Supplier optimal promotion path query method based on government procurement knowledge graph

Abstract

The present invention relates to a supplier optimal promotion path query method based on a government procurement knowledge graph, and belongs to the field of knowledge graph construction and application technology. The present invention utilizes the advantages of knowledge graphs that are good at processing data with complex relationships, good at integrating data with different structures, and good at using knowledge applications to solve practical problems. The entities in the government procurement and "government procurement loan" financing process are abstracted as nodes, the frequent business transactions between entities are abstracted as the relationship between nodes, and the entire process is abstracted as a data graph model. The financing platform can use which partners to promote the target supplier, which can be converted into a path query problem in the data graph, and the knowledge graph is enabled to promote the government procurement supplier on the financing platform.

Description

Technical Field

[0001] The present invention relates to the field of knowledge graph construction and application technology, and specifically to the construction of a government procurement knowledge graph and a method for querying the optimal promotion path of suppliers. Background Art

[0002] The concept of knowledge graphs was born in 2012, when Google proposed and built the original knowledge graph. Because knowledge graphs accurately reflect real-world facts and effectively express abstract knowledge such as concepts and hierarchies, they have been applied to various fields in recent years, garnering extensive research.

[0003] Graph models can be applied in urban road information networks, geographic traffic information systems, biological systems, computer networks, and operations research. For example, social network graphs can be used to represent the degree of intimacy between people in a network, urban road traffic networks can be used to describe the connectivity between urban roads, and bioinformatics networks can be used to describe the order of proteins and genes in natural organisms. Real-world objects (such as people, urban hubs, protein molecules, gene fragments, etc.) are abstracted as vertices in a data graph, and specific relationships and connections between objects are abstracted as edges between two vertices. In this way, various real-world problems can be abstracted into data graph models, and then various real-world problems can be studied and analyzed based on data graph models, simplifying complex real-world problems and making research and analysis more convenient.

[0004] In recent years, government procurement has played an increasingly prominent role in supporting the development of small and medium-sized enterprises. To address the funding shortages faced by successful government procurement suppliers, a new financing platform has emerged, addressing the shortcomings of the existing "Government Procurement Loan" system. This platform allows suppliers, especially small and medium-sized enterprises, to enjoy unsecured, low-interest, simple, and fast financing services with a single government procurement contract. This addresses the difficulties, slowness, and high costs associated with government procurement suppliers' financing. However, these excellent platforms and services are far from being widely known by suppliers. While there are many precedents for the application of knowledge graphs in government affairs, existing government knowledge graphs lack specificity, suffer from low query efficiency, and lack integration with data from internet-based big data and specific application platforms. Summary of the Invention

[0005] The patent of this invention designs a method for constructing a government procurement knowledge graph and a method for querying the optimal promotion path of the financing platform to suppliers.

[0006] The present invention first provides a method for constructing a government procurement knowledge graph, and the steps of the construction method are:

[0007] Step 101. A top-down approach is used to construct a knowledge graph. First, based on government procurement data, enterprise business data, and promotion needs, and with expert advice, a graph database model is constructed, centered on government procurement suppliers and aimed at better promoting suppliers on the financing platform. Based on the model, the entities and entity attributes required to construct the knowledge graph are extracted. Next, relationships and relationship attributes are extracted based on the entities.

[0008] Step 102. Data collection and preprocessing,

[0009] (1) Data collection: including data export and crawling;

[0010] (2) Data cleaning: Clean and deduplicate the collected data, including removing duplicate records and erroneous data to ensure the accuracy and completeness of the data;

[0011] (3) Data integration: Integrate the collected data into a dataset, including unifying the data format and naming;

[0012] Step 103. Knowledge storage: Use a graph database to store the knowledge graph, and build a graph database based on the knowledge graph model based on the data stored in the intermediate library.

[0013] The above-mentioned method for constructing the government procurement knowledge graph of the present invention can be applied to the query of the optimal promotion path from the financing platform to the supplier.

[0014] The present invention also provides a method for querying the optimal promotion path for suppliers based on a government procurement knowledge graph. The query method is based on the knowledge graph constructed by the above-mentioned method for constructing a government procurement knowledge graph, and measures the closeness of the relationship between entity nodes by calculating the number of transactions and transaction amounts between the two, thereby selecting the optimal promotion path. The method includes the following steps:

[0015] Step 201: Use the breadth-first algorithm to search for all promotion paths within three layers from the financing platform node to the target supplier node as candidate paths;

[0016] Step 202: Calculate the weights between nodes based on the number of business transactions and the total amount of business between nodes;

[0017] Step 203: Define the calculation method of the promotion cost between nodes;

[0018] Step 204: Define the weighted promotion cost of each layer and calculate the average weight of each path;

[0019] Step 205: The path with the largest average weight among the alternative paths is the optimal promotion path.

[0020] Furthermore, the specific steps of step 201 are:

[0021] (1) Treat the government procurement knowledge graph as an undirected and unweighted graph, traverse the first-level neighbor nodes of the financing platform node, and if there is a target supplier node, stop traversing and return the path information and store it in a list;

[0022] (2) Continue to traverse the secondary neighbor nodes of the financing platform node in sequence. If there is a target supplier node, return the path information and store it in the list;

[0023] (3) Continue to traverse the third-level neighbor nodes of the financing platform node in sequence. If there is a target supplier node, return the path information and store it in the list;

[0024] (4) All reachable paths within three layers from the financing platform node to the target supplier node are obtained. If there is no path, no path is returned.

[0025] Furthermore, in step 202, the benchmark item amount is first determined based on the bank's loan preference, and the relationship weight is calculated using the number of transactions as the primary evaluation indicator and the total transaction amount as the secondary evaluation indicator. The calculation process is as follows:

[0026] (1) Calculate the average amount AVA of business items between two nodes:

[0027]

[0028] In the formula, number is the number of transactions, amount is the total transaction amount;

[0029] (2) Calculate the weight of the total amount of business between the two nodes amount , the calculation formula is as follows:

[0030]

[0031] In the formula, the benchmark project amount is set to M, and the value is assigned based on the minimum financing amount preferred by commercial banks. When AVA ≤ M, the function will directly return 0, which means there is no additional weight; otherwise, it will calculate the weight corresponding to the average amount AVA. To control the weight of the business amount, a logarithmic function is used to convert the total amount. The logarithmic function can effectively narrow the gap between amounts, and for larger amounts, the corresponding weight growth rate will slow down. To limit the upper limit of the weight of the project amount of suppliers with large project amounts, the min function is used to ensure that the weight does not exceed N, that is, the business amount is at most equivalent to N business times;

[0032] (3) Calculate the weight between two nodes. The calculation formula is as follows:

[0033] weight=number+weight amount

[0034] Furthermore, the calculation formula of the promotion cost between the two nodes in step 203 is as follows:

[0035]

[0036] The weight between the two nodes obtained in step 202 is used to measure the promotion cost. The larger the weight between the two nodes, the smaller the promotion cost between the two nodes. Therefore, a variant of the Logistic function of the S-shaped function is used to calculate the cost. k controls the steepness of the function curve. The larger k is, the steeper the function curve is. b and c are translation parameters, and a is a scaling parameter, which can control the position and size of the function.

[0037] Furthermore, the step 204 uses the average value of the weight after deducting the weighted promotion cost As a measure of the optimal path. In a path, as the number of path hops increases, its promotion cost also increases sharply, so an exponential function is used to weight the promotion cost, and the weighted promotion cost is defined as cost×2 i-1 , using the weight weight between the two nodes and the weighted promotion cost cost×2 i-1 The difference between the two nodes reflects the actual promotion effect between the two nodes, and the average value of the difference in the entire path is used. As a measure of the optimal path. Average path weight The calculation formula is as follows:

[0038]

[0039] Among them, hop refers to the number of hops in the path, cost×2 i-1 Represents the promotion cost after weighting the number of hops, weight-cost×2 i-1 represents the weight of the promotion cost after deducting the weighted hop count, It represents the sum of the weights of the promotion cost after deducting the number of hops in the path and then finding the average value. It is the evaluation index of the optimal promotion path.

[0040] In the alternative path of step 205 The largest one is the optimal promotion path.

[0041] Compared to the existing technology, the advancements of the method for constructing a government procurement knowledge graph and the method for querying the optimal promotion path from the platform to suppliers provided by the patent of this invention are: the invention provides a knowledge graph and path query in the field of government procurement financing platform analysis and promotion, summarizes entity and relationship models, provides a construction model for the specific construction process, and provides a graph visualization for related decision analysis. The constructed knowledge graph is then applied to the promotion of the financing platform to suppliers, making the promotion of the financing platform more accurate and efficient. This makes the circulation and sharing of financing platform data more efficient, and can promote business restructuring and process reengineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Attachment Figure 1 It is a flowchart for building a knowledge graph;

[0043] Attachment Figure 2 It is a flow chart of the method for querying the optimal promotion path of government procurement suppliers;

[0044] Attachment Figure 3 It is a schematic diagram of the overall model of the knowledge graph of the relationship between government procurement suppliers;

[0045] Attachment Figure 4 It is the graph of the promotion cost function. DETAILED DESCRIPTION

[0046] The present invention is further described below with reference to the accompanying drawings and specific embodiments. This embodiment, based on a certain government procurement financing platform, provides a clear and complete description of the technical solution. The described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0047] like Figure 1 As shown, this embodiment describes a method for constructing a government procurement knowledge graph, the purpose of which is to extract entities and relationships from data from various government parties to construct a knowledge graph of government procurement supplier relationships, and apply it to the financing platform's query on the optimal promotion path for government procurement suppliers.

[0048] The present invention first provides a method for constructing a government procurement knowledge graph, and the steps of the construction method are:

[0049] Step 101. Use a top-down approach to build a knowledge graph. First, based on government procurement data, corporate industrial and commercial data, financing platform data and business needs, refer to the opinions of platform experts, and build a graph database model, such as Figure 2 As shown;

[0050] Step 102. Data collection and preprocessing,

[0051] (1) Data collection: Use the Python library requests to interact with the API interface provided by the platform to obtain data; obtain enterprise information from the API interface of the third-party enterprise information query platform; use the Python library Scrapy to crawl government procurement information from the government procurement information disclosure platform.

[0052] (2) Data cleaning: Data cleaning is to remove noise, errors and duplicate records in the data to ensure the accuracy and completeness of the data. The main data cleaning methods include:

[0053] ① Data deduplication: Identify and delete duplicate records in the data set to ensure the uniqueness of the data.

[0054] ②Data format conversion: Convert data into a format suitable for analysis, such as converting text data into numeric data, or converting date and time into a standard format.

[0055] ③ Missing value processing: Identify and process missing values in the data by filling default values, interpolating, or deleting rows / columns with missing values.

[0056] ④ Outlier processing: Detect and process outliers in the data by deleting or replacing them with reasonable values or using statistical methods.

[0057] (3) Data integration: Data integration is the process of integrating the collected data into a data set, making its format and naming consistent for subsequent data use. This mainly includes the following methods:

[0058] ① Standardized naming: unify the naming rules of data fields, using unified uppercase and lowercase, underscore or camel case naming.

[0059] ②Data format conversion: Convert data from different data sources into the same format, and convert date formats into a unified date format.

[0060] ③Data merging: Data comes from multiple sources and is merged into a unified data set through key field ID.

[0061] Step 103. Knowledge storage: construct a graph database based on the knowledge graph model for the data stored in the intermediate library, and use Neo4j graph database to store the knowledge graph.

[0062] like Figure 3As shown, the present invention is based on the knowledge graph constructed by the above-mentioned government procurement knowledge graph construction method, and also proposes a supplier optimal promotion path query method based on the government procurement knowledge graph. The government procurement financing platform often has multiple reachable paths for suppliers that need to be promoted. The closeness of the relationship between entity nodes is measured using a method based on frequency and amount, which is mainly reflected by the number of transactions and the amount of transactions between the two nodes. The more times and the larger the amount, the closer the relationship between the two entities. The closeness of the relationship between the two entity nodes, the financing platform and the supplier, is measured by calculating the number of transactions and the amount of transactions between the two entity nodes, and the optimal promotion path is selected, which includes the following steps:

[0063] Step 201: Use the breadth-first algorithm to search for all promotion paths within three layers from the financing platform node to the target supplier node as candidate paths;

[0064] (1) Treat the government procurement knowledge graph as an undirected and unweighted graph, traverse the first-order neighbor nodes of the financing platform node, and if there is a target supplier node, stop traversing and return the path information and store it in a list;

[0065] (2) Continue to traverse the second-order neighbor nodes of the financing platform node in sequence. If there is a target supplier node, return the path information and store it in the list;

[0066] (3) Continue to traverse the third-order neighbor nodes of the financing platform node in sequence. If there is a target supplier node, return the path information and store it in the list;

[0067] (4) All reachable paths within three layers from the financing platform node to the target supplier node are obtained. If there is no path, no path is returned.

[0068] Step 202: Calculate the weights between nodes based on the number of transactions between nodes and the total transaction amount.

[0069] Based on real cases on the platform, banks are more inclined to lend to projects with a value of more than 500,000 yuan. From a promotional perspective, we can appropriately increase the weight of projects with a value of more than 500,000 yuan. Therefore, we use the number of transactions as the primary evaluation indicator and the total transaction amount as the secondary evaluation indicator to calculate the relationship weight. The calculation process is as follows:

[0070] (1) Calculate the average amount AVA of business items between two nodes:

[0071]

[0072] In the formula, number is the number of transactions, amount is the total transaction amount;

[0073] (2) Calculate the weight of the total amount of business between the two nodes amount , the calculation formula is as follows:

[0074]

[0075] In the formula, the base project amount is set to 500,000 yuan. When AVA ≤ 500,000, the function will directly return 0, which means there is no additional weight. Otherwise, it will calculate the weight corresponding to the average amount AVA. To limit the upper limit of the weight of the project amount of suppliers with large project amounts, the min function is used to ensure that the weight does not exceed 10, that is, the business amount is at most equivalent to 10 business times.

[0076] (3) Calculate the weight between two nodes. The calculation formula is as follows:

[0077] weight=number+weight amount

[0078] Step 203: Define the calculation method of the promotion cost between nodes;

[0079] Finding the most suitable promotion path from all candidate paths is essentially finding the optimal promotion path. It's important to consider that the weight between nodes is negatively correlated with the closeness between them: the greater the weight, the closer the relationship, and vice versa. Using the concept of a distance-weighted calculation function, the promotion cost between nodes should be inversely proportional to the weight between the nodes. The greater the weight between the two nodes, the lower the promotion cost for that hop.

[0080] like Figure 4 As shown in the figure, the calculation formula of the promotion cost between two nodes is as follows:

[0081]

[0082] The weight between the two nodes obtained in step 202 is used as the input of the function. To control the value range of the promotion cost between 0 and 6, let k = 1, c = -5, b = 1.1, a = 5. The specific function is as follows: Figure 4 As shown in the graph of the promotion cost function. Figure 4 As shown, when there are 1 or 2 transactions between two nodes, the cost of promotion is high; when the weight is between 3-7, the cost of promotion decreases rapidly; when the weight is above 7, the cost decreases slowly.

[0083] Step 204: Calculate the average path weight;

[0084] Considering that in actual promotion, the corresponding promotion cost will increase with each additional hop, the product of promotion cost and exponential function is used to express the high-level promotion cost cost×2 i-1 ;

[0085] The average value of the weight after deducting the weighted promotion cost As a measure of the optimal path, the average weight value The largest reachable path within three layers is the optimal promotion path; the average weight of the path The calculation formula is as follows:

[0086]

[0087] Among them, hop refers to the number of hops in the path, cost×2 i-1 Represents the promotion cost after weighting the number of hops, weight-cost×2 i-1 represents the weight of the promotion cost after deducting the weighted hop count, It represents the sum of the weights of the promotion cost after deducting the number of hops in the path and then finding the average value. It is the evaluation index of the optimal promotion path.

[0088] Step 205: The path with the largest average weight among the alternative paths is the optimal promotion path.

[0089] In addition, the present invention also proposes a computer device for constructing a government procurement knowledge graph and a method for querying the optimal promotion path from the platform to suppliers.

[0090] The computer device includes a memory and one or more processors. The memory stores executable code, and when the processor executes the executable code, it is used to implement the above-mentioned government procurement knowledge graph construction method and optimal promotion path query method.

[0091] In this embodiment, the computer device is any device or apparatus with data processing capability.

[0092] An embodiment of the present invention also provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, it is used to implement the above-mentioned government procurement knowledge graph construction method and optimal promotion path query method.

[0093] The computer-readable storage medium can be an internal storage unit of any device or apparatus with data processing capabilities, such as a hard disk or memory, or an external storage device of any device with data processing capabilities, such as a plug-in hard disk, smart media card, SD card, flash card, etc. equipped on the device.

[0094] The above contents are merely preferred embodiments of the present invention and are not intended to limit the scope of implementation of the present invention. That is, any simple equivalent changes and modifications made in accordance with the claims and description of the present invention are still within the scope of the present invention.

Claims

1. A supplier optimal promotion path query method based on government procurement knowledge graph, characterized by: The query method is based on a knowledge graph constructed using a method for constructing a government procurement knowledge graph. By calculating the number of transactions and transaction amounts between entity nodes in the knowledge graph, the closeness of the relationship is measured, and the optimal promotion path for the financing platform to reach suppliers is selected. The query method includes the following steps: Step 201: Use a breadth-first algorithm to search for all promotion paths within three layers from the financing platform node to the target supplier node as candidate paths; Step 202. Calculate the weights between nodes based on the number of transactions and the total transaction amount between nodes; Step 203. Define a calculation method for promotion costs between nodes; Step 204. Define the weighted promotion cost of each layer and calculate the average weight of each path; Step 205. The path with the largest average weight among the alternative paths is the optimal promotion path; The calculation formula of the promotion cost between two nodes in step 203 is as follows: The weight between the two nodes obtained in step 202 is used to measure the promotion cost. The larger the weight between the two nodes, the smaller the promotion cost between the two nodes. Therefore, a variant of the logistic function of the sigmoid function is used to calculate the cost. k controls the steepness of the function curve. The larger k is, the steeper the function curve is. b and c are translation parameters, and a is a scaling parameter that can control the position and size of the function. The step 204 uses the average value of the weight after deducting the weighted promotion cost As a measure of the optimal path, in a path, as the number of path hops increases, its promotion cost also increases sharply. The exponential function is used to weight the promotion cost, and the weighted promotion cost is defined as , using the weight between two nodes and the weighted promotion cost The difference between the two nodes reflects the actual promotion effect between the two nodes, and the average value of the difference in the entire path is used As a measure of the optimal path; average path weight The calculation formula is as follows: ， Among them, hop refers to the number of hops in the path. represents the promotion cost after hop weighting, represents the weight of the promotion cost after deducting the weighted hop count, It represents the sum of the weights of the promotion cost after deducting the number of hops in the path and then finding the average value. , which is the evaluation index of the optimal promotion path; In the step 205, the alternative path The largest one is the optimal promotion path.

2. A supplier optimal promotion path query method based on government procurement knowledge graph according to claim 1, characterized in that: The steps of the method for constructing the government procurement knowledge graph are as follows: Step 101. Build a knowledge graph using a top-down approach. First, based on government procurement data, supplier business data, and promotion needs, and with expert advice, build a graph database model centered on government procurement suppliers and aiming to better promote suppliers on the financing platform. Based on this model, extract the entities and entity attributes required to build the knowledge graph. Next, extract relationships and relationship attributes based on the entities. Step 102. Data collection and preprocessing, (1) Data collection: including data export and crawling; (2) Data cleaning: Clean and deduplicate the collected data, including removing duplicate records and erroneous data to ensure the accuracy and completeness of the data; (3) Data integration: Integrate the collected data into a data set, including unifying the data format and naming; Step 103. Knowledge storage: construct a graph database based on the knowledge graph model for the data stored in the intermediate library, and use the graph database to store the knowledge graph.

3. The method for querying supplier's optimal promotion path based on government procurement knowledge graph according to claim 2 is characterized in that: The specific steps of step 201 are: Treat the government procurement knowledge graph as an undirected and unweighted graph, traverse the first-level neighbor nodes of the financing platform node, and if there is a target supplier node, stop traversal and return the path information and store it in a list; Continue to traverse the secondary neighbor nodes of the financing platform node in sequence. If there is a target supplier node, return the path information and store it in the list; Continue to traverse the third-level neighbor nodes of the financing platform node in sequence. If there is a target supplier node, return the path information and store it in the list; This results in all reachable paths within three layers from the financing platform node to the target supplier node. If there is no path, no path is returned.

4. The method for querying supplier optimal promotion paths based on government procurement knowledge graph according to claim 3 is characterized in that: In step 202, the benchmark item amount is first determined based on the commercial bank's loan preference, and the relationship weight is calculated using the number of transactions as the primary evaluation indicator and the total transaction amount as the secondary evaluation indicator. The calculation process is as follows: (1) Calculate the average amount AVA of business items between two nodes: In the formula, number is the number of transactions, amount is the total transaction amount; (2) Calculate the weight of the total amount of business between two nodes , the calculation formula is as follows: In the formula, the benchmark project amount is set to M. The value of this value refers to the minimum financing amount preferred by commercial banks. When AVA≤M, the function will directly return 0, which means there is no additional weight. Otherwise, it will calculate the weight corresponding to the average amount AVA. In order to limit the upper limit of the weight of the project amount of suppliers with large project amounts, it uses the min function to ensure that the weight does not exceed N, that is, the business amount is at most equivalent to N business times; (3) Calculate the weight between two nodes. The calculation formula is as follows: 。 5. Application of the supplier optimal promotion path query method based on the government procurement knowledge graph as described in claim 4 in querying the supplier's optimal promotion path on a financing platform.

Images