Data resource recommendation method and system for trusted data space

Abstract

The invention discloses a data resource recommendation method and system for a trusted data space. The method comprises the following steps: receiving metadata description update from a data market, and constructing an association graph according to the updated metadata description; constructing a user query vector based on the behavior information of the user and the association map; acquiring a data set matrix of the data market; calculating according to the user query vector and the data set matrix to obtain a recommendation result; sorting and filtering the recommendation results; and feeding back the sorted and filtered recommendation result to the user. According to the method, privacy protection without data out of the domain is realized. The behavior data and preference information of the user are both kept locally, so that the risk of sensitive data leakage is avoided, and the security of user privacy is ensured. The data market not only stores data, but also can directly undertake calculation tasks, for example, a data set matrix is generated based on a TF-IDF algorithm, the burden of local calculation nodes is relieved, and the performance and efficiency of a recommendation system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data resource recommendation, and in particular to a data resource recommendation method and system in a trusted data space. Background Art

[0002] With the continuous development of information technology, the amount of data has shown explosive growth. In order to quickly obtain information related to users from massive data information and reduce users' search costs, recommendation systems have come into being. The construction of a recommendation system is the practical application of algorithms in engineering. Therefore, in a recommendation system, it is generally possible to consider determining the similarity of words based on the distance between word vectors, or construct a user relationship graph based on the relationship between users, the relationship between resources, and the relationship between users and resources, and quickly mine users' interests and hobbies through the relationship graph. However, in a recommendation system, a large amount of computing resources are usually required. If a traditional recommendation system is moved to a data space, the following problems may arise:

[0003] 1. User behavior data is not allowed to leave the domain, that is, user-purchased data resources and user collection resources are not allowed to be stored in centralized nodes, nor are they allowed to be recommended to competitors, in order to avoid leaking user behavior and inferring the purpose of using data, thereby exposing commercial purposes.

[0004] 2. As a centralized node in the data market, it may only provide data storage functions but no data computing resources. If all data recommendations are placed in the centralized node, it will cause computing overload, and the recommendation algorithm performance and recommendation efficiency will not meet the requirements.

[0005] There is no effective solution to the problems of privacy protection difficulties and low computational efficiency in recommendation methods in existing technologies. Summary of the invention

[0006] The embodiments of the present invention provide a data resource recommendation method and system for a trusted data space to solve the problems of difficulty in privacy protection and low computational efficiency in the recommendation method in the prior art.

[0007] To achieve the above-mentioned object, on the one hand, the present invention provides a data resource recommendation method of a trusted data space, the method comprising: S1, receiving metadata description updates from a data market, and constructing a correlation map according to the updated metadata description; S2, constructing a user query vector based on user behavior information and the correlation map; S3, obtaining a data set matrix from the data market;

[0008] S4. Calculate the recommendation results according to the user query vector and the data set matrix; S5. Sort and filter the recommendation results; S6. Feedback the sorted and filtered recommendation results to the user.

[0009] Optionally, S2 includes: dynamically assigning weights to metadata descriptions of different categories according to user behavior information; and constructing a user query vector according to the user behavior information, weight assignment and association graph.

[0010] Optionally, S5 includes: sorting the recommendation results according to the user's behavior information and based on the relationship between the user and the data resources, and the relationship between the data resources; and filtering the sorted recommendation results.

[0011] Optionally, filtering the sorted recommendation results includes: filtering purchased data, filtering data that has been blacklisted by the user.

[0012] Optionally, constructing an association graph according to the updated metadata description includes: periodically constructing an association graph according to the updated metadata description.

[0013] On the other hand, the present invention provides a data resource recommendation system in a trusted data space, the system comprising: an association graph construction unit, used to receive metadata description updates from a data market, and to construct an association graph based on the updated metadata description; a query vector construction unit, used to construct a user query vector based on user behavior information and the association graph; a data set matrix acquisition unit, used to acquire a data set matrix from a data market; a calculation unit, used to calculate a recommendation result based on the user query vector and the data set matrix; an optimization unit, used to sort and filter the recommendation results; and a feedback unit, used to feed back the sorted and filtered recommendation results to the user.

[0014] Optionally, the query vector construction unit includes: a weight allocation subunit, used to dynamically allocate weights to metadata descriptions of different categories according to user behavior information; and a query vector construction subunit, used to construct a user query vector according to user behavior information, weight allocation and association graph.

[0015] Optionally, the optimization unit includes: a sorting subunit, used to sort the recommendation results according to user behavior information and based on the relationship between users and data resources, and the relationship between data resources; and a filtering subunit, used to filter the sorted recommendation results.

[0016] Optionally, filtering the sorted recommendation results includes: filtering purchased data, filtering data that has been blacklisted by the user.

[0017] Optionally, constructing an association graph according to the updated metadata description includes: periodically constructing an association graph according to the updated metadata description.

[0018] Beneficial effects of the present invention:

[0019] The present invention constructs a user query vector on the user's local connector in a trusted data space and obtains a data set matrix from the data market; calculates the recommendation result based on the user query vector and the data set matrix; and implements privacy protection without leaving the domain. The user's behavior data and preference information are retained locally, avoiding the risk of sensitive data leakage and ensuring the security of user privacy. The data market of the present invention not only stores data, but also directly undertakes computing tasks, such as generating a data set matrix based on the TF-IDF algorithm, which reduces the burden on local computing nodes and improves the performance and efficiency of the recommendation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of a data resource recommendation method for a trusted data space provided by an embodiment of the present invention;

[0021] Figure 2 It is a structural diagram of a data resource recommendation system of a trusted data space provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] If the traditional recommendation system is moved to the data space, problems such as difficulty in privacy protection and low computational efficiency may arise.

[0024] A trusted data space is a data space used for data circulation, which includes data providers, data demanders, and a centralized metadata description center. In order to maximize the promotion of data circulation in the data space, we need to explore the interests of data demanders as much as possible. The more accurate the query and recommendation of data, the more beneficial it is to data transactions between data providers and data demanders. Therefore, the present invention proposes a logical framework based on the data space and designs a recommendation system that conforms to the current framework. While realizing data resources within the trusted data space, it can maximize the use of user local behavior data, combine the characteristics of data resources, and mine the best recommendation results for users, thereby promoting the circulation of data resources in the data space.

[0025] Taking the agricultural trusted data space as an example, suppose there is an agricultural user named Zhang San in our agricultural trusted data space, who focuses on growing wheat. In the agricultural trusted data space, Zhang San has queried data related to soil, climate, and market prices in the past, and has paid attention to information such as soil fertility and irrigation plans. Based on this data, the system can provide Zhang San with personalized recommendations to help him make the best decision.

[0026] Figure 1 A data resource recommendation method for a trusted data space provided by an embodiment of the present invention is as follows: Figure 1 As shown, the method includes:

[0027] S1. Receive metadata description updates from the data market and build a correlation graph based on the updated metadata description;

[0028] In an optional embodiment, suppose that the agricultural data market receives a new set of metadata descriptions. Here, "metadata descriptions" refer to the description information of the data itself, without involving specific, sensitive original data content. Metadata descriptions include information about the title, source, category, change, data relationship, etc. of the data, but do not contain specific values ​​such as actual agricultural production data, user behavior data, market prices, etc. This design ensures that no specific agricultural data, user data or other sensitive information will be leaked during the operation of the data market, thereby effectively protecting user privacy and data security.

[0029] The updated metadata descriptions include "soil fertility report", "wheat growth monitoring data", "irrigation plan recommendation", "soil moisture data", etc. The system calculates the correlation between these metadata descriptions through algorithms and builds a correlation map for them. For example: soil fertility report is highly correlated with wheat growth monitoring data, because soil fertility directly affects the growth status of wheat, and irrigation plan recommendation is closely related to soil moisture data. This correlation map provides an important source of information for subsequent recommendations.

[0030] Furthermore, before S1, the method further includes:

[0031] All metadata descriptions published by users are published to the data market.

[0032] By only publishing metadata description information rather than the original data itself, the present invention effectively avoids the risk of user data leakage. Even if the data market or other third parties access this information, they cannot directly obtain the user's sensitive data content (such as actual soil parameters, market prices, etc.).

[0033] Furthermore, constructing a correlation graph according to the updated metadata description includes:

[0034] Regularly build a relationship graph based on updated metadata descriptions.

[0035] That is, the local connector of the user (Zhang San) synchronizes with the data market at regular intervals to obtain the latest metadata description and builds a correlation graph based on the latest metadata description.

[0036] Since metadata descriptions are updated regularly, the user's local connector will continuously and dynamically update the association graph based on the new description information. This update mechanism enables the recommendation system to always make recommendations based on the latest resource associations, thereby improving the relevance and accuracy of the recommendation results.

[0037] S2, constructing a user query vector based on the user's behavior information and association graph;

[0038] The S2 includes: dynamically assigning weights to metadata descriptions of different categories according to the user's behavior information; and constructing a user query vector according to the user's behavior information, weight assignment and association graph.

[0039] In an optional implementation, user behavior information refers to a record of the user's interaction with the system (local connector), including but not limited to:

[0040] Browsing history: which data resources the user has visited (e.g., a certain type of soil data, market price data, etc.);

[0041] Click history: which recommendations the user clicked in the recommendation system;

[0042] Purchase history: resources that users have purchased or downloaded;

[0043] Collection information: data resources collected by users;

[0044] Search keywords: Keywords entered by users in the search box reflect the direction of their interests.

[0045] These behavioral information can reflect the user's interest preferences and demand directions. Therefore, this information will serve as an important basis when constructing query vectors.

[0046] The association graph reflects the relationship and similarity between different data resources, and can reveal which data resources may be related or interdependent. By adding data associated with the user's behavioral information, it reflects the user's possible related interests and provides richer information for constructing the query vector.

[0047] The query vector is a mathematical expression used to describe the user's interest preferences and current needs. Based on the user's behavior information and association graph, the system converts this information into a vector representation to form a multi-dimensional query vector.

[0048] Behavior information is converted into vectors: Each user's behavior (such as clicks, browsing, and favorites) will be converted into a vector, indicating the contribution of the behavior to the user's interest. For example, if a user frequently accesses soil data, the vector component of this part will increase, indicating the user's interest in soil data.

[0049] Adding related graph information to the query vector: By analyzing the relationship between data resources, related graph information will affect the construction of the query vector. If a user frequently views a certain type of data (such as soil data), the system will add data related to that data resource (such as meteorological data or irrigation plans) to the query vector to reflect the user's possible related interests.

[0050] Ultimately, the user query vector integrates the user's behavior history, interest preferences, and associations with other data resources. It is a multi-dimensional vector representation that can accurately reflect the user's current needs.

[0051] Furthermore, based on the construction of user query vectors based on user behavior information and association graphs, the personalization and accuracy of the recommendation system are further enhanced by assigning weights to different categories of data. This weighting mechanism adjusts the weights of different categories of data based on the user's historical behavior (such as browsing, collection, click, etc.), so that the recommendation system can prioritize the data resources that users are more interested in.

[0052] The specific weight distribution rules are as follows:

[0053] Soil data: For example, Zhang San has collected data related to soil quality, and the system will recognize that he has a strong interest in soil data. Therefore, the system sets the weight of soil data resources to 3, indicating that this category of data is what Zhang San is most interested in.

[0054] Crop growth data: If Zhang San browses relevant data about crop growth, the system will also weight it according to the user's browsing behavior. Assume that the weight of crop growth data is 2, which reflects the user's interest in this category.

[0055] Other categories of data: For other categories of data, assuming that Zhang San has not clearly expressed interest, the weight of this data will be set to 1, which is the default weight.

[0056] When constructing a query vector, in addition to constructing a basic vector based on user behavior information and association graphs, the vectors of each data category are weighted according to the above weight assignment. By assigning weights to different categories of data, the system can better match the user's actual interests and dynamically adjust the recommended content based on these preferences.

[0057] S3, obtain the data set matrix of the data market;

[0058] Specifically, the data set matrix of the data market is generated by all metadata descriptions of the data market through the TF-IDF algorithm.

[0059] Metadata description is a textual description of data resources. The TF-IDF (Term Frequency-Inverse Document Frequency) algorithm converts text into vector representation by calculating the importance of each word in a document collection. It is an important tool for generating dataset matrices.

[0060] The TF-IDF algorithm is used to convert each metadata description into a vector, and the vectors converted from all metadata descriptions form a data set matrix. Through the data set matrix generated by the TF-IDF algorithm, the system can effectively identify and evaluate the similarities between different data resources, thereby providing more accurate and relevant recommendation results.

[0061] S4. Calculate and obtain a recommendation result based on the user query vector and the data set matrix;

[0062] In the user's local connector, the recommendation result is determined by calculating the similarity between the user query vector and the dataset matrix;

[0063] Similarity calculation: Calculate the similarity between the user query vector and each row in the dataset matrix (i.e. each metadata description). By calculating the similarity between the user query vector and each metadata description, the system will calculate a similarity score for each metadata description. Usually, the system will select the top N metadata descriptions with the highest similarity as the recommendation results.

[0064] S5. Sort and filter the recommendation results;

[0065] Specifically, S5 includes:

[0066] Sorting the recommendation results according to the user's behavior information and based on the relationship between the user and the data resources, and the relationship between the data resources;

[0067] Specifically, if the recommended results include soil fertility report, market price trend, and irrigation scheme recommendation, and Zhang San clicks on the collection and browses the soil fertility report, the system will sort the recommended results according to these behaviors. The data resources that Zhang San is interested in recently are recommended first, and the priority is recommended according to his browsing behavior of market price trends and the close correlation between soil fertility report and irrigation method recommendation.

[0068] The recommended sorting results may be:

[0069] 1. Soil fertility report

[0070] 2. Market price trends

[0071] 3. Recommended irrigation plan

[0072] The sorted recommendation results are filtered.

[0073] Filtering the sorted recommendation results includes: filtering purchased data and filtering data that has been blacklisted by the user.

[0074] If Zhang San has already purchased certain data (for example, he has purchased the "Soil Fertility Report"), the system will filter out the purchased data and will not push it. If Zhang San shows no interest in certain types of data (such as blacklisting the market price trend in a certain area), the system will automatically remove the data to avoid pushing irrelevant content.

[0075] S6. Feedback the sorted and filtered recommendation results to the user.

[0076] After S6, the method further includes:

[0077] After the user operates, the user behavior data is fed back to the user behavior center.

[0078] After Zhang San receives the recommended data, operations such as clicks, browsing, favorites, and blacklisting will be fed back to Zhang San's local connector and fed back to the behavior recording center, which will be recorded and used for the next query of related data.

[0079] Figure 2 is a schematic diagram of the structure of a data resource recommendation system for a trusted data space provided by an embodiment of the present invention. Figure 2 As shown, the system includes:

[0080] The association graph construction unit 201 is used to receive metadata description updates from the data market and construct an association graph according to the updated metadata description;

[0081] The step of constructing a correlation graph according to the updated metadata description includes:

[0082] Regularly build a relationship graph based on updated metadata descriptions.

[0083] A query vector construction unit 202, configured to construct a user query vector based on the user's behavior information and the association graph;

[0084] The query vector construction unit includes:

[0085] A weight assignment subunit, used to dynamically assign weights to metadata descriptions of different categories based on user behavior information;

[0086] The query vector construction subunit is used to construct a user query vector based on the user's behavior information, weight distribution and association graph.

[0087] A data set matrix acquisition unit 203, used to acquire a data set matrix of a data market;

[0088] A calculation unit 204, configured to calculate a recommendation result according to the user query vector and the data set matrix;

[0089] An optimization unit 205, configured to sort and filter the recommendation results;

[0090] The optimization unit comprises:

[0091] A sorting subunit, used to sort the recommendation results according to the user's behavior information and based on the relationship between the user and the data resources, and the relationship between the data resources;

[0092] The filtering subunit is used to filter the sorted recommendation results.

[0093] Filtering the sorted recommendation results includes: filtering purchased data and filtering data that has been blacklisted by the user.

[0094] The feedback unit 206 is used to feed back the sorted and filtered recommendation results to the user.

[0095] The system provided by the present invention corresponds to the above method, and the specific implementation of the system will not be repeated here.

[0096] The present invention effectively improves the personalization, accuracy and efficiency of the recommendation system by combining user behavior analysis, data set matrix generated by TF-IDF algorithm, user query vector, similarity calculation, sorting and filtering, etc., which not only ensures the protection of user privacy, but also improves the response speed and processing capacity of the system. By providing more accurate and intelligent recommendation results, the present invention can significantly optimize the resource circulation of the agricultural data market, promote the digital transformation of the agricultural industry, and improve user experience and satisfaction.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data resource recommendation method for a trusted data space, characterized in that: include: S1. Receive metadata description updates from the data market and build a correlation graph based on the updated metadata description; S2, constructing a user query vector based on the user's behavior information and association graph; S3, obtain the data set matrix of the data market; S4. Calculate and obtain a recommendation result based on the user query vector and the data set matrix; S5. Sort and filter the recommendation results; S6. Feedback the sorted and filtered recommendation results to the user.

2. The method according to claim 1, characterized in that: The S2 includes: Dynamically assign weights to different categories of metadata descriptions based on user behavior information; Construct user query vector based on user behavior information, weight distribution and association graph.

3. The method according to claim 1, characterized in that: The S5 includes: Sorting the recommendation results according to the user's behavior information and based on the relationship between the user and the data resources, and the relationship between the data resources; The sorted recommendation results are filtered.

4. The method according to claim 3, characterized in that: Filtering the sorted recommendation results includes: filtering purchased data and filtering data that has been blacklisted by the user.

5. The method according to claim 1, characterized in that The step of constructing a correlation graph according to the updated metadata description includes: Regularly build a relationship graph based on updated metadata descriptions.

6. A data resource recommendation system for a trusted data space, characterized in that: include: A correlation graph construction unit, used to receive metadata description updates from the data market and construct a correlation graph according to the updated metadata description; A query vector construction unit, used to construct a user query vector based on the user's behavior information and the association graph; A data set matrix acquisition unit, used to acquire the data set matrix of the data market; A calculation unit, used to calculate the recommendation result according to the user query vector and the data set matrix; An optimization unit, used for sorting and filtering the recommendation results; The feedback unit is used to feed back the sorted and filtered recommendation results to the user.

7. The system according to claim 6, characterized in that The query vector construction unit includes: A weight assignment subunit, used to dynamically assign weights to metadata descriptions of different categories based on user behavior information; The query vector construction subunit is used to construct a user query vector based on the user's behavior information, weight distribution and association graph.

8. The system according to claim 6, characterized in that The optimization unit comprises: A sorting subunit, used to sort the recommendation results according to the user's behavior information and based on the relationship between the user and the data resources, and the relationship between the data resources; The filtering subunit is used to filter the sorted recommendation results.

9. The system according to claim 8, characterized in that: Filtering the sorted recommendation results includes: filtering purchased data and filtering data that has been blacklisted by the user.

10. The system according to claim 6, characterized in that The step of constructing a correlation graph according to the updated metadata description includes: Regularly build a relationship graph based on updated metadata descriptions.

Images