An information distribution method and system based on data access and big data classification

Abstract

The present invention relates to the field of data processing technology, and specifically to an information distribution method and system based on data access and big data classification. First, collect user information data, store it in a database, and retrieve and update each data in real time; secondly, identify and classify the user information data to obtain hierarchical classification information; then input the information in the database into a first information similarity analysis model and a second information similarity analysis model, and input the output into a first information distribution model and a second information distribution model respectively to obtain a first matching information set and a second matching information set; combine the two sets into a comprehensive matching information set, calculate the distribution weight value, and sort each matching information in the comprehensive matching information set according to the distribution weight value; finally, select a number of matching information for distribution, collect user feedback, and update the database. The present invention can improve the efficiency and accuracy of information distribution.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an information distribution method and system based on data access and big data classification. Background Art

[0002] The demand for information distribution stems from the continuous development of the information society, and people's demand for information acquisition and processing continues to increase. The information distribution system aims to deliver information to target users effectively and timely to meet their personalized needs and improve information utilization efficiency.

[0003] The original bucketing information distribution method, that is, the distribution strategy based on fixed rules, is centered on presetting a set of fixed rules to distribute information into different "buckets" or user groups. The advantages of this method are its simplicity of implementation and ease of management, but its limitation is that it lacks adaptability to changes in user needs. When user needs become more diverse and the amount of information continues to increase, the fixed rule distribution method will appear too rigid and difficult to meet personalized needs. The relevance of information may decrease, resulting in the content received by users not being in line with their actual interests and needs, thus affecting the dissemination effect of information and user experience.

[0004] With the emergence of big data technology, information distribution methods have begun to shift to data-driven strategies. This approach uses big data technology to conduct in-depth analysis of user behavior, interests, and preferences. By processing large amounts of user data, the distribution system can accurately segment users and achieve personalized information push. Data-driven distribution methods can not only improve the relevance of information, but also adjust push strategies in real time to adapt to changes in user needs. This dynamic adjustment capability makes information distribution more flexible and effective, and can maintain efficient information delivery in a rapidly changing environment, thereby improving user experience and information dissemination.

[0005] However, existing methods for distributing information based on big data classification also have some significant drawbacks. Although big data classification can significantly improve the accuracy of information distribution, it still faces the problem of inaccurate delivery in practical applications. The variability of user behavior and interests makes it possible for information recommendation systems to not always maintain a high level of accuracy. In addition, existing methods often ignore differences in user identities, resulting in information distribution that does not fully consider the user's personal background, role, and specific needs, thus affecting the effectiveness of information.

[0006] To this end, an information distribution method and system based on data access and big data classification are proposed. Summary of the invention

[0007] The purpose of the present invention is to provide an information distribution method and system based on data access and big data classification, which is used to improve the efficiency and accuracy of information distribution. First, collect user information data, including user needs, first user identity information, second user identity information and user historical information records; store the user information data in a database, and retrieve and update each data in real time; secondly, identify and classify the user information data, read the first user identity information and the second user identity information in the database, classify the users hierarchically, and obtain hierarchical classification information; then input the information in the database into the first information similarity analysis model and the second information similarity analysis model to obtain a user similarity matrix and an interest similarity matrix respectively; input the user similarity matrix and the interest similarity matrix into The first information distribution model and the second information distribution model are used to obtain a first matching information set and a second matching information set; the first matching information set and the second matching information set are combined into a comprehensive matching information set, and a user similarity matrix and an interest similarity matrix are used to perform comprehensive weight calculation to obtain a distribution weight value of each matching information in the comprehensive matching information set; each matching information in the comprehensive matching information set is sorted according to the distribution weight value; finally, a number of matching information are selected from the sorted comprehensive matching information set to accurately distribute information to users, and feedback from users after obtaining the information is collected to update each data information in the database.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] An information distribution method based on data access and big data classification, comprising:

[0010] Collecting user information data, including user needs, first user identity information, second user identity information, and user history information records;

[0011] Store user information data in the database and retrieve and update the data in real time;

[0012] Identify and classify the user information data, read the first user identity information and the second user identity information in the database, classify the users hierarchically, and obtain hierarchical classification information;

[0013] Inputting the information in the database into a first information similarity analysis model and a second information similarity analysis model to obtain a user similarity matrix and an interest similarity matrix respectively; inputting the user similarity matrix and the interest similarity matrix into a first information distribution model and a second information distribution model respectively to obtain a first matching information set and a second matching information set;

[0014] The first matching information set and the second matching information set are combined into a comprehensive matching information set, and a user similarity matrix and an interest similarity matrix are used to perform comprehensive weight calculation to obtain a distribution weight value of each matching information in the comprehensive matching information set; and each matching information in the comprehensive matching information set is sorted according to the distribution weight value;

[0015] A number of matching information are selected from the sorted comprehensive matching information set to accurately distribute information to users, and feedback from users after obtaining the information is collected to update the data information in the database.

[0016] The storage of the user information data adopts distributed storage technology, and the specific implementation process is as follows:

[0017] The user information data is divided into multiple data blocks; the data blocks are distributed to multiple storage nodes; copies of the data blocks are saved on the multiple storage nodes to prevent data loss when a single node fails; a consistency protocol is used to ensure that the data blocks saved on the multiple storage nodes remain consistent and that all copies of the data blocks are synchronized; data requests are distributed through a load balancing mechanism to prevent a node from being overloaded; and update operations on the data blocks are synchronized to all related storage nodes.

[0018] Further, the first user identity information and the second user identity information in the database are read to classify the users into different levels; wherein the first user identity information is the role of the user, and the second user identity information is the authority of the user;

[0019] The specific process of the hierarchical classification is as follows:

[0020] Extract the first user identity information and the second user identity information of each user from the database to obtain ; Assume that the user information feature matrix is , the label matrix is , where user information features refer to the user's role features and the user's permission features, and the label matrix contains hierarchical labels; the decision tree classification algorithm is used to classify users, and the formula is ,in, is the predicted level category, is a given feature When the user belongs to a certain level category The probability of Classification to get hierarchical labels , the formula is ,in, is a decision tree classifier;

[0021] Finally, label each user's level Convert to hierarchical category scores and storing the hierarchical category scores in the database.

[0022] Furthermore, the first information similarity analysis model describes the similarity between users; the second information similarity analysis model describes the similarity between demand information; the first information similarity analysis model and the second information similarity analysis model both use cosine similarity to calculate the user similarity matrix and the interest similarity matrix respectively;

[0023] Among them, the user similarity matrix The calculation formula is:

[0024] ;

[0025] in, and The users are and The eigenvector of

[0026] Interest Similarity Matrix The calculation formula is:

[0027] ;

[0028] in, and The users are and The eigenvector of

[0029] The user similarity matrix and the data of the database are input into the first information distribution model to obtain a first matching information set; the interest similarity matrix and the data of the database are input into the second information distribution model to obtain a second matching information set;

[0030] Among them, in the first information distribution model, first, calculate the user On demand Demand score , the demand score It includes actual rating and predicted rating; the actual rating is the user's The actual score given; the formula for the predicted score is:

[0031] ;

[0032] Among them, the prediction score Used similar users' needs ratings To predict; Is with the user Similar user groups, Is a user and users The user similarity matrix between ;

[0033] Second, using the needs score Calculating Demand Forecast Scores ;

[0034] Then, score based on demand forecast Screen the user requirements as the first matching information set ;

[0035] Furthermore, in the second information distribution model, first, the user About content Interest rating , the interest score Calculated based on content features and user preference features:

[0036] ;

[0037] in, Is a user The eigenvector of It is the content The eigenvector of

[0038] Second, using the interest rating Calculating contextual ratings ;

[0039] Then, based on the content-related rating Screen the user requirements as the second matching information set .

[0040] Further, the first matching information set and the second matching information set are combined into a comprehensive matching information set. , and perform comprehensive weight calculation on the user similarity matrix and the interest similarity matrix to obtain the distribution weight value of each matching information in the comprehensive matching information set ,in, It is the first in the comprehensive matching information set. The distribution weight value of each piece of information; is the hierarchical category score corresponding to each user; It is the user weight matrix calculated based on the user similarity matrix; is the interest weight matrix calculated based on the interest similarity matrix;

[0041] Then, based on the distribution weight value of each matching information corresponding to each user The matching information in the comprehensive matching information set is sorted.

[0042] An information distribution system based on data access and big data classification, comprising:

[0043] A user information data collection unit, wherein the user information data includes user needs, first user identity information, second user identity information, and user history information records;

[0044] A user information data storage unit stores the user information data in a database and retrieves and updates the data in real time;

[0045] A user information data identification and classification unit, which reads the first user identity information and the second user identity information in the database and classifies the users into different levels;

[0046] A similarity calculation unit, inputting the information in the database into a first information similarity analysis model and a second information similarity analysis model to obtain a user similarity matrix and an interest similarity matrix respectively;

[0047] A matching information collection unit, inputting the user similarity matrix and the interest similarity matrix into a first information distribution model and a second information distribution model respectively, to obtain a first matching information set and a second matching information set;

[0048] a weight calculation unit, combining the first matching information set and the second matching information set into a comprehensive matching information set, performing comprehensive weight calculation using a user similarity matrix and an interest similarity matrix, and obtaining a distribution weight value of each matching information in the comprehensive matching information set; and sorting each matching information in the comprehensive matching information set according to the distribution weight value;

[0049] The information distribution unit selects a number of matching information from the sorted comprehensive matching information set to accurately distribute information to users, collects feedback from users after obtaining the information, and updates the data information in the database.

[0050] The first user identity information is the user's role, and the second user identity information is the user's authority. The specific process of the hierarchical classification is:

[0051] Assume that the user information feature matrix is , the label matrix is , where user information features refer to the user's role features and the user's permission features, and the label matrix contains hierarchical labels;

[0052] Use decision tree classification algorithm to classify users. ,in, is the predicted level category, is a given feature When the user belongs to a certain level category The probability of Classification to get hierarchical labels ,in, is a decision tree classifier; finally, the level label of each user Convert to hierarchical category scores and storing the hierarchical category scores in the database.

[0053] Furthermore, the user similarity matrix The calculation formula is ,in, and The users are and The eigenvector of

[0054] The interest similarity matrix The calculation formula is ,in, and The users are and The feature vector of .

[0055] Furthermore, the process of obtaining the first matching information set is as follows: first, calculating the user On demand Demand score , the demand score It includes actual rating and predicted rating; the actual rating is the user's The actual score given; the formula for the predicted score is , where the predicted score Used similar users' needs ratings To predict; Is with the user Similar user groups, Is a user and users The user similarity matrix between them; secondly, using the demand scoring Calculating Demand Forecast Scores ; Then, score based on demand forecast Screen the user requirements as the first matching information set ;

[0056] The process of obtaining the second matching information set is as follows: first, calculate the user About content Interest rating , the interest score Calculated based on content features and user preference features: ,in, Is a user The eigenvector of It is the content Secondly, using the interest score Calculating contextual ratings ,in, Is a user and users Then, based on the content-related score Screen the user requirements as the second matching information set .

[0057] Further, the first matching information set and the second matching information Combined into a comprehensive matching information set , and perform comprehensive weight calculation on the user similarity matrix and the interest similarity matrix to obtain the distribution weight value of each matching information in the comprehensive matching information set ,in, It is the first in the comprehensive matching information set. The distribution weight value of each piece of information; is the hierarchical category score corresponding to each user; It is the user weight matrix calculated based on the user similarity matrix; is the interest weight matrix calculated based on the interest similarity matrix; then, based on the distribution weight value of each matching information corresponding to each user The matching information in the comprehensive matching information set is sorted.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1. The present invention uses the first user identity information and the second user identity information in the database to classify users in a hierarchical manner and obtain hierarchical classification information. For high-level users, the system will distribute more important or higher-value information, while for low-level users, the system will distribute more basic or conventional information, thereby more effectively utilizing system resources. This hierarchical classification can more accurately understand the needs and interests of users, thereby providing a more reliable basis for information matching and distribution, and improving the efficiency and accuracy of information distribution.

[0060] 2. The present invention obtains a user similarity matrix and an interest similarity matrix respectively by introducing the first and second information similarity analysis models. Through independent analysis of users and interests, we can deeply understand the needs and interests of users and ensure that information distribution is more in line with the actual needs of users; using the similarity matrix to classify information can improve the efficiency and accuracy of distribution and reduce the probability of users receiving irrelevant information; inputting these two matrices into the corresponding information distribution model can achieve more accurate matching, thereby improving user satisfaction and the relevance of information.

[0061] 3. The present invention uses the user similarity matrix and the interest similarity matrix to perform comprehensive weight calculation, obtains the distribution weight value of each matching information in the comprehensive matching information set, and then sorts the distribution weight value to screen a number of information for distribution. This method combines the user similarity matrix and the interest similarity matrix, and performs weight calculation on the comprehensive matching information set, so as to ensure that the distributed information not only meets the actual needs of the user, but also closely matches his or her interests. This comprehensive weight calculation can effectively improve the pertinence of information and reduce the interference of irrelevant information. In addition, by sorting the information, the most relevant information can be displayed first, further improving the efficiency and accuracy of information distribution. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic diagram of a flow chart of an information distribution method based on data access and big data classification of the present invention;

[0063] Figure 2 A flowchart of obtaining a first matching information set according to the present invention;

[0064] Figure 3 A flowchart of obtaining a second matching information set according to the present invention;

[0065] Figure 4 This is a system structure diagram of an information distribution system based on data access and big data classification according to the present invention. DETAILED DESCRIPTION

[0066] 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.

[0067] In order to improve the accuracy and effectiveness of information distribution, the present invention provides an information distribution method and system based on data access and big data classification. In order to illustrate the function of the present invention, the effectiveness of the present invention will be illustrated from the following embodiments.

[0068] Embodiment 1

[0069] Company A's enterprise system adopts an information distribution method based on data access and big data classification to provide different information to users at different levels.

[0070] Reference Figure 1 In step S01, user information data is collected, including user needs, first user identity information, second user identity information and user history information records.

[0071] Reference Figure 1 Step S02, storing the user information data in a database, wherein each data is retrieved and updated in real time.

[0072] The storage of the user information data adopts distributed storage technology, and the specific implementation process is as follows:

[0073] Divide the user information data into multiple data blocks; distribute the data blocks to multiple storage nodes; save copies of the data blocks on the multiple storage nodes to prevent data loss when a single node fails; use a consistency protocol to ensure that the data blocks stored on the multiple storage nodes remain consistent, and ensure that all copies of the data blocks are synchronized; distribute data requests through a load balancing mechanism to prevent a node from being overloaded; synchronize the update operation of the data block to all related storage nodes. The distributed storage technology used in this embodiment is a prior art and will not be described in detail here.

[0074] Storing user information data using distributed storage technology can improve data reliability and availability, and prevent data loss caused by single point failures through data block copies; in addition, using consistency protocols to ensure data synchronization and consistency, and improving system scalability and performance through distributed architecture, supporting efficient storage and processing of large-scale data.

[0075] Reference Figure 1 In step S03, user information data is identified and classified. Specifically, the first user identity information and the second user identity information in the database are read to classify the users into different levels; wherein the first user identity information is the user's role, and the second user identity information is the user's authority. In this embodiment, the user's role can be a senior manager, a middle manager, and an ordinary employee; the user's authority is that senior managers need strategic summary reports, middle managers need operational data analysis, and ordinary employees need daily tasks and operational guidance.

[0076] Furthermore, the specific process of hierarchical classification is:

[0077] Extract the first user identity information and the second user identity information of each user from the database to obtain ; Assume that the user information feature matrix is , the label matrix is , where user information features refer to the user's role features and the user's permission features, and the label matrix contains hierarchical labels.

[0078] In this embodiment, a decision tree classification algorithm is used to classify users:

[0079] ;

[0080] in, is the predicted level category, is a given feature When the user belongs to a certain level category The probability of

[0081] For each user Classification to get hierarchical labels :

[0082] ;

[0083] in, is a decision tree classifier;

[0084] Finally, label each user's level Convert to hierarchical category scores , and save the hierarchical category scores in the database. In this embodiment, some hierarchical classification results are shown in Table 1.

[0085] Table 1 Classification results of some levels

[0086] User Level Tier Labels Senior Managers 1 Middle Managers 2 Ordinary employees 3 ... ...

[0087] By classifying users according to levels, users can be accurately located. This method can improve the pertinence and effectiveness of information distribution, ensure that users receive content that meets their needs and interests, and make the subsequent information matching and distribution process more efficient and accurate. In addition, this hierarchical classification method can optimize resource allocation and reduce computational complexity when processing large amounts of data.

[0088] Reference Figure 1 In step S04, the information in the database is input into the first information similarity analysis model and the second information similarity analysis model to obtain a user similarity matrix and an interest similarity matrix respectively; the user similarity matrix and the interest similarity matrix are input into the first information distribution model and the second information distribution model respectively to obtain a first matching information set and a second matching information set.

[0089] Among them, the first information similarity analysis model describes the similarity between users; the second information similarity analysis model describes the similarity between demand information; the first information similarity analysis model and the second information similarity analysis model both use cosine similarity to calculate the user similarity matrix and the interest similarity matrix respectively.

[0090] User similarity matrix The calculation formula is:

[0091] ;

[0092] in, and The users are and The feature vector of .

[0093] Interest Similarity Matrix The calculation formula is:

[0094] ;

[0095] in, and The users are and The feature vector of .

[0096] The user similarity matrix and the data of the database are input into the first information distribution model to obtain a first matching information set; the interest similarity matrix The data in the database is input into the second information distribution model to obtain a second matching information set.

[0097] The flowchart for obtaining the first matching information set is as follows: Figure 2 First, calculate the user On demand Demand score , the demand score It includes actual rating and predicted rating; the actual rating is the user's The actual score given; the formula for the predicted score is:

[0098] ;

[0099] Among them, the prediction score Used similar users' needs ratings To predict; Is with the user Similar user groups, Is a user and users The user similarity matrix between ;

[0100] Second, using the needs score Calculating Demand Forecast Scores ;

[0101] Then, score based on demand forecast Screen the user requirements as the first matching information set .

[0102] The flowchart for obtaining the second matching information set is shown in FIG. Figure 3 First, calculate the user About content Interest rating , the interest score Calculated based on content features and user preference features:

[0103] ;

[0104] in, Is a user The eigenvector of It is the content The eigenvector of

[0105] Second, using the interest rating Calculating contextual ratings ;

[0106] Then, based on the content-related rating Screen the user requirements as the second matching information set .

[0107] First matching information set Including historical demand information of other users who are closer to the user level. For ordinary employees, historical demand information includes daily task instructions, operation manuals or solutions to common problems similar to other ordinary employees; for senior managers, historical demand information includes areas that other senior managers have paid attention to in the past, such as similar strategic summary reports, market trends or financial analysis information; for middle-level managers, historical demand information includes operational data analysis reports, business process improvements, efficiency improvements and other information similar to their departments or business lines.

[0108] The second matching information set The collection includes demand information related to the user's interests. For ordinary employees, the demand information includes training and skill improvement resources, such as training courses, skill improvement resources or career development suggestions related to their positions; for senior managers, the demand information includes relevant industry trends and technological developments, such as the latest industry trends, technological innovations or competitor analysis related to corporate strategies; for middle-level managers, the demand information includes best practices and case analyses, such as management skills, best practices for process optimization or success stories in the industry.

[0109] Input user data information into the similarity analysis model to obtain the user similarity matrix and interest similarity matrix respectively, so as to accurately identify the matching degree between user needs and interests. This can effectively classify and filter information, ensure that the distributed content meets the current needs of users, and improve the relevance and accuracy of information distribution.

[0110] Reference Figure 1 In step S05, the first matching information set and the second matching information set are combined into a comprehensive matching information set, and a user similarity matrix and an interest similarity matrix are used to perform comprehensive weight calculation to obtain a distribution weight value of each matching information in the comprehensive matching information set; and each matching information in the comprehensive matching information set is sorted according to the distribution weight value.

[0111] Specifically, the first matching information set and the second matching information set are combined into a comprehensive matching information set. , and calculate the user weight matrix based on the user similarity matrix , the calculation formula is:

[0112] ;

[0113] Furthermore, the interest weight matrix is ​​calculated based on the interest similarity matrix , the calculation formula is:

[0114] ;

[0115] The user similarity matrix and the interest similarity matrix are subjected to comprehensive weight calculation to obtain the distribution weight value of each matching information in the comprehensive matching information set:

[0116] ;

[0117] in, It is the first in the comprehensive matching information set. The distribution weight value of each piece of information; is the hierarchical category score corresponding to each user.

[0118] Then, based on the distribution weight value of each matching information corresponding to each user The matching information in the comprehensive matching information set is sorted.

[0119] The calculation method of the distribution weight value integrates the user weight matrix and the interest weight matrix, that is, it integrates the user information and interest information of the relevant levels, which can improve the accuracy of information matching, make the information obtained by users more relevant and in line with actual needs, and ensure that the information distribution is more targeted and personalized.

[0120] Reference Figure 1 In step S06, a number of matching information are selected from the sorted comprehensive matching information set to accurately distribute information to users, and feedback from users after obtaining the information is collected to update the data information in the database.

[0121] This information distribution method based on data access and big data classification significantly improves the accuracy and hierarchical differentiation of information distribution. First, through the comprehensive collection and real-time update of user information, user needs and behavior patterns can be accurately understood. In the hierarchical classification process, hierarchical division is carried out according to user identity information to ensure that information distribution is targeted. For example, VIP and SVIP users will receive higher-end recommendations, while ordinary users will receive regular information, which enhances the personalization and hierarchical differentiation of services. In addition, by using user similarity and interest similarity matrices to generate two matching information sets and sorting them according to comprehensive weights, the relevance and accuracy of information can be improved. The comprehensive matching information set combines user needs and interests, making the distributed information more in line with the user's actual preferences. Overall, this method improves the accuracy of information distribution while taking into account the personalized needs of users at different levels.

[0122] Embodiment 2

[0123] An e-commerce platform adopts an information distribution system based on data access and big data classification to distribute different shopping information to users at different levels.

[0124] Reference Figure 4 , which is a system structure diagram of an information distribution system based on data access and big data classification, including a user information data acquisition unit, wherein the user information data includes user needs, first user identity information, second user identity information, and user historical information records.

[0125] An information distribution system based on data access and big data classification also includes a user information data storage unit, which stores the user information data in a database and retrieves and updates each data in real time.

[0126] Furthermore, an information distribution system based on data access and big data classification also includes a user information data identification and classification unit, which reads the first user identity information and the second user identity information in the database and classifies users into different levels.

[0127] The first user identity information is the user's role, and the second user identity information is the user's authority. In this embodiment, the user's role is a common user, a VIP customer, and an SVIP customer; the user's authority includes discount information, exclusive event information, and preferential information.

[0128] The specific process of the hierarchical classification is as follows: Assume that the user information feature matrix is , the label matrix is , where user information features refer to the user's role features and the user's permission features, and the label matrix contains hierarchical labels;

[0129] Use decision tree classification algorithm to classify users. ,in, is the predicted level category, is a given feature When the user belongs to a certain level category The probability of Classification to get hierarchical labels ,in, is a decision tree classifier; finally, the level label of each user Convert to hierarchical category scores The hierarchical classification scores are stored in the database. Some hierarchical classification results are shown in Table 2.

[0130] Table 2 Classification results of some levels

[0131] User Level Tier Labels SVIP Customers 1 VIP Customers 2 Normal User 3 ... ...

[0132] Furthermore, an information distribution system based on data access and big data classification also includes a similarity calculation unit, which inputs the information in the database into a first information similarity analysis model and a second information similarity analysis model to obtain a user similarity matrix and an interest similarity matrix respectively.

[0133] Specifically, the user similarity matrix The calculation formula is:

[0134] ;

[0135] in, and The users are and Eigenvector of; Interest similarity matrix The calculation formula is:

[0136] ;

[0137] in, and The users are and The feature vector of .

[0138] Furthermore, an information distribution system based on data access and big data classification also includes a matching information acquisition unit, which inputs the user similarity matrix and the interest similarity matrix into the first information distribution model and the second information distribution model respectively to obtain a first matching information set and a second matching information set.

[0139] Specifically, the process of obtaining the first matching information set is as follows: first, calculate the user On demand Demand score , the demand score It includes actual rating and predicted rating; the actual rating is the user's The actual score given; the formula for the predicted score is

[0140] ;

[0141] Among them, the prediction score Used similar users' needs ratings To predict; Is with the user Similar user groups, Is a user and users Second, using the demand scoring Calculating Demand Forecast Scores ; Then, score based on demand forecast Screen the user requirements as the first matching information set .

[0142] Furthermore, the process of obtaining the second matching information set is as follows: first, calculate the user About content Interest rating , the interest score Calculated based on content features and user preference features: ,in, Is a user The eigenvector of It is the content Secondly, using the interest score Calculating contextual ratings ,in, Is a user and users Then, based on the content-related score Screen the user requirements as the second matching information set .

[0143] First matching information set Including historical demand information of other users who are closer to the user level. For ordinary users, historical demand information includes regular discounts and product recommendations similar to those of other ordinary users, such as popular discounts, best-selling product recommendations or popular promotions; for VIP users, historical demand information includes exclusive discounts, high-end product recommendations and customized service information similar to those of other VIP users, such as special discounts related to high-end products, exclusive new product launches or exclusive recommendations of high-end products; for SVIP users, historical demand information includes more advanced services and information, such as priority customer service information, invitations to exclusive events and unique personalized service information in addition to advanced discounts and product recommendations.

[0144] The second matching information set The collection includes demand information related to the user's interests. For ordinary users, demand information includes regular discounts and product recommendations related to their interests, such as products related to their purchase history, discount information that matches their browsing behavior, or regular commodity and service information; for VIP users, demand information includes more targeted exclusive discounts and high-end product recommendation information. If a VIP user shows interest in a certain category of high-end goods, the system will recommend the latest products or unique promotions in that category, and may include additional value-added service information; for SVIP users, demand information includes highly personalized high-end recommendations and exclusive service information, such as unique customized product recommendations, exclusive limited edition product information, special offers, and personalized service plans. In addition, the system will also distribute information on high-end activities or customer experience improvement opportunities that SVIP users can participate in first.

[0145] Furthermore, an information distribution system based on data access and big data classification also includes a weight calculation unit, which combines the first matching information set and the second matching information set into a comprehensive matching information set, uses a user similarity matrix and an interest similarity matrix to perform a comprehensive weight calculation, and obtains a distribution weight value of each matching information in the comprehensive matching information set; and sorts each matching information in the comprehensive matching information set according to the distribution weight value.

[0146] Specifically, the first matching information set and the second matching information set are combined into a comprehensive matching information set. , and calculate the user weight matrix based on the user similarity matrix , the calculation formula is:

[0147] ;

[0148] Furthermore, the interest weight matrix is ​​calculated based on the interest similarity matrix , the calculation formula is:

[0149] ;

[0150] The user similarity matrix and the interest similarity matrix are subjected to comprehensive weight calculation to obtain the distribution weight value of each matching information in the comprehensive matching information set:

[0151] ;

[0152] in, It is the first in the comprehensive matching information set. The distribution weight value of each piece of information; is the hierarchical category score corresponding to each user.

[0153] Then, based on the distribution weight value of each matching information corresponding to each user The matching information in the comprehensive matching information set is sorted.

[0154] Furthermore, an information distribution system based on data access and big data classification also includes an information distribution unit, which selects several matching information from the sorted comprehensive matching information set to accurately distribute information to users, collects feedback from users after obtaining the information, and updates the data information in the database.

[0155] 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.

Claims

1. An information distribution method based on data access and big data classification, characterized in that: include: Collecting user information data, including user needs, first user identity information, second user identity information, and user history information records; Store user information data in the database and retrieve and update the data in real time; Identify and classify the user information data, read the first user identity information and the second user identity information in the database, classify the users into different levels, obtain the level category scores, and store them in the database; Inputting the information in the database into a first information similarity analysis model and a second information similarity analysis model to obtain a user similarity matrix and an interest similarity matrix respectively; inputting the user similarity matrix and the interest similarity matrix into a first information distribution model and a second information distribution model respectively to obtain a first matching information set and a second matching information set; Combining the first matching information set and the second matching information set into a comprehensive matching information set, and performing comprehensive weight calculation using a user similarity matrix and an interest similarity matrix to obtain a distribution weight value of each matching information in the comprehensive matching information set; Sort each matching information in the comprehensive matching information set according to the distribution weight value; Selecting a number of matching information from the sorted comprehensive matching information set to accurately distribute information to users, collecting feedback from users after obtaining the information, and updating each data information in the database; The first user identity information is the user's role, and the second user identity information is the user's authority; The first information similarity analysis model describes the similarity between users; the second information similarity analysis model describes the similarity between demand information; the first information similarity analysis model and the second information similarity analysis model both use cosine similarity to calculate a user similarity matrix and an interest similarity matrix respectively; Among them, the user similarity matrix S U The calculation formula for (i,j) is: Among them, u i and u j are the feature vectors of users i and j respectively; Interest similarity matrix S D The calculation formula for (k,l) is: Among them, d k and d l are the feature vectors of requirements k and l respectively; The user similarity matrix S U (i, j) and the data of the database are input into the first information distribution model to obtain a first matching information set; the interest similarity matrix S D (k, l) and the data of the database are input into the second information distribution model to obtain a second matching information set; In the first information distribution model, first, the demand score R of user j for demand k is calculated. j,k , the demand score R j,k It includes actual rating and predicted rating; the actual rating is the real rating given by the user to the requirement k; the formula of the predicted rating is: Among them, the prediction score Use the demand score R of similar users j,k To predict; Neighbors(i) is a set of users similar to user i, S U (i,j) is the user similarity matrix between user i and user j; Secondly, using the demand score R j,k Calculate the demand forecast score M1(i,k)=∑ j S U (i,j)·R j,k ; Then, the user needs are screened according to the demand prediction score M1(i,k) as the first matching information set N; In the second information distribution model, first, the interest score C of user i for demand l is calculated. i,l : C i,l =u i ·c l ; Among them, u i is the feature vector of user i; c l is the characteristic vector of demand l; Secondly, using the interest score C i,l Calculate the content relevance score M2(i,k)=∑ i S D (k,l)·C i,l ; Then, the user requirements are screened according to the content-related scores M2(i,k) as the second matching information set M.

2. The information distribution method based on data access and big data classification according to claim 1 is characterized in that: The storage of the user information data adopts distributed storage technology, and the specific implementation process is as follows: Dividing the user information data into a plurality of data blocks; distributing the data blocks to a plurality of storage nodes; and storing copies of the data blocks on the plurality of storage nodes to prevent data loss when a single node fails; A consistency protocol is adopted to ensure that the data blocks stored on the multiple storage nodes remain consistent and that all copies of the data blocks are synchronized; data requests are distributed through a load balancing mechanism to prevent a node from being overloaded; and update operations on the data blocks are synchronized to all relevant storage nodes.

3. The information distribution method based on data access and big data classification according to claim 1 is characterized in that: The specific process of the hierarchical classification is as follows: Extract the first user identity information and the second user identity information of each user from the database to obtain {U1, U2, ..., U I }; Assume that the user information feature matrix is ​​X and the label matrix is ​​Y, where the user information features refer to the user's role features and the user's permission features, and the label matrix contains hierarchical labels; Use the decision tree classification algorithm to classify users. The formula is: in, is the predicted level category, P(Y k |X) is the probability that a user belongs to a certain level category k given feature X; For each user U i Classification to get level labels The formula is Among them, Classifier() is a decision tree classifier; Finally, label each user's level Convert to hierarchical category scores And the hierarchical category scores are stored in the database.

4. The information distribution method based on data access and big data classification according to claim 1 is characterized in that: The first matching information set and the second matching information set are combined into a comprehensive matching information set A=M∪N, and the user similarity matrix and the interest similarity matrix are subjected to comprehensive weight calculation to obtain the distribution weight value of each matching information in the comprehensive matching information set. Among them, W i is the distribution weight value of the i-th information in the comprehensive matching information set; is the hierarchical category score corresponding to each user; It is the user weight matrix calculated based on the user similarity matrix; is the interest weight matrix calculated based on the interest similarity matrix; Then, based on the distribution weight value W of each matching information corresponding to each user i The matching information in the comprehensive matching information set is sorted.

5. An information distribution system based on data access and big data classification, characterized in that: include: A user information data collection unit, wherein the user information data includes user needs, first user identity information, second user identity information, and user history information records; A user information data storage unit stores the user information data in a database and retrieves and updates the data in real time; A user information data identification and classification unit, which reads the first user identity information and the second user identity information in the database, classifies the user into different levels, obtains the level category scores, and stores them in the database; A similarity calculation unit, inputting the information in the database into a first information similarity analysis model and a second information similarity analysis model to obtain a user similarity matrix and an interest similarity matrix respectively; A matching information collection unit, inputting the user similarity matrix and the interest similarity matrix into a first information distribution model and a second information distribution model respectively, to obtain a first matching information set and a second matching information set; a weight calculation unit, combining the first matching information set and the second matching information set into a comprehensive matching information set, performing comprehensive weight calculation using a user similarity matrix and an interest similarity matrix, and obtaining a distribution weight value of each matching information in the comprehensive matching information set; Sort each matching information in the comprehensive matching information set according to the distribution weight value; An information distribution unit selects a number of matching information from the sorted comprehensive matching information set to accurately distribute information to users, collects feedback from users after obtaining the information, and updates each data information in the database; The first user identity information is the user's role, and the second user identity information is the user's authority; The first information similarity analysis model describes the similarity between users; the second information similarity analysis model describes the similarity between demand information; the first information similarity analysis model and the second information similarity analysis model both use cosine similarity to calculate a user similarity matrix and an interest similarity matrix respectively; Among them, the user similarity matrix S U The calculation formula for (i,j) is: Among them, u i and u j are the feature vectors of users i and j respectively; Interest similarity matrix S D The calculation formula for (k,l) is: Among them, d k and d l are the feature vectors of requirements k and l respectively; The user similarity matrix S U (i, j) and the data of the database are input into the first information distribution model to obtain a first matching information set; the interest similarity matrix S D (k, l) and the data of the database are input into the second information distribution model to obtain a second matching information set; The process of obtaining the first matching information set is as follows: first, calculate the demand score R of user j for demand k j,k , the demand score R j,k It includes actual rating and predicted rating; the actual rating is the real rating given by the user to the requirement k; the formula for the predicted rating is: Among them, the prediction score Use the demand score R of similar users j,k To predict; Neighbors(i) is a set of users similar to user i, S U (i, j) is the user similarity matrix between user i and user j; secondly, using the demand score R j,k Calculate the demand forecast score M1(i,k)=∑ j S U (i,j)·R j,k ; Then, the user needs are screened according to the demand prediction score M1(i,k) as the first matching information set N; The process of obtaining the second matching information set is as follows: first, calculate the interest score C of user i for requirement l i,l :C i,l =u i ·c l , where u i is the feature vector of user i; c l is the feature vector of demand l; secondly, using the interest score C i,l Calculate the content relevance score M2(i,k)=∑ i S D (k,l)·C i,l , where S D (k, l) is the interest similarity matrix between requirements k and l; then, the user requirements are screened according to the content-related scores M2(i, k) as the second matching information set M.

6. The information distribution system based on data access and big data classification according to claim 5 is characterized in that: The specific process of the hierarchical classification is as follows: assuming that the user information feature matrix is ​​X and the label matrix is ​​Y, wherein the user information features refer to the user's role features and the user's authority features, and the label matrix contains hierarchical labels; Use decision tree classification algorithm to classify users. in, is the predicted level category, P(Y k |X) is the probability that the user belongs to a certain level category k when the feature X is given; then for each user U i Classification to get level labels Among them, Classifier() is a decision tree classifier; finally, the level label of each user Convert to hierarchical category scores And the hierarchical category scores are stored in the database.

7. The information distribution system based on data access and big data classification according to claim 5 is characterized in that: The first matching information set N and the second matching information M are combined into a comprehensive matching information set A=M∪N, and the user similarity matrix and the interest similarity matrix are subjected to comprehensive weight calculation to obtain the distribution weight value of each matching information in the comprehensive matching information set. Among them, W i is the distribution weight value of the i-th information in the comprehensive matching information set; is the hierarchical category score corresponding to each user; It is the user weight matrix calculated based on the user similarity matrix; is the interest weight matrix calculated based on the interest similarity matrix; then, based on the distribution weight value W of each matching information corresponding to each user i The matching information in the comprehensive matching information set is sorted.

Images