, see Figure 3 As shown, the electronic device also includes: The input interface 300 is connected to the processor 200 and is used to obtain computer programs, parameters and instructions imported from the outside, and is saved in the memory 100 under the control of the processor 200. The input interface 300 can be connected to an input device to receive parameters or instructions manually input by a user. The input device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the terminal shell, or a keyboard, touchpad or mouse. The display unit 400 is connected to the processor 200 and is used to display the data processed by the processor 200 and to display a visual user interface. The display unit 400 can be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The network port 500 is connected to the processor 200 and is used to communicate with various external terminal devices. The communication technology used for the communication connection can be a wired communication technology or a wireless communication technology, such as mobile high-definition link technology (MHL), universal serial bus (USB), high-definition multimedia interface (HDMI), wireless fidelity technology (WiFi), Bluetooth communication technology, low-power Bluetooth communication technology, communication technology based on IEEE802.11s, etc.

[0072] Figure 2 and 3 Only an electronic device having components 100-500 is shown, and those skilled in the art will appreciate that Figure 2 and 3 The structure shown does not constitute a limitation on the electronic device, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0073] In summary, the user intelligent matching method, device and equipment based on the data collision algorithm of this application, combined with the user interface input box, effectively solves the pain points of complexity, professional threshold, subjective error and lack of flexibility in traditional matching modeling through intelligent and accurate methods, and can achieve the goal of intelligent interactive, high-precision, real-time dynamic and multi-dimensional matching. This method not only improves the accuracy of matching and the game experience, but also ensures the long-term effectiveness and security of the system through continuous model optimization and data protection. .

[0074] Technical points of the present invention: 1. Multi-dimensional feature encoding Technical highlights: Encode multiple user characteristics (such as skill level, geographic location, network latency, game preferences, etc.) to generate a comprehensive score.

[0075] Advantages: Compared with the traditional single-dimensional matching algorithm, multi-dimensional feature coding can more comprehensively reflect the actual situation of users and improve the accuracy and fairness of matching. Through this technical means, users with similar skill levels, close geographical locations, and low network latency can be better matched to improve the competition experience.

[0076] 2. Data collision algorithm Technical highlights: Through data collision technology, users' comprehensive scores can be quickly compared to achieve efficient matching.

[0077] Advantages: The computational complexity of the data collision algorithm is low, and it can maintain a fast matching speed in a high-concurrency environment. Compared with traditional matching algorithms, such as Elo rating or TrueSkill, the data collision algorithm is more efficient in processing multi-dimensional data, allowing users to quickly enter the game and reduce waiting time.

[0078] 3. Dynamic adjustment mechanism Technical highlights: Dynamically adjust the weights and parameters of the matching algorithm based on real-time data to adapt to the ever-changing competition environment and user behavior.

[0079] Advantages: Compared with static algorithms, dynamic adjustment mechanisms can be optimized according to the actual situation of the game to ensure the rationality and timeliness of matching. For example, it can increase the matching speed during peak periods and improve the matching quality during off-peak periods. This technical means enables the algorithm to adjust adaptively, improving the overall matching effect and user satisfaction.

[0080] 4. Real-time data collection and analysis Technical highlights: Continuously collect and analyze various user data, such as game performance, network status, device information, etc., to provide accurate input data for the algorithm.

[0081] Advantages: Compared with algorithms that rely on historical data, real-time data collection and analysis can ensure that the algorithm matches based on the latest user information, improving the accuracy and timeliness of matching. Through this technical means, the algorithm can quickly respond to changes in user status and provide better matching results.

[0082] 5. Security and Privacy Protection Technical points: During the data collection and processing process, encryption, anonymization and other technical means are used to protect users' personal privacy and data security.

[0083] Advantages: Compared with matching algorithms that do not focus on privacy protection, this invention can effectively prevent user data leakage through security and privacy protection measures, thereby improving user trust and satisfaction with the platform. This technical means ensures that user data is fully protected during the matching process, in line with modern data privacy regulations.

[0084] This patent solution can achieve Multi-dimensional data analysis: Accurate user portraits can be achieved through analysis of multi-dimensional user data (such as basic information, technical applications, historical behavior, etc.).

[0085] Refined matching: Use data collision algorithms to perform refined matching of users to ensure that the recommended users meet specific conditions and requirements.

[0086] Label management: Personalized recommendations can be achieved by managing multiple user labels (such as age, gender, geographic location, technical level, etc.).

[0087] Privacy computing: Privacy computing technologies (such as multi-party secure computing and homomorphic encryption) are used in the data processing and matching process to ensure user data privacy.

[0088] Cross-domain application: This method is not only suitable for agricultural technology promotion, but can also be applied to multiple fields such as big data, government services, and data analysis.

[0089] Automated matching: Achieve efficient user matching and recommendations through automated algorithms, reducing manual intervention.

[0090] Possible future application scenarios Agricultural technology promotion: Accurately recommend agricultural technologies and solutions to improve farmers' production efficiency and income. Recommend appropriate agricultural machinery and management technology based on farmers' specific conditions (such as crop type, planting area, soil type, etc.). Provide personalized pest and disease control solutions to reduce pesticide use and improve crop quality.

[0091] Healthcare: Recommend personalized medical plans and drugs based on the patient's medical history, genetic data, and lifestyle habits. Intelligently match doctors and patients to ensure that patients can get the best medical services. Provide health management and prevention advice to help users stay healthy.

[0092] Education and vocational training: Recommend appropriate courses and training programs based on students’ interests, academic performance, and career plans. Intelligently match teachers and students to provide personalized teaching and guidance. Help companies conduct training and career development planning based on employees’ skills and needs.

[0093] Smart Manufacturing and Industry 4.0: Recommend optimization solutions based on the operating status of production lines and equipment to improve production efficiency. Intelligently match workers and positions to ensure the best configuration of the production process. Provide equipment maintenance and fault prediction to reduce downtime.

[0094] Smart cities and public services: Provide personalized public services and policy recommendations based on citizens’ needs and behavior data. Intelligently match citizens and government services to improve the efficiency and satisfaction of public services. Provide urban management and decision-making support to improve urban operation efficiency.

[0095] Human Resources and Recruitment: Recommend suitable jobs based on job seekers’ skills and career experience. Intelligently match companies and job seekers to improve recruitment efficiency and matching. Provide employee training and career development advice to help companies improve employee skills.

[0096] Finance and insurance: Recommend personalized investment and financial products based on the user's financial status and risk preferences. Intelligently match users and insurance products to provide personalized insurance plans. Provide credit assessment and risk management to improve the security and reliability of financial services.

[0097] Glossary 1. Data collision algorithm Data Collision Algorithm is an algorithm used to match and compare large data sets. It finds the best match by analyzing the feature vectors of the data set and calculating the similarity or difference. In the user matching system, this algorithm is used to find the best match for the user's needs.

[0098] 2. User Tags User tags are a set of attributes that describe user characteristics. These tags can include personnel attribute tags, behavior tags, production information tags, land environment information tags, socio-economic tags, special identity tags, policy subsidy tags, time tags, etc. User tags help the system identify and classify users during the matching process.

[0099] 3. Eigenvector A feature vector is a numerical vector used to represent user attributes. Each user's feature vector contains several numerical elements, each of which represents the value of a user attribute (such as match duration, win rate, chat frequency, etc.). Feature vectors are the basis for matching calculations.

[0100] 4. Euclidean distance Euclidean distance is a metric used to calculate the distance between two vectors.

[0101] 5. One-Hot Encoding One-Hot Encoding is a method of converting categorical data into binary vectors. In the user matching system, discrete features such as gender, role, and playing habits are converted into numerical vectors through One-Hot Encoding to facilitate subsequent calculations.

[0102] 6. Standardization Standardization is a data preprocessing technique used to convert feature data into a standard normal distribution with a mean of 0 and a standard deviation of 1. Standardized data can avoid the impact of differences in the magnitude of different feature values ​​on matching calculations.

[0103] 7. Training set, validation set and test set The training set is a dataset used to train the model, which contains a large number of labeled samples. The validation set is used to adjust the model parameters and evaluate the model performance during the training process. The test set is a dataset used to finally evaluate the generalization ability of the model, which is usually not involved in model training.

[0104] 8. Random Forest Model The Random Forest Model is an ensemble learning method that improves prediction accuracy and stability by building multiple decision trees and taking their average. In the user matching system, the Random Forest Model is used to make predictions and matches based on user characteristics.

[0105] 9. Mean Squared Error (MSE) Mean Squared Error (MSE) is a metric for evaluating model prediction error.

[0106] 10. Model Optimization Model optimization refers to the process of improving model performance by adjusting model parameters and structure. In the user matching system, model optimization gradually improves matching accuracy and match experience by continuously introducing new data and retraining the model.

Claims

1. A user intelligent matching method based on data collision algorithm, characterized in that The steps include: Step 1: Data collection and preprocessing Collect user data and annotated data from various data sources, perform data cleaning and standardization, and encrypt sensitive information; Step 2: User tag generation Generate user tags based on user data, including age, gender, geographic location, education level, years of experience, technical level, and historical behavior; Step 3: Feature Extraction Feature extraction is to extract feature information from the labeled data for use by the model. Feature information is divided into basic features, behavioral features, and achievement features. Extract all features from the labeled data and perform standardization; Step 4: Model training Use the random forest model for training. The random forest model processes a large number of features and samples, divides the data set into training set, validation set, and test set, and uses the random forest for training; Decision tree training: For each tree t, the selection of the optimal split point s*, (D, х j , s) is the splitting point of the labeled data of the sample set; Information gain: Information gain IG is defined as: D is the sample set of the current node, D1 and D2 are the two subsets after the split, and H(D) is the entropy of the node; Entropy H is defined as: Among them, C is the number of categories, the total number of categories in the feature information, and pi is the probability that category i accounts for the total number of categories; c. Prediction of random forest: Let T be the number of decision trees, is the prediction result of the tth tree, then the final prediction of the random forest is: Step 5: Data Collision Data collision is to match the newly input user needs with the existing user data to find the user who best meets the needs; The data collision algorithm is used: Euclidean distance, where xi and yi are two feature vectors respectively; Step 6: Model optimization algorithm: mean square error MSE, where N is the number of samples, is the true value, is the predicted value; Root mean square error RMSE formula: The data is used to extract features and retrain the model for optimization, and the hyperparameters are adjusted through the validation set so that the MSE of the model is continuously reduced and the model is continuously updated with new data.

2. According to claim 1, a user intelligent matching method based on data collision algorithm is characterized in that The user tags in step 2 include personnel attribute tags, behavior tags, production information tags, land environment information tags, social and economic tags, special identity tags, policy subsidy tags, and time tags; the feature information is divided into basic features, behavior features, and achievement features; The basic features of the feature information are associated with the user tags. The behavior features are the specific behaviors of the user, and the achievement features are the results of the specific behaviors. The behavior features and the achievement features are associated with the user tags through the basic features.

Images