Big model agent-based investment attraction clue mining and analysis system

Abstract

The invention relates to the technical field of investment attraction, and discloses an investment attraction clue mining and analysis system based on a large model agent, and the system comprises the following modules: a key link information collection module which is used for collecting key link information; the first graph structure data encoding module is used for encoding the key link information into first graph structure data; the key link acquisition module is used for inputting the first graph structure data into a key link model and outputting key links representing an industrial chain required by an economic development area; the investment attraction clue information acquisition module is used for acquiring investment attraction clue information; the second graph structure data encoding module is used for encoding the investment attraction clue information into second graph structure data; and the investment attracting plan acquisition module is used for inputting the second graph structure data into the investment attracting model and outputting and representing an investment attracting plan. According to the method, the investment attraction work can be more accurately carried out for key links of different industrial chains in different regions according to the result.

Description

Technical Field

[0001] The present invention relates to the technical field of investment promotion, and more specifically, to an investment promotion clue mining and analysis system based on a large model intelligent body. Background Art

[0002] The new energy vehicle industry is an important direction for the current global economic development and is of great significance to the transformation and upgrading of local economies and sustainable development. Many regions are actively developing new energy vehicle industries, hoping to seize the opportunity and create a new engine for local economic development. Against this background, economic development zones, as an important platform for undertaking industrial transfer and cultivating emerging industries, have proposed a plan to build a new energy vehicle industry cluster.

[0003] In order to achieve this goal, the economic development zone needs to introduce a number of key enterprises in the key links of the new energy vehicle industry chain to form a more complete industrial ecology. What the economic development zone needs is a more complete industrial chain. It may not be possible to form a complete industrial chain by only introducing specific key links. Moreover, the conditions of each economic development zone are different, and the key links required at different times may also be different. Therefore, in order to attract investment for key enterprises in these key links more accurately, it is necessary to conduct in-depth research on the key links and key enterprises required by the economic development zone, and a more complete investment promotion plan is needed to carry out investment promotion work in a targeted manner. Summary of the invention

[0004] The present invention provides an investment promotion clue mining and analysis system based on a large model intelligent body, which solves the technical problems in the related technology that the key links of the industrial chain required by different economic development zones in different periods are different and it is difficult to obtain a more complete investment promotion plan.

[0005] The present invention provides a method for mining and analyzing investment leads based on a large model agent, comprising the following steps:

[0006] Step 100, collecting key link information, which includes all links in the industrial chain and information on economic development zones;

[0007] The industry chain links include: battery material supply, motor material supply, power electronic device supply, lightweight material supply, power battery production, drive motor production, power electronic control system production, vehicle design, body manufacturing, vehicle assembly, charging station and charging pile construction, operator service system, battery recycling, material reuse, core software system development, testing and inspection services, logistics and marketing services;

[0008] The information of economic development zones includes: location advantage information, resource information, industrial foundation information, human resources information, supporting facilities information, policy support information, and development goal information;

[0009] Step 200, encoding the key link information into No. 1 graph structure data; the graph structure data includes nodes and edges connecting the nodes, and a node represents an industrial chain link or location advantage or resource information or industrial foundation or human resources or supporting facilities or policy support or development goals;

[0010] It indicates that there are edges between the nodes of all industrial chain links, that there are edges between industrial chain links and location advantages when the location advantage brings convenience to the industrial chain links, that there are edges between industrial chain links and resource information when resource information is beneficial to the production and processing of industrial chain links, that there are edges between industrial chain links and industrial base when the existing industrial base and industrial chain links have upstream and downstream relationships or collaborative relationships, that there are edges between industrial chain links and human resources when the quality or quantity of human resources meets the production needs of the industrial chain, that there are edges between industrial chain links and supporting facilities when supporting facilities can support the smooth operation of industrial chain links, that there are edges between industrial chain links and policy support when policy support can attract and promote the development of industrial chain links, and that there are edges between industrial chain links and development goals when development goals attach importance to industrial chain links;

[0011] Indicates that the characteristics of the industrial chain links, location advantages, resource information, industrial base, human resources, supporting facilities, policy support, and development goals are equal to the corresponding node characteristics;

[0012] The characteristics of industrial chain links, location advantages, resource information, industrial base, human resources, supporting facilities, policy support, and development goals are obtained through feature engineering;

[0013] Step 300, input the No. 1 graph structure data into the key link model, the key link model includes a first hidden layer and a first output layer, wherein the first hidden layer inputs the No. 1 graph structure data and outputs the first hidden state to the first output layer, and the first output layer outputs the key link of the industrial chain required for the economic development zone.

[0014] In a preferred embodiment, the calculation formula of the first hidden layer is as follows:

[0015] ;

[0016] in Indicates the first picture structure data The first hidden state of a node, It is the node with the structure data of graph No.1 The set of nodes with edges between them, Indicates the first picture structure data The node characteristics of the nodes, is a normalizing constant (such as node degree), represents the weight parameter of the No. 1 graph, represents the bias parameter of Figure 1, It is a S-shaped function.

[0017] In a preferred embodiment, the calculation formula of the first output layer is as follows:

[0018] ;

[0019] in Represents a vector, the cth component of the vector represents the probability of the cth key link, Indicates the first picture structure data The first hidden state of a node, Indicates the concatenation of the first hidden states of all nodes. A collection of nodes representing the links in the industrial chain. is the first weight parameter, is the first bias parameter, Represents a sigmoid function.

[0020] Step 400, collecting investment information, which includes key links of the industrial chain, enterprise information and economic development zone information;

[0021] Enterprise information includes: enterprise name, type, scale, establishment time, legal representative, business scope, registered address, registered capital, number of employees, core technology, number of patents, and market share.

[0022] Step 500, encoding the investment lead information into No. 2 graph structure data, the graph structure data includes nodes and edges connecting the nodes, and a node represents a key link of the industrial chain or an enterprise or an economic development zone;

[0023] There are edges between the nodes representing all key links. When there are supply chain relationships or competitive relationships between enterprises, there are edges between enterprises. When an enterprise is engaged in a key link, there is an edge between the enterprise and this key link. There are edges between the economic development zone and all nodes representing key links.

[0024] The node characteristics of the node representing the enterprise are equal to the enterprise characteristics, the node characteristics of the node representing the key link are equal to the key link characteristics, and the node characteristics of the node representing the economic development zone are equal to the economic development zone characteristics;

[0025] Enterprise characteristics, key link characteristics and economic development zone characteristics are obtained through feature engineering;

[0026] Encode the investment lead information of all nodes into sequence data, the sequence data includes n sequence units, and the t-th sequence unit represents the investment lead information of all nodes at the t-th moment;

[0027] Step 600, input the No. 2 graph structure data into the investment promotion model, the model includes a second hidden layer, a third hidden layer and a second output layer, wherein the second hidden layer inputs the No. 2 graph structure data, outputs the second hidden state to the third hidden layer, the third hidden layer outputs the third hidden state to the second output layer, and the second output layer outputs the investment promotion plan.

[0028] In a preferred embodiment, the calculation formula of the second hidden layer is as follows:

[0029] ;

[0030] in The tth sequence unit of the second sequence data is represented by The second hidden state of the node of nodes, The tth sequence unit of the second sequence data is represented by The aggregation coefficient of nodes, represents the set of nodes that have edges with the vth node of the tth sequence unit of the second sequence data, represents the activation function, represents the graph weight coefficient;

[0031] ;

[0032] ;

[0033] ;

[0034] in and The tth sequence unit of the second sequence data is represented by and The node characteristics of the nodes, and The tth sequence unit of the second sequence data is represented by and The linear transformation characteristics of nodes, represents the aggregation weight coefficient, represents the splicing weight coefficient, T represents transposition, represents an exponential function with a natural constant as the base, LeakyRelu represents a modified linear unit function, Indicates The set of nodes with which there are edges.

[0035] In a preferred embodiment, the calculation formula of the third hidden layer is as follows:

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] in, , , represents the first, second and third sequence weight parameters, , , represents the first, second and third sequence bias parameters, represents the dot product, , and Represent the first, second, and third intermediate states respectively, where , The tth sequence unit of the second sequence data is represented by The second hidden state of the node, Indicates the first t-th sequence unit of the second sequence data The set of nodes with which there are edges. and They represent the t-th and t-1-th third hidden states respectively, n≥t≥1, n represents the total number of sequence units of the second sequence data, and when t=1 , tanh is the hyperbolic tangent function, Represents a sigmoid function.

[0041] In a preferred embodiment, the calculation formula of the second output layer is as follows:

[0042] ;

[0043] in Represents a vector, the i-th component of the vector represents the probability value of the i-th action; the set of all actions is called the action space, which contains all possible actions that may be selected. represents the tth third hidden state, represents the concatenation of the 1st to nth third hidden states, n represents the total number of sequence units of the second sequence data, is the second weight parameter, is the second bias parameter, Represents a sigmoid function.

[0044] In a preferred embodiment, both the key link model and the investment promotion model are trained using reinforcement learning, wherein the model to be trained is the key link model or the investment promotion model;

[0045] The steps of training the model to be trained include:

[0046] Step 201, constructing a target network, the target network copies the model to be trained;

[0047] Step 202: randomly obtain an experience from the experience pool. The experience includes the state at time t. , the action performed at time t , the state at time e+1 , perform the action Rewards ;

[0048] The states at time e and e+1 respectively represent the key link information or investment lead information at the e-th and e+1-th time input to the model to be trained;

[0049] Then calculate the state The target network output corresponds to The probability value ;

[0050] ;

[0051] express The maximum probability value of the target network output;

[0052] Step 203, according to and Calculate the loss value loss and update the model to be trained;

[0053] ;

[0054] express The output of the model to be trained corresponds to The probability value of

[0055] is the discount factor, a value between 0 and 1;

[0056] Step 204, updating the target network every fixed number of network updates so that its parameters are the same as those of the current model to be trained;

[0057] Step 205, until the number of collected experiences reaches a set threshold or the model to be trained converges, the step is terminated.

[0058] In a preferred embodiment, a large model agent-based investment lead mining and analysis system is characterized by comprising the following modules:

[0059] A No. 1 graph structure data encoding module, used to encode key link information into No. 1 graph structure data; the graph structure data includes nodes and edges connecting nodes, and a node represents an industrial chain link or location advantage or resource information or industrial foundation or human resources or supporting facilities or policy support or development goals;

[0060] A key link acquisition module is used to input the No. 1 graph structure data into a key link model, the key link model includes a first hidden layer and a first output layer, wherein the first hidden layer inputs the No. 1 graph structure data and outputs a first hidden state to the first output layer, and the first output layer outputs a key link of the industrial chain required by the economic development zone;

[0061] The investment clue information collection module is used to collect investment clue information, which includes key links of the industrial chain, enterprise information and economic development zone information;

[0062] A second graph structure data encoding module, used to encode the investment lead information into second graph structure data, the graph structure data includes nodes and edges connecting the nodes, and a node represents a key link of the industrial chain or an enterprise or an economic development zone;

[0063] The investment promotion plan acquisition module is used to input the No. 2 graph structure data into the investment promotion model. The model includes a second hidden layer, a third hidden layer and a second output layer. The second hidden layer inputs the No. 2 graph structure data and outputs the second hidden state to the third hidden layer. The third hidden layer outputs the third hidden state to the second output layer. The second output layer outputs the investment promotion plan.

[0064] The beneficial effects of the present invention are:

[0065] The present invention first determines the key links of the industrial chain required locally, and then determines the investment promotion plan required for each key link through the key links. According to the results, investment promotion work can be carried out more accurately for key links of different industrial chains in different regions. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a flow chart of a method for mining and analyzing investment leads based on a large model intelligent agent of the present invention;

[0067] Figure 2 It is a module schematic diagram of a large-model intelligent agent-based investment lead mining and analysis system of the present invention. DETAILED DESCRIPTION

[0068] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0069] At least one embodiment of the present invention discloses a method for mining and analyzing investment leads based on a large model agent, such as Figure 1 As shown, the following steps are included:

[0070] Step 100, collecting key link information, which includes all links in the industrial chain and information on economic development zones;

[0071] The industry chain links include: battery material supply, motor material supply, power electronic device supply, lightweight material supply, power battery production, drive motor production, power electronic control system production, vehicle design, body manufacturing, vehicle assembly, charging station and charging pile construction, operator service system, battery recycling, material reuse, core software system development, testing and inspection services, logistics and marketing services;

[0072] The information of economic development zones includes: location advantage information, resource information, industrial foundation information, human resources information, supporting facilities information, policy support information, and development goal information;

[0073] Step 200, encoding the key link information into No. 1 graph structure data; the No. 1 graph structure data includes nodes and edges connecting the nodes, and a node represents an industrial chain link or location advantage or resource information or industrial foundation or human resources or supporting facilities or policy support or development goals;

[0074] It indicates that there are edges between the nodes of all industrial chain links, that there are edges between industrial chain links and location advantages when the location advantage brings convenience to the industrial chain links, that there are edges between industrial chain links and resource information when resource information is beneficial to the production and processing of industrial chain links, that there are edges between industrial chain links and industrial base when the existing industrial base and industrial chain links have upstream and downstream relationships or collaborative relationships, that there are edges between industrial chain links and human resources when the quality or quantity of human resources meets the production needs of the industrial chain, that there are edges between industrial chain links and supporting facilities when supporting facilities can support the smooth operation of industrial chain links, that there are edges between industrial chain links and policy support when policy support can attract and promote the development of industrial chain links, and that there are edges between industrial chain links and development goals when development goals attach importance to industrial chain links;

[0075] Indicates that the characteristics of the industrial chain links, location advantages, resource information, industrial base, human resources, supporting facilities, policy support, and development goals are equal to the corresponding node characteristics;

[0076] The characteristics of industrial chain links, location advantages, resource information, industrial base, human resources, supporting facilities, policy support, and development goals are obtained through feature engineering.

[0077] Step 300, inputting the No. 1 graph structure data into a key link model, the key link model comprising a first hidden layer and a first output layer, wherein the first hidden layer inputs the No. 1 graph structure data and outputs a first hidden state to the first output layer, and the first output layer outputs a key link of the industrial chain required by the economic development zone;

[0078] In one embodiment of the present invention, the calculation formula of the first hidden layer is as follows:

[0079] ;

[0080] in Indicates the first picture structure data The first hidden state of a node, It is the node with the structure data of graph No.1 The set of nodes with edges between them, Indicates the first picture structure data The node characteristics of the nodes, is a normalizing constant (such as node degree), represents the weight parameter of the No. 1 graph, represents the bias parameter of Figure 1, It is a S-shaped function.

[0081] In one embodiment of the present invention, the calculation formula of the first output layer is as follows:

[0082] ;

[0083] in Represents a vector, the cth component of the vector represents the probability of the cth key link, Indicates the first picture structure data The first hidden state of a node, Indicates the concatenation of the first hidden states of all nodes. A collection of nodes representing the links in the industrial chain. is the first weight parameter, is the first bias parameter, Represents a sigmoid function.

[0084] In one embodiment of the present invention, the key links of the industrial chain required by the economic development zone output by the key link model are decoded to obtain specific key links before proceeding to subsequent steps.

[0085] Step 400, collecting investment information, which includes key links of the industrial chain, enterprise information and economic development zone information;

[0086] Enterprise information includes: enterprise name, type, scale, establishment time, legal representative, business scope, registered address, registered capital, number of employees, core technology, number of patents, and market share.

[0087] Step 500, encoding the investment lead information into No. 2 graph structure data, the No. 2 graph structure data includes nodes and edges connecting the nodes, and a node represents a key link of the industrial chain or an enterprise or an economic development zone;

[0088] There are edges between the nodes representing all key links. When there are supply chain relationships or competitive relationships between enterprises, there are edges between enterprises. When an enterprise is engaged in a key link, there is an edge between the enterprise and this key link. There are edges between the economic development zone and all nodes representing key links.

[0089] The node characteristics of the node representing the enterprise are equal to the enterprise characteristics, the node characteristics of the node representing the key link are equal to the key link characteristics, and the node characteristics of the node representing the economic development zone are equal to the economic development zone characteristics;

[0090] Enterprise characteristics, key link characteristics and economic development zone characteristics are obtained through feature engineering;

[0091] Encode the investment lead information of all nodes into No. 2 sequence data, the No. 2 sequence data includes n sequence units, and the t-th sequence unit represents the investment lead information of all nodes at the t-th moment;

[0092] Step 600, inputting the No. 2 graph structure data into the investment attraction model, the model includes a second hidden layer, a third hidden layer and a second output layer, wherein the second hidden layer inputs the No. 2 graph structure data, outputs the second hidden state to the third hidden layer, the third hidden layer outputs the third hidden state to the second output layer, and the second output layer outputs the investment attraction plan;

[0093] In one embodiment of the present invention, the key link information and investment lead information are in text mode, and the WordEmbedding algorithm is used as a feature engineering method.

[0094] In one embodiment of the present invention, the calculation formula of the second hidden layer is as follows:

[0095] ;

[0096] in The tth sequence unit of the second sequence data is represented by The second hidden state of the node of nodes, The tth sequence unit of the second sequence data is represented by The aggregation coefficient of nodes, represents the set of nodes that have edges with the vth node of the tth sequence unit of the second sequence data, represents the activation function, represents the graph weight coefficient;

[0097] ;

[0098] ;

[0099] ;

[0100] in and The tth sequence unit of the second sequence data is represented by and The node characteristics of the nodes, and The tth sequence unit of the second sequence data is represented by and The linear transformation characteristics of nodes, represents the aggregation weight coefficient, represents the splicing weight coefficient, T represents transposition, represents an exponential function with a natural constant as the base, LeakyRelu represents a modified linear unit function, Indicates The set of nodes that have edges.

[0101] In one embodiment of the present invention, the calculation formula of the third hidden layer is as follows:

[0102] ;

[0103] ;

[0104] ;

[0105] ;

[0106] in, , , represents the first, second and third sequence weight parameters, , , represents the first, second and third sequence bias parameters, represents the dot product, , and Represent the first, second, and third intermediate states respectively, where , The tth sequence unit of the second sequence data is represented by The second hidden state of the node, Indicates the first t-th sequence unit of the second sequence data The set of nodes with which there are edges. and They represent the t-th and t-1-th third hidden states respectively, n≥t≥1, n represents the total number of sequence units of the second sequence data, and when t=1 , tanh is the hyperbolic tangent function, Represents a sigmoid function.

[0107] In one embodiment of the present invention, the calculation formula of the second output layer is as follows:

[0108] ;

[0109] in Represents a vector, the i-th component of the vector represents the probability value of the i-th action; the set of all actions is called the action space, which contains all possible actions that may be selected. represents the tth third hidden state, represents the concatenation of the 1st to nth third hidden states, n represents the total number of sequence units of the second sequence data, is the second weight parameter, is the second bias parameter, Represents a sigmoid function.

[0110] In one embodiment of the present invention, an action is a business investment plan, and an action is represented as a matrix. The first row and j column of the matrix represent the types of enterprises required for the jth key link, the second row and j column represent the number of enterprises required for the jth key link, the third row and j column represent the scale of the enterprises required for the jth key link, and the fourth row and j column represent the priority of the enterprises required for the jth key link.

[0111] In one embodiment of the present invention, the actions in the action space are to randomly combine the enterprise type, the number of enterprises, the enterprise scale, and the investment promotion priority. The enterprise type and the industrial chain link division are the same. The number of enterprises is an integer value between 1-10. The enterprise scale is divided into three levels: small, medium, and large. The investment promotion priority is divided into three levels: one, two, and three.

[0112] In one embodiment of the present invention, the investment promotion plan output by the investment promotion model is decoded to obtain a specific investment promotion plan, and then the investment promotion plan is implemented by relevant departments of the local government.

[0113] In one embodiment of the present invention, Figure 2 As shown, a large model-based agent-based investment lead mining and analysis system includes the following modules:

[0114] The key link information collection module 101 is used to collect key link information, which includes information on all links in the industrial chain and economic development zones;

[0115] The first graph structure data encoding module 102 is used to encode key link information into the first graph structure data; the graph structure data includes nodes and edges connecting nodes, and a node represents an industrial chain link or location advantage or resource information or industrial foundation or human resources or supporting facilities or policy support or development goals;

[0116] The key link acquisition module 103 is used to input the No. 1 graph structure data into the key link model, the key link model includes a first hidden layer and a first output layer, wherein the first hidden layer inputs the No. 1 graph structure data and outputs the first hidden state to the first output layer, and the first output layer outputs the key link of the industrial chain required by the economic development zone;

[0117] The investment clue information collection module 104 is used to collect investment clue information, which includes key links of the industrial chain, enterprise information and economic development zone information;

[0118] The second graph structure data encoding module 105 is used to encode the investment promotion clue information into the second graph structure data, the graph structure data includes nodes and edges connecting nodes, and a node represents a key link of the industrial chain or an enterprise or an economic development zone;

[0119] The investment promotion plan acquisition module 106 is used to input the No. 2 graph structure data into the investment promotion model, and the model includes a second hidden layer, a third hidden layer and a second output layer, wherein the second hidden layer inputs the No. 2 graph structure data, outputs the second hidden state to the third hidden layer, the third hidden layer outputs the third hidden state to the second output layer, and the second output layer outputs the investment promotion plan.

[0120] In one embodiment of the present invention, both the key link model and the investment promotion model are trained using reinforcement learning, wherein the model to be trained is the key link model or the investment promotion model;

[0121] The steps of training the model to be trained include:

[0122] Step 201, constructing a target network, the target network copies the model to be trained;

[0123] Step 202: randomly obtain an experience from the experience pool. The experience includes the state at time t. , the action performed at time t , the state at time e+1 , perform the action Rewards ;

[0124] The states at time e and e+1 respectively represent the key link information or investment lead information at the e-th and e+1-th time input to the model to be trained;

[0125] Then calculate the state The target network output corresponds to The probability value ;

[0126] ;

[0127] express The maximum probability value of the target network output;

[0128] Step 203, according to and Calculate the loss value loss and update the model to be trained;

[0129] ;

[0130] express The output of the model to be trained corresponds to The probability value of

[0131] is the discount factor, a value between 0 and 1;

[0132] Step 204, updating the target network every fixed number of network updates so that its parameters are the same as those of the current model to be trained;

[0133] Step 205, until the number of collected experiences reaches a set threshold or the model to be trained converges, the step is terminated.

[0134] In one embodiment of the present invention, the experience is derived from the record of historical action execution.

[0135] In one embodiment of the present invention, the reward of the key clue model The calculation formula is as follows:

[0136]

[0137] ;

[0138] in represents the number of key links from time e to e+1, represents the total number of links in the industrial chain, represents the location adaptation score given by experts from time e to e+1, represents the natural resource adaptation score given by experts from time e to e+1, represents the human resource adaptation score given by experts from time e to e+1, represents the policy support adaptation score given by experts from time e to e+1, , , , , They represent the first, second, third, fourth, and fifth weight coefficients respectively.

[0139] In one embodiment of the present invention, the incentives of the investment model The calculation formula is as follows:

[0140]

[0141] ;

[0142] in It represents the new tax revenue of the economic development zone from time e to e+1. It represents the total tax revenue of the economic development zone in the previous year. It represents the newly added employment population in the economic development zone from time e to e+1. represents the total number of employees in the economic development zone in the previous year, It represents the average salary of the investment projects from time e to e+1. represents the average wage in the economic development zone, It represents the capital investment in the economic development zone from time e to time e+1. Indicates the total annual capital investment in the economic development zone, It represents the land area investment of the economic development zone from time e to e+1. It represents the total annual land area invested in the economic development zone. , , , , They represent the sixth, seventh, eighth, ninth and tenth weight coefficients respectively.

[0143] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A method for mining and analyzing investment leads based on a large model agent, characterized in that: The following steps are involved: Step 100, collecting key link information, which includes all links in the industrial chain and information on economic development zones; Step 200, encoding the key link information into No. 1 graph structure data; the graph structure data includes nodes and edges connecting the nodes, and a node represents an industrial chain link or location advantage or resource information or industrial foundation or human resources or supporting facilities or policy support or development goals; It indicates that there are edges between the nodes of all industrial chain links, that there are edges between industrial chain links and location advantages when the location advantage brings convenience to the industrial chain links, that there are edges between industrial chain links and resource information when resource information is beneficial to the production and processing of industrial chain links, that there are edges between industrial chain links and industrial base when the existing industrial base and industrial chain links have upstream and downstream relationships or collaborative relationships, that there are edges between industrial chain links and human resources when the quality or quantity of human resources meets the production needs of the industrial chain, that there are edges between industrial chain links and supporting facilities when supporting facilities can support the smooth operation of industrial chain links, that there are edges between industrial chain links and policy support when policy support can attract and promote the development of industrial chain links, and that there are edges between industrial chain links and development goals when development goals attach importance to industrial chain links; Indicates that the characteristics of the industrial chain links, location advantages, resource information, industrial base, human resources, supporting facilities, policy support, and development goals are equal to the corresponding node characteristics; The characteristics of industrial chain links, location advantages, resource information, industrial base, human resources, supporting facilities, policy support, and development goals are obtained through feature engineering; Step 300, input the No. 1 graph structure data into the key link model, the key link model includes a first hidden layer and a first output layer, wherein the first hidden layer inputs the No. 1 graph structure data and outputs the first hidden state to the first output layer, and the first output layer outputs the key link of the industrial chain required for the economic development zone.

2. According to the method of claim 1, the method is characterized in that: The following steps are also included: Step 400, collecting investment information, which includes key links of the industrial chain, enterprise information and economic development zone information; Step 500, encoding the investment lead information into No. 2 graph structure data, the graph structure data includes nodes and edges connecting the nodes, and a node represents a key link of the industrial chain or an enterprise or an economic development zone; There are edges between the nodes representing all key links. When there are supply chain relationships or competitive relationships between enterprises, there are edges between enterprises. When an enterprise is engaged in a key link, there is an edge between the enterprise and this key link. There are edges between the economic development zone and all nodes representing key links. The node characteristics of the node representing the enterprise are equal to the enterprise characteristics, the node characteristics of the node representing the key link are equal to the key link characteristics, and the node characteristics of the node representing the economic development zone are equal to the economic development zone characteristics; Enterprise characteristics, key link characteristics and economic development zone characteristics are obtained through feature engineering; Encode the investment lead information of all nodes into sequence data, the sequence data includes n sequence units, and the t-th sequence unit represents the investment lead information of all nodes at the t-th moment; Step 600, input the No. 2 graph structure data into the investment promotion model, the model includes a second hidden layer, a third hidden layer and a second output layer, wherein the second hidden layer inputs the No. 2 graph structure data, outputs the second hidden state to the third hidden layer, the third hidden layer outputs the third hidden state to the second output layer, and the second output layer outputs the investment promotion plan.

3. The method for mining and analyzing investment leads based on a large model agent according to claim 1, characterized in that: The industry chain links include: battery material supply, motor material supply, power electronic device supply, lightweight material supply, power battery production, drive motor production, power electronic control system production, vehicle design, body manufacturing, vehicle assembly, charging station and charging pile construction, operator service system, battery recycling, material reuse, core software system development, testing and inspection services, logistics and marketing services; The information of economic development zones includes: location advantage information, resource information, industrial foundation information, human resources information, supporting facilities information, policy support information, and development goal information; Enterprise information includes: enterprise name, type, scale, establishment time, legal representative, business scope, registered address, registered capital, number of employees, core technology, number of patents, and market share.

4. The method for mining and analyzing investment leads based on a large model agent according to claim 1, characterized in that: The calculation formula for the first hidden layer is as follows: ； in Indicates the first picture structure data The first hidden state of a node, It is the node with the structure data of graph No.1 The set of nodes with edges between them, Indicates the first picture structure data The node characteristics of the nodes, is a normalizing constant (such as node degree), represents the weight parameter of the No. 1 graph, represents the bias parameter of Figure 1, It is a S-shaped function.

5. The method for mining and analyzing investment leads based on a large model agent according to claim 1, characterized in that: The calculation formula of the first output layer is as follows: ； in Represents a vector, the cth component of the vector represents the probability of the cth key link, Indicates the first picture structure data The first hidden state of a node, Indicates the concatenation of the first hidden states of all nodes. A collection of nodes representing the links in the industrial chain. is the first weight parameter, is the first bias parameter, Represents a sigmoid function.

6. The method for mining and analyzing investment leads based on a large model agent according to claim 1, characterized in that: The calculation formula for the second hidden layer is as follows: ； in The tth sequence unit of the second sequence data is represented by The second hidden state of the node of nodes, The tth sequence unit of the second sequence data is represented by The aggregation coefficient of nodes, represents the set of nodes that have edges with the vth node of the tth sequence unit of the second sequence data, represents the activation function, represents the graph weight coefficient; ； ； ； in and The tth sequence unit of the second sequence data is represented by and The node characteristics of the nodes, and The tth sequence unit of the second sequence data is represented by and The linear transformation characteristics of nodes, represents the aggregation weight coefficient, represents the splicing weight coefficient, T represents transposition, represents an exponential function with a natural constant as the base, LeakyRelu represents a modified linear unit function, Indicates The set of nodes with which there are edges.

7. The method for mining and analyzing investment leads based on a large model agent according to claim 1, characterized in that: The calculation formula for the third hidden layer is as follows: ； ； ； ； in, , , represents the first, second and third sequence weight parameters, , , represents the first, second and third sequence bias parameters, represents the dot product, , and Represent the first, second, and third intermediate states respectively, where , The tth sequence unit of the second sequence data is represented by The second hidden state of the node, Indicates the first t-th sequence unit of the second sequence data The set of nodes with which there are edges. and They represent the t-th and t-1-th third hidden states respectively, n≥t≥1, n represents the total number of sequence units of the second sequence data, and when t=1 , tanh is the hyperbolic tangent function, Represents a sigmoid function.

8. The method for mining and analyzing investment leads based on a large model agent according to claim 1, characterized in that: The calculation formula of the second output layer is as follows: ； in Represents a vector, the i-th component of the vector represents the probability value of the i-th action; the set of all actions is called the action space, which contains all possible actions that may be selected. represents the tth third hidden state, represents the concatenation of the 1st to nth third hidden states, n represents the total number of sequence units of the second sequence data, is the second weight parameter, is the second bias parameter, Represents a sigmoid function.

9. The method for mining and analyzing investment leads based on a large model agent according to claim 1, characterized in that: Both the key link model and the investment promotion model are trained using reinforcement learning, where the model to be trained is the key link model or the investment promotion model; The steps of training the model to be trained include: Step 201, constructing a target network, the target network copies the model to be trained; Step 202: randomly obtain an experience from the experience pool. The experience includes the state at time t. , the action performed at time t , the state at time e+1 , perform the action Rewards ; The states at time e and e+1 respectively represent the key link information or investment lead information at the e-th and e+1-th time input to the model to be trained; Then calculate the state The target network output corresponds to The probability value ; ； express The maximum probability value of the target network output; Step 203, according to and Calculate the loss value loss and update the model to be trained; ； express The output of the model to be trained corresponds to The probability value of is the discount factor, a value between 0 and 1; Step 204, updating the target network every fixed number of network updates so that its parameters are the same as those of the current model to be trained; Step 205, until the number of collected experiences reaches a set threshold or the model to be trained converges, the step is terminated.

10. A system for mining and analyzing investment leads based on a large model agent, characterized in that: Includes the following modules: A No. 1 graph structure data encoding module, used to encode key link information into No. 1 graph structure data; the graph structure data includes nodes and edges connecting nodes, and a node represents an industrial chain link or location advantage or resource information or industrial foundation or human resources or supporting facilities or policy support or development goals; A key link acquisition module is used to input the No. 1 graph structure data into a key link model, the key link model includes a first hidden layer and a first output layer, wherein the first hidden layer inputs the No. 1 graph structure data and outputs a first hidden state to the first output layer, and the first output layer outputs a key link of the industrial chain required by the economic development zone; The investment clue information collection module is used to collect investment clue information, which includes key links of the industrial chain, enterprise information and economic development zone information; A second graph structure data encoding module, used to encode the investment lead information into second graph structure data, the graph structure data includes nodes and edges connecting the nodes, and a node represents a key link of the industrial chain or an enterprise or an economic development zone; The investment promotion plan acquisition module is used to input the No. 2 graph structure data into the investment promotion model. The model includes a second hidden layer, a third hidden layer and a second output layer. The second hidden layer inputs the No. 2 graph structure data and outputs the second hidden state to the third hidden layer. The third hidden layer outputs the third hidden state to the second output layer. The second output layer outputs the investment promotion plan.

Images