Recommendation method, recommendation system and electronic device based on graph knowledge

Abstract

The present invention discloses a recommendation method, a recommendation system and an electronic device based on graph knowledge. The recommendation method based on graph knowledge includes: step S1: obtaining data provided by a user and sorting the data; step S2: extracting graph elements from the sorted data and constructing knowledge graph data based on the graph elements; step S3: training a model according to the knowledge graph data; step S4: obtaining a recommendation result through the trained model according to real-time data. The present invention displays the correlation between previous behaviors and results in the form of an association relationship, and can clearly see that the relationship between certain behaviors and entities is relatively close. This close relationship can be subsequently bound to make relevant adjustments in a targeted manner in the subsequent target optimization scenario.

Description

Technical Field

[0001] The present invention relates to a recommendation method, a recommendation system and an electronic device, and more specifically, to a recommendation method, a recommendation system and an electronic device based on a knowledge graph interpretation displayed by a group of users. Background Art

[0002] With the rapid development of the Internet, the rapid expansion of information, and the accelerated pace of user behavior, there is an increasing demand to quickly obtain information about users' direct or potential needs. Under this demand, the application of recommendation systems is becoming more and more extensive.

[0003] Customers not only need to view the various data indicators of the recommended location, but also want to view the associations in the recommendation process. Through the knowledge graph, a clear association display is constructed to view the rationality of the recommended content, and deeper insights can be drawn through the analysis of associations.

[0004] In the prior art, the data report corresponding to the recommendation position displays data indicators of different behaviors (such as exposure, clicks, purchases, likes, etc.). However, it is found in practice that there are still the following shortcomings:

[0005] 1. The tables of data indicators must be completely consistent in time and cannot show the characteristics of behavioral sequences and correlations. Viewing behavioral data separately may result in omissions of data phenomena or lead to wrong conclusions.

[0006] 2. Convert the user's own characteristics and behaviors into labels, and display the data of various dimensions of a single label or multiple labels.

[0007] 3. There is no time sequence or correlation between tags, and the data cannot be displayed layer by layer, resulting in the effect of analyzing from a single dimension downward.

[0008] Therefore, there is an urgent need to develop a recommendation method, recommendation system and electronic device based on graph knowledge to overcome the above-mentioned defects. Summary of the invention

[0009] In view of the above problems, the present invention provides a recommendation method based on graph knowledge, which includes:

[0010] Step S1: Obtain data provided by the user and organize the data;

[0011] Step S2: extracting graph elements from the sorted data and constructing knowledge graph data based on the association relationship between the graph elements;

[0012] Step S3: Perform model training according to the knowledge graph data;

[0013] Step S4: Obtain recommendation results through the trained model based on real-time data.

[0014] The above-mentioned recommendation method based on graph knowledge further includes:

[0015] Step S5: Output a single user association trajectory by combining the user's historical behavior and recommendation results.

[0016] The above-mentioned recommendation method based on graph knowledge further includes:

[0017] Step S6: extract the data of each user, aggregate and count them, and output the statistical results.

[0018] In the above-mentioned recommendation method based on graph knowledge, the data includes: user data and material data, and the user data includes user behavior and user characteristics.

[0019] In the above-mentioned recommendation method based on graph knowledge, the step S1 includes defining a label structure for the data.

[0020] In the above-mentioned recommendation method based on graph knowledge, step S2 includes:

[0021] Step S21: Processing the data through word segmentation and semantic processing;

[0022] Step S22: performing entity extraction, relationship extraction and event extraction;

[0023] Step S23: Match the labels, association relationships and entities to form knowledge graph data.

[0024] In the above-mentioned recommendation method based on graph knowledge, the step S5 includes selecting the historical behavior of the user referenced by the model and recording the classification of different behaviors of the user, and outputting the association trajectory of a single user according to the historical behavior, classification and the recommendation result.

[0025] In the above-mentioned recommendation method based on graph knowledge, step S6 includes extracting the user behavior and user characteristics of each user, performing aggregation statistics based on different characteristics as dimensions and outputting the statistical results.

[0026] The present invention also provides a recommendation system based on graph knowledge, which includes:

[0027] Data processing unit: obtains data provided by users and organizes the data;

[0028] Construction unit: extracting graph elements from the sorted data and constructing knowledge graph data based on the association relationship between the graph elements;

[0029] Training unit: performing model training according to the knowledge graph data;

[0030] Result output unit: obtains recommendation results through the trained model based on real-time data;

[0031] Trajectory output unit: combines user historical behavior and recommendation results to output a single user's associated trajectory;

[0032] Statistical result output unit: extracts data from each user for aggregate statistics and outputs statistical results.

[0033] The present invention also provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the graph knowledge-based recommendation method described in any one of the items through the computer program.

[0034] In summary, compared with the prior art, the effectiveness of the present invention lies in that: the present invention can use the labeling of entity relationships of the knowledge graph and the results with strong correlation relationships, summarize the behavioral characteristics of individual users, summarize the common behavioral trajectories and behavioral characteristics of group users, and use graphical displays to clearly show the correlation relationships, which is convenient for quickly discovering data phenomena and performing data analysis; at the same time, based on the diversity and complexity of the correlation relationships of the present invention, the relationship graph structure can be displayed with emphasis.

[0035] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 is a flow chart of a recommended method of the present invention;

[0038] Figure 2 for Figure 1 Flow chart of step S2;

[0039] Figure 3 It is a schematic diagram of the application of the recommended method of the present invention;

[0040] Figure 4 This is an exemplary diagram showing the recommended method of the present invention;

[0041] Figure 5 It is a schematic diagram of the structure of the recommendation system of the present invention;

[0042] Figure 6 It is a schematic diagram of the hardware structure of the electronic device of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0044] The exemplary embodiments and descriptions of the present invention are used to explain the present invention, but are not intended to limit the present invention. In addition, elements / components with the same or similar reference numerals used in the drawings and embodiments are used to represent the same or similar parts.

[0045] The terms “first”, “second”, “S1”, “S2”, etc. used in this document do not specifically refer to an order or sequence, nor are they used to limit the present invention. They are only used to distinguish elements or operations described with the same technical terms.

[0046] The directional terms used in this document, such as up, down, left, right, front or back, etc., are only used to refer to the directions of the drawings. Therefore, the directional terms used are used to illustrate and not to limit the present invention.

[0047] The words “include,” “including,” “have,” “contain,” etc. used in this document are open-ended terms, meaning including but not limited to.

[0048] As used herein, "and / or" includes any or all combinations of the items described.

[0049] As used herein, “plurality” includes “two” and “more than two”; as used herein, “plurality of groups” includes “two groups” and “more than two groups”.

[0050] The terms "substantially" and "approximately" used herein are used to modify any quantity or error that may vary slightly, but these slight changes or errors do not change their essence. Generally speaking, the range of slight changes or errors modified by such terms may be 20% in some embodiments, 10% in some embodiments, 5% in some embodiments, or other values. Those skilled in the art should understand that the aforementioned values ​​can be adjusted according to actual needs and are not limited thereto.

[0051] Certain terms used to describe the present application are discussed below or elsewhere in this specification to provide additional guidance to those skilled in the art regarding the description of the present application.

[0052] Knowledge Graph, known as knowledge domain visualization or knowledge domain mapping map in the library and information industry, is a series of various graphs that show the development process and structural relationship of knowledge. It uses visualization technology to describe knowledge resources and their carriers, and to mine, analyze, construct, draw and display knowledge and their interrelationships.

[0053] Knowledge graph is a modern theory that combines the theories and methods of applied mathematics, graphics, information visualization technology, information science and other disciplines with methods such as metrology citation analysis and co-occurrence analysis, and uses visual graphs to vividly display the core structure, development history, frontier fields and overall knowledge architecture of disciplines to achieve the purpose of multidisciplinary integration. It can provide practical and valuable reference for disciplinary research.

[0054] The knowledge graph is a graph-based data structure consisting of nodes (points) and edges (Edges). Each node represents an "entity", and each edge is a "relationship" between entities. The knowledge graph is essentially a semantic network. Entities can refer to things in the real world, such as people, place names, companies, telephones, animals, etc.; relationships are used to express a certain connection between different entities. From the above figure, you can see that entities include place names and people; Dali belongs to Yunnan, Xiao Ming lives in Dali, Xiao Ming and Xiao Qin are friends, these are all relationships between entities. Popular definition: The knowledge graph is a relationship network obtained by connecting all different types of information together, so the knowledge graph provides the ability to analyze problems from the perspective of "relationships".

[0055] The graph knowledge-based recommendation method of the present invention, after constructing a knowledge graph and training an applicable model in the early stage, can automatically combine user behavior data, output and store user behavior trajectories, extract key dimensions, and perform summary statistics on key dimensions. At the same time, it can display the relationship between key entities in different dimensions, and can circle the scope of different dimensions for detailed viewing and analysis.

[0056] Its nodes represent entities or concepts, and its edges represent various semantic relationships between entities / concepts.

[0057] Please refer to Figure 1 , Figure 1 Flow chart of the recommended method of the present invention. Figure 1 As shown, the recommendation method based on graph knowledge of the present invention includes:

[0058] Step S1: Obtain the data provided by the user and organize the data. Wherein, the step S1 includes defining a label structure for the data. Specifically, the data includes: user data and material data, and the user data includes user behavior and user characteristics. Organize the data according to the user data and material data provided by the customer. If there are clear and regular labels in the provided data, they can be used directly. If the labels in the provided data are unclear, they need to be reprocessed. Define the label structure for the data first: for example, determine how many labels to distinguish, a course can have labels such as lecturer, price, and category; for example, what is the dictionary value of each label and what is the range, such as the specific age of the user can be within a range of 10 years old.

[0059] Step S2: extract graph elements from the sorted data and construct knowledge graph data based on the association relationship between the graph elements.

[0060] Please refer to Figure 2 , Figure 2 for Figure 1 Flow chart of step S2 in FIG. Figure 2 As shown, the step S2 includes:

[0061] Step S21: Processing the data through word segmentation and semantic processing;

[0062] Step S22: performing entity extraction, relationship extraction and event extraction;

[0063] Step S23: Match the labels, association relationships and entities to form knowledge graph data.

[0064] Specifically, in this step, for information such as titles and introductions, word segmentation and semantic processing can be used. If the basic information exists in the link, the basic information can be first obtained through a crawler, and then text processing can be performed to correspond labels, associations, etc. with entities.

[0065] Relation extraction:

[0066] Definition: Automatically identify a semantic relationship between entities. According to the number of participating entities, it can be divided into binary relationship extraction (two entities) and multi-relation extraction (three or more entities).

[0067] By focusing on the semantic relationship between two entities, we can get the (arg1, relation, arg2) triple, where arg1 and arg2 represent two entities, and relation represents the semantic relationship between the entities. (For example, the semantic relationship between words in a sentence can be obtained through the Hanlp analysis tool).

[0068] Relation extraction method:

[0069] Limited domain relationship extraction method:

[0070] Template-based relation extraction method: extract and identify entity relationships in text through templates obtained through manual editing or learning. Limited by the quality and coverage of templates, it is not very scalable. (Court documents made by myself belong to template-based extraction)

[0071] Relation extraction method based on machine learning: regard relationship extraction as a classification problem

[0072] Among them, the relationship extraction methods based on machine learning can be divided into supervised and weakly supervised.

[0073] Supervised relation extraction method:

[0074] Feature engineering-based methods: It is necessary to explicitly convert relation instances into feature vectors that the classifier can accept.

[0075] Kernel function-based method: directly takes the structure tree as the processing object, and uses the kernel function instead of the inner product of the eigenvector when calculating the distance between relationships.

[0076] Neural network-based methods: Automatically learn effective feature representations directly from the input text, end-to-end

[0077] Weakly supervised relationship extraction method: does not require manual labeling of large amounts of data.

[0078] Distance supervision: Use open knowledge graphs to automatically annotate training samples without the need for manual annotation. This is a type of weakly supervised relationship extraction.

[0079] Open domain relation extraction method:

[0080] There is no need to predefine relationship categories, and some words in the context of entity pairs are used to describe the relationship between entities.

[0081] Event extraction and method:

[0082] Definition: Extract events of interest to users from text describing event information and present them in a structured form.

[0083] Steps: First, identify the event and its type, then identify the elements involved in the event (usually entities), and finally determine the role each element plays in the event.

[0084] Event extraction related concepts:

[0085] Event reference: A natural language description of a specific objective event, usually a sentence or sentence group.

[0086] Event trigger word: refers to the word that best represents the occurrence of an event in an event reference. It is an important feature that determines the event category. It is usually a verb or noun.

[0087] Event elements: participants in an event, mainly composed of entities, time and attribute values

[0088] Element role: What role does the event element play in the corresponding event?

[0089] Event category: Event elements and trigger words determine the category of the event (the category defines several subcategories)

[0090] Limited domain event extraction: Before extraction, the type of target event and the specific structure of each type (which specific event elements are included) are pre-defined, and usually a certain amount of labeled data is given.

[0091] Limited domain event extraction method:

[0092] Pattern matching-based method: The identification and extraction of a certain type of event is carried out under the guidance of some patterns (steps: pattern acquisition, pattern matching)

[0093] Supervised event pattern matching: Pattern acquisition is completely based on manually annotated corpus

[0094] Weakly supervised event pattern matching: No need to fully annotate the corpus, only manual pre-classification of the corpus or formulation of a small number of seed patterns is required

[0095] Machine learning based methods

[0096] Supervised event extraction method: Modeling event extraction as a multi-classification problem

[0097] Feature engineering-based methods: It is necessary to explicitly convert event instances into feature vectors that can be accepted by the classifier. The research focus is on how to extract discriminative features.

[0098] Neural network-based method: automatically obtains features from text to complete event extraction, avoiding the error accumulation problem caused by using traditional natural language processing tools

[0099] Entity extraction:

[0100] Entity extraction or named entity recognition (NER) plays an important role in information extraction. It mainly extracts atomic information elements in text, such as names of people, names of organizations / institutions, geographic locations, events / dates, character values, monetary values, etc. There are two keywords for the entity extraction task: find&classify, which means finding named entities and classifying them.

[0101] There are three methods for entity extraction: rule-based methods usually require writing templates for the target entity and then matching them in the original corpus; statistical machine learning-based methods mainly train the original corpus through machine learning methods, and then use the trained model to identify entities; open domain-oriented extraction will be oriented towards massive Web corpora.

[0102] Step S3: Perform model training based on the knowledge graph data. Specifically, user data and material data are input into the model to train the model and obtain the initial access recommendation results.

[0103] Step S4: Obtain recommendation results through the trained model based on real-time data. Specifically, in this step, the trained model is put online and combined with real-time user behavior data to output recommendation results, that is, the recommendation results are output through the combination of continuously added new users and users' real-time data and real-time user behavior data. For example, maternal and child-related products can be recommended to users based on the user's new search for maternal and child products within 1 hour.

[0104] Step S5: Output the single user association trajectory by combining the user's historical behavior and recommendation results. Specifically, in this step, since the user's recommendation results are generally a black box for customers, this part of the content can be converted into explainable content, and the historical behavior of the user referenced by the model is selected to record the classification of different user behaviors. For example, user A purchased a finance course (expanding the price range, course duration range, lecturer information, etc. of this finance course), searched for art courses (also expanding the price range, course duration range, lecturer information, etc. of this art course), and clicked on an art course (marking whether the click / other behavior is a product recommended to the user).

[0105] Analyze user behavior trajectories, study user preferences, and provide technical support for the active service model of "information looking for users". According to the analysis results of user behavior trajectories, obtain user preference information, recommend content that suits the user's preferences, and realize the active service of "information looking for users". According to the evaluation mechanism of content recommendation, it is convenient to collect user evaluations on active services and facilitate the optimization of subsequent behavior trajectory analysis models. Optimize the trajectory analysis model in a loop and feed back the optimization results to users to realize closed-loop sustainable and precise active services.

[0106] Application scenarios:

[0107] (1) Application content recommendation. During the application process, users find that there are too many dimension conditions, the frequently used dimensions are not at the front, and the condition values ​​need to be reselected each time, which is inconvenient. Through analysis, it is found that users only frequently use a few dimensions such as city, time and brand. Solution: Recommend and display the dimensions that users frequently use, hide the dimensions that are not frequently used, and automatically arrange the dimensions according to the order of the dimensions that users frequently operate, and automatically recommend and fill in the condition values ​​that users frequently select or enter.

[0108] (2) Pay attention to application recommendations. During the application process, users find that the system is too large and has too many function nodes. They have to click through many layers each time to find the function page they need to use. In addition, the applications they pay attention to are scattered in different places, which is inconvenient to use. According to the user's application attention model, the applications that users frequently visit in their daily work are analyzed. It is found that there are too many function nodes and the tree nodes on the left side of the system are too large, which makes it very inconvenient for users to use. In addition, the applications that users pay attention to are scattered in different places, which is inconvenient to use. In addition, there are only a few KPIs and ad hoc query applications that users frequently use. Solution: Actively recommend the KPIs and ad hoc query applications that users frequently use to the user's personalized office desktop, or favorites, or homepage and other places where users can directly use the application, so that users can use these applications conveniently.

[0109] (3) Assist in key application protection. Users found that some key applications were often searched repeatedly for many times but could not find data. Key applications were confirmed based on the user's application focus model and system optimization model, and the concentrated access time period of key applications was analyzed. Solution: The background UTAP adjusts the task priority according to the user's access time requirements to ensure that key applications are executed first.

[0110] Step S6: Extract the data of each user, aggregate and count them, and output the statistical results. Specifically, in this step, extract the user and behavior characteristics of each user, aggregate and count them with different characteristics as dimensions (for example, view the aggregated data with region, category, price range, etc. as the main dimensions), and according to the result of step S5, supplement the user's label, such as the region where the behavior of a single user occurs, the user's gender, age and other demographic information, the user's occupation and other work information, and then aggregate the data of users with the same behavior and the same label to obtain the correlation between the user label and the recommended content, so as to obtain the reasons for group user recommendations and the recommendation effect, which can be displayed in a visual way on the interface for customers to understand their own user behavior preferences and recommendation effects, which can in turn guide the formulation of operation strategies and assist in targeted promotion activities.

[0111] Please refer to Figure 3 , Figure 3Schematic diagram of the application of the recommended method of the present invention. Figure 3 As shown, the recommendation method based on graph knowledge of the present invention generally has the following process when applied:

[0112] 1. Obtain data provided by customers and organize the data;

[0113] 2. Graph construction, including entity extraction, relationship extraction, and event extraction;

[0114] 3. Use the constructed data to train the model;

[0115] 4. After the trained model is launched, it will be combined with real-time user behavior data to output recommendation results;

[0116] 5. Combine user historical behavior and recommendation results to output a single user association trajectory;

[0117] 6. Extract the user and behavior characteristics of each user, and aggregate statistics based on different characteristics (for example, view summary data based on region, category, price range, etc. as the main dimensions).

[0118] Please refer to Figure 4 , Figure 4 This is an example diagram showing the recommended method of the present invention. Figure 4 As shown, the recommendation method of the present invention displays the correlation between previous behaviors and results in the form of association relationships, which can clearly show that the relationship between certain behaviors and entities is relatively close.

[0119] In another embodiment of the present invention, a label marking method can also be used, and multiple label dimensions can be combined to view possible association relationships.

[0120] Please refer to Figure 5 , Figure 5 Schematic diagram of the structure of the recommendation system of the present invention. Figure 5 As shown, the recommendation system of the present invention includes:

[0121] Data processing unit 11: obtains data provided by the user and organizes the data, wherein the data processing unit 11 defines a tag structure for the data;

[0122] Construction unit 12: extracting graph elements from the sorted data and constructing knowledge graph data based on the association relationship between the graph elements;

[0123] Training unit 13: performing model training according to the knowledge graph data;

[0124] Result output unit 14: obtains recommendation results through the trained model according to real-time data;

[0125] Trajectory output unit 15: outputs a single user associated trajectory in combination with the user's historical behavior and the recommendation result, wherein the trajectory output unit 15 selects the historical behavior of the user referenced by the model and records the classification of different user behaviors, and outputs a single user associated trajectory according to the historical behavior, classification and the recommendation result;

[0126] The statistical result output unit 16 extracts the data of each user, aggregates and counts them, and outputs the statistical results. The statistical result output unit 16 extracts the user behavior and user characteristics of each user, aggregates and counts them based on different characteristics, and outputs the statistical results.

[0127] The data include: user data and material data, and the user data include user behavior and user characteristics.

[0128] Furthermore, after the construction unit 12 processes the data through word segmentation and semantic processing, the construction unit 12 performs entity extraction, relationship extraction and event extraction; then the construction unit 12 matches the labels, association relationships and entities to form knowledge graph data.

[0129] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored program, wherein when the program is run, the recommendation method based on graph knowledge described in any one of the above items is executed.

[0130] The present invention also provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the graph knowledge-based recommendation method described in any one of the items through the computer program.

[0131] Specifically, combined with Figure 1 The entity recommendation method described in the embodiment of the present application can be implemented by an electronic device. Figure 6 It is a schematic diagram of the hardware structure of the electronic device of the present invention.

[0132] The electronic device may include a processor 81 and a memory 82 storing computer program instructions.

[0133] Specifically, the processor 81 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0134] Among them, the memory 82 may include a large-capacity memory for data or instructions. By way of example and not limitation, the memory 82 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 82 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 82 may be inside or outside the data processing device. In a specific embodiment, the memory 82 is a non-volatile memory. In a specific embodiment, the memory 82 includes a read-only memory (ROM) and a random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (Programmable Read-Only Memory, PROM for short), an erasable PROM (Erasable ProgrammableRead-Only Memory, EPROM for short), an electrically erasable PROM (Electrically Erasable ProgrammableRead-Only Memory, EEPROM for short), an electrically alterable ROM (Electrically Alterable Read-Only Memory, EAROM for short) or a flash memory (FLASH) or a combination of two or more of these. Under appropriate circumstances, the RAM can be a static random access memory (SRAM) or a dynamic random access memory (DRAM), wherein the DRAM can be a fast page mode dynamic random access memory (FPMDRAM), an extended data output dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0135] The memory 82 may be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 81 .

[0136] The processor 81 implements any one of the recommendation methods in the above embodiments by reading and executing computer program instructions stored in the memory 82 .

[0137] In some of the embodiments, the electronic device may further include a communication interface 83 and a bus 80. Figure 6 As shown, the processor 81, the memory 82, and the communication interface 83 are connected via a bus 80 and communicate with each other.

[0138] The communication interface 83 is used to implement communication between the modules, devices, units and / or equipment in the embodiment of the present application. The communication port 83 can also implement data communication with other components such as: external devices, image / data acquisition equipment, databases, external storage, and image / data processing workstations.

[0139] The bus 80 includes hardware, software or both, and couples the components of the electronic device to each other. The bus 80 includes but is not limited to at least one of the following: a data bus, an address bus, a control bus, an expansion bus, and a local bus. By way of example and not limitation, bus 80 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses or a combination of two or more of these. Where appropriate, bus 80 may include one or more buses. Although embodiments of the present application describe and illustrate a particular bus, the present application contemplates any suitable bus or interconnect.

[0140] The electronic device can display the correlation between previous behaviors and results in the form of association relationships, so as to clearly see that certain behaviors and entities are closely related. Subsequently, this close relationship can be bound to make relevant adjustments in the subsequent target optimization scenario, thereby achieving the combination of Figure 1 Describe the recommended approach.

[0141] In summary, the present invention displays the correlation between previous behaviors and results in the form of association relationships, which can clearly show that certain behaviors and entities have a close relationship. This close relationship can be subsequently bound to make relevant targeted adjustments in subsequent target optimization scenarios.

[0142] 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 substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A recommendation method based on graph knowledge, characterized in that: include: Step S1: obtaining data provided by the user and sorting the data, defining a tag structure for the data, determining the type of tags to be distinguished, the size of the dictionary value of each tag, and the range of the dictionary value of each tag; Step S2: extracting graph elements from the sorted data and constructing knowledge graph data based on the association relationship between the graph elements; Step S3: Perform model training according to the knowledge graph data; Step S4: Obtain recommendation results through the trained model according to real-time data; Step S5: Outputting a single user association trajectory based on the user's historical behavior and recommendation results; Step S6: extract the data of each user, aggregate and count them, and output the statistical results; The data include: user data and material data, and the user data include user behavior and user characteristics.

2. The recommendation method based on graph knowledge as claimed in claim 1, characterized in that: The step S2 includes: Step S21: Processing the data through word segmentation and semantic processing; Step S22: performing entity extraction, relationship extraction and event extraction; Step S23: Match the labels, association relationships and entities to form knowledge graph data.

3. The recommendation method based on graph knowledge as claimed in claim 1, characterized in that: The step S5 includes selecting the historical behavior of the user referenced by the model and recording the classification of different behaviors of the user, and outputting a single user association trajectory according to the historical behavior, classification and the recommendation result.

4. The recommendation method based on graph knowledge as claimed in claim 1, characterized in that: The step S6 includes extracting the user behavior and user features of each user, performing aggregation statistics based on different features as dimensions, and outputting the statistical results.

5. A recommendation system based on graph knowledge, characterized in that: include: Data processing unit: obtains data provided by the user and organizes the data, defines a tag structure for the data, determines the type of tags to be distinguished, the size of the dictionary value of each tag, and the range of the dictionary value of each tag; Construction unit: extracting graph elements from the sorted data and constructing knowledge graph data based on the association relationship between the graph elements; Training unit: performing model training according to the knowledge graph data; Result output unit: obtains recommendation results through the trained model based on real-time data; Trajectory output unit: combines user historical behavior and recommendation results to output a single user's associated trajectory; Statistical result output unit: extracts data from each user for aggregation and statistics and outputs statistical results; The data include: user data and material data, and the user data include user behavior and user characteristics.

6. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor is configured to execute the graph knowledge-based recommendation method described in any one of claims 1 to 4 through the computer program.

Images