CN117010929B

CN117010929B - Agricultural product public opinion information construction method

Info

Publication number: CN117010929B
Application number: CN202310920732.1A
Authority: CN
Inventors: 孙彤; 黄桂恒
Original assignee: Brick Suzhou Agricultural Internet Co ltd
Current assignee: Brick Suzhou Agricultural Internet Co ltd
Priority date: 2022-05-30
Filing date: 2022-05-30
Publication date: 2024-04-26
Anticipated expiration: 2042-05-30
Also published as: CN115082129A; CN117010929A; CN114881706A; CN114936884A; CN115082129B

Abstract

The embodiment of the specification provides a method for constructing agricultural product public opinion information, which comprises the steps of obtaining a propagation path of a predicted price; constructing a knowledge graph based on the propagation path; and performing influence calculation on the knowledge graph to generate a public opinion matrix.

Description

Agricultural product public opinion information construction method

Description of the division

The application relates to a Chinese application with the filing number 202210596614.5 and the name of "a method and a system for automatically adjusting agricultural product quotation", which is proposed for the Chinese application with the filing number 2022, 05 and 30, 202210830227.3 and the name of "a method and a system for automatically quoting agricultural products".

Technical Field

The specification relates to the field of agricultural product data processing, in particular to a construction method of agricultural product public opinion information.

Background

In the present informatization era, big data penetrate the aspects of people's society life. By properly using the data, people can become more efficient and quick to handle daily transactions. And the method is also beneficial to predicting future data of things and making related plans based on the existing mass data. The prior art (grant bulletin number CN 102982229B) discloses a data preprocessing method for predicting the prices of multiple varieties of commodities based on a neural network, an improved RBF neural network and a BP neural network are utilized to calculate the optimal order of magnitude of commodity price data mined by a webpage, and the calculated optimal order of magnitude is utilized to perform preprocessing of normalized data orders of the commodity price data, so that the prediction accuracy of the RBF neural network and the BP neural network is improved, and the universality of the RBF neural network and the BP neural network for predicting the prices of different varieties of commodities is also improved. The prior art (grant bulletin number CN 112651832B) provides an artificial intelligence futures price prediction system and prediction method based on blockchain. The method aims at helping the common investors to predict the futures price from the technical level. The prior art (authorized bulletin number CN 109829742B) provides a thermal power generating unit optimal quotation calculation method based on declaration price settlement, and the technical scheme provided by the method can solve the bidding problem of thermal power enterprises participating in peak shaving auxiliary service markets, improve the participation level of the thermal power generating unit in the markets, improve the running efficiency of the markets and increase the income of the thermal power enterprises. The prior art (authorized bulletin number CN 109961314B) discloses a system and a method for evaluating and quoting jewelry materials based on big data, which can integrate global jewelry material transaction information by the system, generate uniform standards and transaction information, regulate product prices according to market feedback, improve market acceptance of all products, promote transaction of both buyers and sellers, and enable middle-or small-sized sellers or individuals to sell or buy at a better and more reasonable price. However, these prior art techniques are not suitable for automatic quotation of agricultural products and all suffer from certain limitations. The quotation of the agricultural products directly affects the production and circulation of the agricultural products, adjusts the quotation of the agricultural products, and has important practical significance for avoiding the market risk of the agricultural products and promoting the adjustment and sustainable development of the agricultural structure.

Therefore, it is desirable to provide a method for constructing public opinion information of agricultural products, which can more accurately automatically offer the agricultural products, thereby promoting the development of the agricultural product market.

Disclosure of Invention

One of the embodiments of the present disclosure provides a method for constructing public opinion information of agricultural products. The method for constructing the public opinion information of the agricultural products comprises the following steps: acquiring a propagation path of a predicted price; constructing a knowledge graph based on the propagation path; and performing influence calculation on the knowledge graph to generate a public opinion matrix.

One of the embodiments of the present specification provides an agricultural product public opinion information construction system including: the acquisition module is used for acquiring a propagation path of the predicted price; constructing a knowledge graph based on the propagation path; and performing influence calculation on the knowledge graph to generate a public opinion matrix.

One of the embodiments of the present specification provides an agricultural product public opinion information construction apparatus, the apparatus including: at least one storage medium storing computer instructions; and the at least one processor executes the computer instructions to realize the agricultural product public opinion information construction method.

One of the embodiments of the present specification provides a computer-readable storage medium storing computer instructions that when read by a computer, the computer performs a method of constructing public opinion information of agricultural products.

Drawings

The present specification will be further elucidated by way of example embodiments, which will be described in detail by means of the accompanying drawings. The embodiments are not limiting, in which like numerals represent like structures, wherein:

FIG. 1 is a schematic illustration of an application scenario of a system for automatically adjusting agricultural product offers according to some embodiments of the present description;

FIG. 2 is an exemplary block diagram of a system for automatically adjusting agricultural product price quotes in accordance with some embodiments of the present description;

FIG. 3 is an exemplary flow chart of a method of automatically adjusting agricultural product offers according to some embodiments of the present description;

FIG. 4 is an exemplary schematic diagram of constructing an impact graph based on impact factors, shown in accordance with some embodiments of the present description;

FIG. 5 is an exemplary schematic diagram of a price fluctuation prediction model for agricultural products according to some embodiments of the present description;

FIG. 6 is an exemplary flow chart for determining price-affecting parameters of agricultural products according to some embodiments of the present description;

FIG. 7 is an exemplary schematic diagram of a model structure for dynamically adjusting a price fluctuation prediction model for agricultural products, according to some embodiments of the present description;

Fig. 8 is an exemplary flow chart of a method of automatically quoting agricultural products according to some embodiments of the present description.

Detailed Description

In order to more clearly illustrate the technical solutions of the embodiments of the present specification, the drawings that are required to be used in the description of the embodiments will be briefly described below. It is apparent that the drawings in the following description are only some examples or embodiments of the present specification, and it is possible for those of ordinary skill in the art to apply the present specification to other similar situations according to the drawings without inventive effort. Unless otherwise apparent from the context of the language or otherwise specified, like reference numerals in the figures refer to like structures or operations.

It will be appreciated that "system," "apparatus," "unit" and/or "module" as used herein is one method for distinguishing between different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

As used in this specification and the claims, the terms "a," "an," "the," and/or "the" are not specific to a singular, but may include a plurality, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.

A flowchart is used in this specification to describe the operations performed by the system according to embodiments of the present specification. It should be appreciated that the preceding or following operations are not necessarily performed in order precisely. Rather, the steps may be processed in reverse order or simultaneously. Also, other operations may be added to or removed from these processes.

Fig. 1 is a schematic illustration of an application scenario of a system for automatically adjusting agricultural product offers according to some embodiments of the present description. As shown in fig. 1, an application scenario 100 of a system for automatically adjusting agricultural product offers may include one or more of a processing device 110, a network 120, a storage device 130, a terminal device 140, and an agricultural product price 150, among others. The system for automatically adjusting the price quote for the agricultural product may include a system for determining price affecting parameters for the agricultural product, a system for automatically quoting the agricultural product, and the like.

In some embodiments, processing device 110 may process information and/or data related to application scenario 100 of a system for automatically adjusting agricultural product offers to perform one or more of the functions described in this specification. For example, processing device 110 may obtain an impact factor fluctuation of the impact factor, predict a price fluctuation of the agricultural product by the price fluctuation prediction model of the agricultural product based on the impact factor fluctuation, and determine an adjustment to the quote based on the price fluctuation of the agricultural product. For another example, the processing device 110 may obtain the impact factors, construct an impact graph based on the impact factors as nodes, and based on the relationships of the impact factors as directed edges; processing device 110 may determine the agricultural product price impact parameters by an impact map. For another example, the processing device 110 may obtain bid information, deal vectors, public opinion information, and the like. Processing device 110 may predict a price of the agricultural product through the agricultural product price prediction model based on at least one of price quotation information, a deal vector, public opinion information, and the like, and determine a price quotation for the agricultural product based on the predicted price of the agricultural product. In some embodiments, processing device 110 may include one or more processing engines (e.g., a single chip processing engine or a multi-chip processing engine). By way of example only, the processing device 110 may include a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), an Application Specific Instruction Processor (ASIP), a Graphics Processor (GPU), a Physical Processor (PPU), a Digital Signal Processor (DSP), a Field Programmable Gate Array (FPGA), an editable logic circuit (PLD), a controller, a microcontroller unit, a Reduced Instruction Set Computer (RISC), a microprocessor, or the like, or any combination thereof.

The network 120 may connect various components of the system and/or connect the system with external resource components. Network 120 allows communication between the various components, as well as with other components outside the system. For example, the processing device 110 may obtain the impact factor, impact factor fluctuations of the impact factor, bid information, deal vectors, public opinion information, and the like through the network 120. For another example, the processing device 110 may obtain, via the network 120, a plurality of data stored in the storage device 130, such as, for example, values of influence factors, quotation information, deal vectors, public opinion information, etc. at some historical point in time. In some embodiments, network 120 may be any one or more of a wired network or a wireless network. For example, the network 120 may include a cable network, a fiber optic network, a telecommunications network, the internet, a Local Area Network (LAN), a Wide Area Network (WAN), a Wireless Local Area Network (WLAN), a Metropolitan Area Network (MAN), a Public Switched Telephone Network (PSTN), a bluetooth network, a ZigBee network, a Near Field Communication (NFC), an intra-device bus, an intra-device line, a cable connection, and the like, or any combination thereof. The network connection between the parts can be in one of the above-mentioned ways or in a plurality of ways. In some embodiments, the network may be a point-to-point, shared, centralized, etc. variety of topologies or a combination of topologies.

Storage device 130 may be used to store data and/or instructions related to application scenario 100 of a system for automatically adjusting agricultural product offers. In some embodiments, storage device 130 may store data and/or information obtained from processing device 110, network 120, and the like. For example, the storage device 130 may store influence factors, influence factor fluctuations of influence factors, quotation information, deal vectors, public opinion information, and the like. In some embodiments, storage device 130 may include one or more storage components, each of which may be a separate device or may be part of another device. In some embodiments, the storage device 130 may be disposed in the processing device 110. In some embodiments, the storage device 130 may include Random Access Memory (RAM), read Only Memory (ROM), mass storage, removable memory, volatile read-write memory, and the like, or any combination thereof. By way of example, mass storage may include magnetic disks, optical disks, solid state disks, and the like. In some embodiments, storage device 130 may be implemented on a cloud platform.

Terminal device 140 may refer to one or more terminal devices or software used by a user. The user may be an individual or a collective related to the agricultural product, for example, a buyer or seller of the agricultural product. The user may include one or more of an agricultural product producer, an agricultural product provider, an agricultural product distributor, an agricultural product retailer, and the like. In some embodiments, the terminal device 140 may include a mobile device 140-1, a tablet computer 140-2, a notebook computer 140-3, a laptop computer 140-4, or the like, or any combination thereof. In some embodiments, the terminal device 140 may include other smart terminals, such as wearable smart terminals and the like. The above examples are only intended to illustrate the broad scope of the terminal device and not to limit its scope.

The price 150 of the agricultural product varies at different times. A variety of factors (e.g., impact factors, etc.) may impact the price of the agricultural product 150. Fluctuations in a variety of factors (e.g., influence factor fluctuations) may influence agricultural product price fluctuations. Processing device 110 may analyze the relevant data for price of agricultural product 150, predict price of agricultural product, and thus predict price fluctuations of agricultural product, determine price-affecting parameters of agricultural product, determine adjustments to the price quote, and determine price quote for agricultural product.

It should be noted that the application scenario 100 of the system for automatically adjusting agricultural product offers is provided for illustrative purposes only and is not intended to limit the scope of the present description. Many modifications and variations will be apparent to those of ordinary skill in the art in light of the present description. For example, application scenario 100 of a system for automatically adjusting agricultural product offers may implement similar or different functionality on other devices. However, such changes and modifications do not depart from the scope of the present specification.

FIG. 2 is an exemplary block diagram of a system for automatically adjusting agricultural product offers according to some embodiments of the present description.

In some embodiments, the system 200 for automatically adjusting agricultural product offers may include an acquisition module 210 and a prediction module 220. In some embodiments, the prediction module 220 may include a determination module 221 and a dynamic adjustment module 222.

In some embodiments, the acquisition module 210 may be configured to acquire an impact factor fluctuation of the impact factor.

In some embodiments, prediction module 220 may be configured to predict the agricultural product price volatility based on the impact factor volatility by an agricultural product price volatility prediction model that is a machine learning model that includes at least one neural network layer.

In some embodiments, determination module 221 may be configured to determine an adjustment to the offer based on the fluctuation in the price of the agricultural product.

In some embodiments, the prediction module 220 may be further configured to construct an impact graph based on the impact factors as nodes, based on the relationships of the impact factors as directed edges, and based on the impact graph, predict the price fluctuation of the agricultural product through the price fluctuation prediction model of the agricultural product.

In some embodiments, the agricultural product price fluctuation prediction model includes a plurality of neural networks, each of the plurality of neural networks corresponding to a node of the impact graph, the connection relationship between each of the neural networks being determined based on the directed edges of the impact graph.

In some embodiments, dynamic adjustment module 222 may be used to obtain a predictive effect of the agricultural product price fluctuation prediction model; based on the prediction effect, the model structure of the agricultural product price fluctuation prediction model is dynamically adjusted.

In some embodiments, the acquisition module 210 may be configured to acquire the impact factor.

In some embodiments, determination module 221 may be configured to determine the agricultural product price impact parameter based on the impact map.

In some embodiments, determination module 221 may also be configured to construct a price fluctuation prediction model for agricultural products based on the impact map; training a price fluctuation prediction model of the agricultural product based on the historical data; and determining the price influence parameters of the agricultural products based on the model parameters of the price fluctuation prediction model of the agricultural products.

In some embodiments, edge features that affect directed edges of the graph include trustworthiness.

In some embodiments, the acquisition module 210 may be configured to acquire quotation information; obtaining an intersection vector; and obtaining public opinion information.

In some embodiments, the prediction module 220 may be configured to predict the price of the agricultural product based on at least one of price quote information, a deal vector, and public opinion information via an agricultural product price prediction model, wherein the agricultural product price prediction model is a machine learning model.

In some embodiments, determination module 221 may be configured to determine a price quote for the agricultural product based on the predicted price of the agricultural product.

In some embodiments, the offer information includes agricultural product provider offer information and a credibility factor, and the acquisition module 210 may be further configured to acquire the agricultural product provider offer information; acquiring a credibility factor; based on the commodity provider pricing information and the credibility factor, a pricing matrix of the pricing information is constructed.

In some embodiments, the public opinion information includes a public opinion matrix constructed based on different degrees of influence, the degrees of influence determined based on a propagation relationship of the predicted price.

It should be understood that the system shown in fig. 2 and its modules may be implemented in a variety of ways.

It should be noted that the above description of the system and its modules is for convenience of description only and is not intended to limit the present description to the scope of the illustrated embodiments. It will be appreciated by those skilled in the art that, given the principles of the system, various modules may be combined arbitrarily or a subsystem may be constructed in connection with other modules without departing from such principles. In some embodiments, the acquisition module 210, the prediction module 220, the determination module 221, and the dynamic adjustment module 222 disclosed in fig. 2 may be different modules in one system, or may be one module to implement the functions of two or more modules described above. For example, each module may share one memory module, or each module may have a respective memory module. Such variations are within the scope of the present description.

FIG. 3 is an exemplary flow chart of a method of automatically adjusting agricultural product offers according to some embodiments of the present description. As shown in fig. 3, the process 300 includes the following steps. In some embodiments, the process 300 may be performed by the processing device 110.

In step 310, influence factor fluctuations of the influence factor are obtained. In some embodiments, step 310 may be performed by the acquisition module 210.

The influencing factor may refer to a factor that influences price fluctuations of the agricultural product. For example, the impact factors may include weather, plant area, fuel price, labor costs, and the like.

In some embodiments, the impact factor includes public opinion information. For more content on public opinion information please see the relevant description of fig. 8.

The influence factor fluctuation may refer to a change in a factor that affects the price fluctuation of the agricultural product. The impact factor fluctuations may include changes in weather, plant area, fuel price, labor costs, etc. The fluctuation of the influence factors may cause the price of agricultural products to fluctuate. For example, weather in the influence factors fluctuates, and the change from ' more sunny days ' to ' rainy days ' and less sunny days ' can influence sweetness of agricultural products, so that price fluctuation (such as price reduction and the like) of the agricultural products is influenced; for another example, the planting area in the influencing factors fluctuates, the planting area is reduced by 10%, the planting area is reduced, and agricultural products are not supplied enough, so that the price fluctuation (such as price rising and the like) of the agricultural products is influenced. For another example, the fuel price in the influencing factors fluctuates, the fuel price increases by 5%, the transportation cost of agricultural products increases, and further the price fluctuation (such as price rising) of the agricultural products is influenced.

In some embodiments, the acquisition module 210 may acquire the influence factor fluctuations of the influence factor in a variety of ways. In some embodiments, the acquisition module 210 may acquire the impact factor fluctuations of the impact factor over the network 120. For example, the obtaining module 210 obtains values of the influence factors at different time points through the network 120, and further obtains fluctuations of the influence factors over a period of time. In some embodiments, the acquisition module 210 may acquire the impact factor fluctuations of the impact factor through the network 120 and a storage device (e.g., the storage device 130). For example, the obtaining module 210 obtains a value of an influence factor of a current time point through a network. The obtaining module 210 may obtain the value of the influence factor at the historical time point from a storage device (such as the storage device 130), and further obtain the influence factor fluctuation.

Step 320, predicting the price fluctuation of the agricultural product by a price fluctuation prediction model of the agricultural product based on the influence factor fluctuation, the price fluctuation prediction model of the agricultural product being a machine learning model comprising at least one neural network layer. In some embodiments, step 330 may be performed by prediction module 220.

The fluctuation of the price of the agricultural product refers to the rising or falling of the price of the agricultural product, etc. In some embodiments, the agricultural product price fluctuations may be represented by a rise or fall in the agricultural product price, e.g., the agricultural product price fluctuations may be represented by a percentage and/or number representing the rise or fall. By way of example, a percentage +10% indicates that the price of the agricultural product fluctuates by 10%; the number +10 indicates that the price fluctuation of the agricultural products is increased by 10 yuan; the percentage of-15% indicates that the price fluctuation of the agricultural products is 15% of the drop; numeral-20 indicates that the price fluctuation of the agricultural product is 20 yuan.

The agricultural product price prediction model refers to a model that can predict the price of agricultural products. In some embodiments, the agricultural product price fluctuation prediction model may be a machine learning model including at least one neural network layer. The selection of the model type may be case-specific. For more on the agricultural product price prediction model, see the relevant description of fig. 5.

In some embodiments, prediction module 220 may predict the agricultural product price volatility by an agricultural product price volatility prediction model based on the impact factor volatility.

In some embodiments, the prediction module 220 may construct an influence graph based on the influence factors as nodes and the relationships of the influence factors as directed edges. Prediction module 220 may predict the price fluctuation of the agricultural product by a price fluctuation prediction model of the agricultural product based on the impact graph. Regarding predicting price fluctuation of agricultural products by the price fluctuation prediction model of agricultural products based on the influence map, reference is made to the related description of fig. 4 for details.

Step 330, an adjustment to the offer is determined based on the agricultural product price fluctuations. In some embodiments, step 330 may be performed by determination module 221.

Adjustment of the offer may refer to an adjustment in the offer to the buyer by the seller of the agricultural product. Adjustments to the bid may include increasing the bid, decreasing the bid, maintaining the bid.

In some embodiments, the determination module 221 may determine an adjustment to the offer based on the fluctuation in the price of the agricultural product. For example, the agricultural product price fluctuation is 10% rise, and the determination module 221 may determine that the adjustment to the price quote is 10% rise in the price quote. For another example, the agricultural product price fluctuation is 15% drop, and the determination module 221 may determine that the adjustment to the offer is 15% drop in the offer.

In some embodiments of the present disclosure, based on the fluctuation of the influencing factor, the price fluctuation of the agricultural product is predicted by the agricultural product price fluctuation prediction model, so that the adjustment of the price quotation is determined, the price fluctuation of the agricultural product can be predicted more accurately, and the price quotation of the agricultural product can be adjusted better.

FIG. 4 is an exemplary schematic diagram illustrating constructing an impact graph based on impact factors according to some embodiments of the present description. In some embodiments, the process 400 may be performed by the prediction module 220.

In step 410, an influence graph is constructed based on the influence factors as nodes and the relation of the influence factors as directed edges.

In some embodiments, the prediction module 220 may construct a graph through nodes and edges. The number of nodes and the number of edges may be two or more. In some embodiments, the prediction module 220 may determine the impact factor as a node. For example, the prediction module 220 determines an impact factor (e.g., weather, plant area, fuel price, labor cost, public opinion information, etc.) as an impact factor node. In some embodiments, prediction module 220 may determine the agricultural product price as an agricultural product price node.

Different nodes have different node characteristics. Node characteristics refer to information describing a node. In some embodiments, when a node is an influence factor node, the node characteristics may include historical fluctuations of the node. The historical fluctuation of the node refers to fluctuation of factors which influence the fluctuation of the price of the agricultural product and are represented by the node in a preset time period, wherein the preset time period refers to a time interval with a certain time length. In some embodiments, the preset time period may be set according to human experience, or may be set by default for the system. The preset time period can be adjusted according to actual needs.

In some embodiments, when the node is an agricultural price node, the node characteristics may include historical price fluctuations for the node. The historical price fluctuation of the node refers to fluctuation of the price of the agricultural product represented by the node in a preset time period, wherein the representation mode of the preset time period is similar to that of the influence factor node.

In some embodiments, the type in the influence graph is an influence factor node of public opinion information, and the node characteristic may include at least one of influence degree, farmer browsing amount and farmer type vector. For more details on node characteristics of influence factor nodes of the influence graph, the type of which is public opinion information, see the relevant description of fig. 8.

In some embodiments, the edges between nodes are directed edges. Directed edges refer to causal links between nodes, and the direction of the directed edge may represent a causal direction. For example, if the price fluctuation causes labor cost fluctuation, the directed edge may be represented as pointing from the price to the labor.

In some embodiments, when a node is an influence factor node, the prediction module 220 may determine whether there is a directed edge between the influence factor nodes based on whether there is a causal relationship between the influence factor nodes, where the causal relationship may include an association between historical fluctuations corresponding to the influence factor nodes.

In some embodiments, when the node is an agricultural product price node, the prediction module 220 may determine a connection between the agricultural product price node and the impact factor node through a directed edge based on a causal relationship between the impact factor node and the agricultural product price, and the connection between the impact factor node and the agricultural product price node through the directed edge may include a direct connection and an indirect connection. Wherein the causal link may include a degree of influence of the influence factor node on the price node of the agricultural product. The degree of influence may be represented by a distance between the influence factor node and the agricultural product price node, which may be a number of hops between the influence factor node and the agricultural product price node. The greater the impact of the impact factor node on the price node of the agricultural product, the smaller the distance between the two, i.e., the fewer hops between the impact factor node and the price node of the agricultural product. For example, 1 hop is a direct connection between an impact factor node and an agricultural product price node; for example, a hop greater than 1 is an indirect connection between an impact factor node and an agricultural product price node.

In some embodiments, edge features affecting directed edges of the graph may include trustworthiness 721.

The confidence 721 may represent the strength of causal association of two connected influence factor nodes corresponding to the directed edge. For example, as in the influence graph of FIG. 4, the trustworthiness of the directed edge 4121 may represent the strength of causal links between the influence factor node 412 and the influence factor node 1. The greater the reliability of the directed edge, the greater the strength of causal association of two connected influence factor nodes.

In some embodiments, the reliability may include a correlation coefficient between the historical data of the two influence factor nodes corresponding to the directed edge, a frequency of occurrence of a certain rule in the historical data of the two influence factor nodes corresponding to the directed edge, or a result of weighted calculation of the correlation coefficient between the historical data of the two influence factor nodes corresponding to the directed edge and the frequency of occurrence of the certain rule, and the like.

In some embodiments, the prediction module 220 may determine the trustworthiness of the directed edge correspondence by historical data between two impact factor nodes of the directed edge correspondence. For example, the prediction module 220 may determine causal links between two influence factor nodes by correlation between historical data of the two influence factor nodes. The causal association between two influence factor nodes may include a positive correlation or a negative correlation, etc. The prediction module 220 may represent the confidence level of the directed edge by a statistically relevant coefficient. For another example, the prediction module 220 may analyze the historical data of the two influence factor nodes corresponding to the directed edge, and determine the change rule of the two influence factor nodes, so as to determine the reliability corresponding to the directed edge. Illustratively, as shown in FIG. 4, each time the influence factor node 412 fluctuates by a fixed value, the fluctuation of the influence factor node 1 is also at or near the fixed value. While influence factor node 412 has no corresponding ripple each time influence factor node 1 fluctuates. The prediction module 220 may determine that there is a causal relationship between the influence factor node 412 and the influence factor node 1, with the directed edge 4121 pointing to the influence factor node 1 for the influence factor node 412. The prediction module 220 may represent the confidence level of the directed edge 4121 between the influence factor node 412 and the influence factor node 1 by the frequency of occurrence of the law of fluctuation described above. For another example, the degree of reliability of a directed edge between two influencing factor nodes may also be determined based on a weighted calculation of the correlation coefficient and the frequency of occurrence of the law of fluctuation described above. In some embodiments, when two or more influence factor nodes collectively influence one influence factor node, the prediction module 220 may determine the trustworthiness of the directed edges associated with the two or more influence factor nodes by the method described above.

In some embodiments, the prediction module 220 may construct the impact graph based on the impact factor nodes, the agricultural product nodes, causal associations between the impact factor nodes, and causal associations between the impact factor nodes and the agricultural product nodes.

In some embodiments, nodes in the influence graph may include root nodes and non-root nodes. A root node refers to a node where there are no edges pointing to the node from other nodes, and a non-root node refers to a node that is not a root node. As shown in FIG. 4, the influence factor nodes 1, 2, N-1, etc. may represent non-root influence factor nodes 1, 2, N-1, etc. The impact factor node N may represent an agricultural product price node N. The influence factor nodes 411, 412, 413, 414, m-1, m, etc. may represent root influence factor nodes 411, 412, 413, 414, m-1, m, etc.

In some embodiments, each time the root influence factor node 411 and the root influence factor node 412 fluctuate by a certain value, the fluctuation generated by the non-root influence factor node 1 each time is relatively close to the value, otherwise, the non-root influence factor node 1 is not established, and causal association exists between the non-root influence factor node 1 and the root influence factor node 411 and the root influence factor node 412. The prediction module 220 may determine that a directed edge 4111 exists between the non-root influencing factor node 1 and the root influencing factor node 411, and that a directed edge 4121 exists between the non-root influencing factor node 1 and the root influencing factor node 412.

In some embodiments, the prediction module 220 may determine that a directed edge 11 exists between the non-root factor node 2 and the non-root factor node 1, a directed edge 4131 exists between the non-root factor node 2 and the root factor node 413, and a directed edge 4141 exists between the non-root factor node 2 and the root factor node according to the methods described above. Similarly, there is a directed edge m-11 between non-root-influencing factor node N-1 and root-influencing factor node m-1, and there is a directed edge N-21 between non-root-influencing factor node N-1 and last hop non-root-influencing factor node N-2.

In some embodiments, for agricultural product price node N, fluctuations of non-root influencing factor node N-1 and root influencing factor node m may directly influence the agricultural product price, and thus non-root influencing factor node N-1 and root influencing factor node m may be directly connected to agricultural product price node N, respectively. The prediction module 220 may determine that a directed edge N-11 exists between the non-root impact factor node N-1 and the agricultural product price node N, and that a directed edge m1 exists between the root impact factor node m and the agricultural product price node N.

It should be noted that the above description with respect to fig. 4 is by way of example only and not by way of limitation. The number of nodes and directed edges in the effect graph may not be limited as shown in fig. 4. In some embodiments, the number of factors affecting the price fluctuation of the agricultural product may determine the number of impact factor nodes in the impact graph.

Step 420, based on the impact graph, predicting the price fluctuation of the agricultural product by the price fluctuation prediction model of the agricultural product.

The model refers to a predictive model that can be directly applied to the graph, for example, a machine learning model that includes at least one neural network layer. In some embodiments, the model may obtain a prediction of a node based on the state of each node in the graph. For example, the model may be an agricultural product price fluctuation prediction model 421 based on an influence map. The agricultural product price fluctuation prediction model 421 may predict agricultural product price fluctuation 422 based on historical fluctuation of the influence factor nodes in the influence graph. In some embodiments, the prediction module 220 may construct the agricultural product price fluctuation prediction model 421 from a machine learning model based on historical fluctuations of the influence factor nodes in the influence graph. For example, the prediction module 220 may establish a training sample based on node characteristics of the influence factor nodes in the influence graph, input the training sample into the machine learning model, and adjust model parameters of the machine learning model based on the model output and the influence factor node history fluctuation, to obtain a trained agricultural product price fluctuation prediction model. For more on the agricultural product price fluctuation prediction model and its training method, see fig. 5 and its related description.

Price fluctuations of agricultural products are affected by a plurality of factors, and at the same time, causal relations exist among different influencing factors. In some embodiments of the present disclosure, an influence graph is constructed based on factors affecting price fluctuation of agricultural products as influence factor nodes, and causal relations among the factors affecting price fluctuation of the agricultural products as directed edges among the influence factor nodes, and price fluctuation of the agricultural products is predicted by an agricultural product price fluctuation prediction model based on the influence graph, so that a rule of price fluctuation of the agricultural products can be obtained more accurately, and further, quotation of the agricultural products can be adjusted better.

FIG. 5 is an exemplary schematic diagram of a price fluctuation prediction model for agricultural products according to some embodiments of the present description. In some embodiments, the process 500 may be performed by the prediction module 220.

The agricultural product price fluctuation prediction model may include a plurality of neural networks. Wherein each of the plurality of neural networks corresponds to a node in the influence graph, and the connection relationship between each of the neural networks may be determined based on the directed edges of the influence graph.

In some embodiments, prediction module 220 may predict agricultural product price fluctuations 422 through agricultural product price fluctuation prediction model 421. The input data of the agricultural product price fluctuation prediction model 421 is a relationship between a plurality of factors affecting the agricultural product price fluctuation and each of the factors affecting the agricultural product price fluctuation. The output of the agricultural product price fluctuation prediction model is agricultural product price fluctuation 422. In some embodiments, the input data may be a data structure made up of nodes and edges, such as the effect graph shown in FIG. 4.

In some embodiments, the agricultural product price fluctuation prediction model may include a plurality of neural networks. For each neural network, the node characteristics are taken as the input of the neural network, and the prediction output corresponding to the node can be obtained. Further, the prediction output corresponding to the node may also be used as an input to other neural networks. For example, prediction module 220 may construct agricultural product price fluctuation prediction model 421 based on the root nodes, non-root nodes, and directed edges in the influence graph. Each non-root node in the influence graph corresponds to one neural network, and the directed edges of the nodes in the influence graph can represent the input and the output of the corresponding neural network. The root node in the effect graph need not receive as input the output from other effect factor nodes, and thus may be a historical fluctuation of the corresponding effect factor node, i.e., the root node may be a value that does not need to correspond to a neural network, and the value may be an input to the neural network corresponding to the node connected to the root node.

As shown in fig. 5, the agricultural product price fluctuation prediction model 421 may include a neural network 1, a neural network 2, a neural network Q-1, a neural network Q, and the like. The neural network 1, the neural network 2, the neural network Q-1 and the like correspond to a non-root influence factor node 1, a non-root influence factor node 2, a non-root influence factor node N-1 and the like in the influence graph respectively. The neural network Q corresponds to the agricultural product price node N in the impact graph.

In some embodiments, the inputs of the neural network Q-1 corresponding to the non-root-influencing factor node N-1 may include historical fluctuations of the root-influencing factor node m-1 directly connected thereto, node characteristics (e.g., historical fluctuations, etc.) of the non-root-influencing factor node N-1, and outputs of the neural network corresponding to other non-root-influencing factor nodes directly connected to the non-root-influencing factor node N-1. The output of the non-root influencing factor node N-1 may be a future fluctuation of the non-root influencing factor node N-1. Future fluctuations refer to fluctuation values of a preset period and/or point in time after a period and/or point in time corresponding to the historical fluctuations in a manner similar to the period and/or point in time corresponding to the historical fluctuation values. For example, the inputs of the neural network 1 corresponding to the non-root-influencing factor node 1 may include historical fluctuations of the root-influencing factor nodes 411 and 412 directly connected thereto, the node characteristics of the non-root-influencing factor node 1, and the outputs of the non-root-influencing factor node may be future fluctuations thereof. For another example, the inputs of the neural network 2 corresponding to the non-root-influencing factor node 2 may include historical fluctuations of the root-influencing factor nodes 413 and 414 directly connected thereto, the node characteristics of the non-root-influencing factor node 2, and the outputs of the neural network 1 corresponding to the non-root-influencing factor node 1 directly connected to the non-root-influencing factor node 2, the outputs of which may be future fluctuations thereof.

In some embodiments, the input of the neural network Q corresponding to the agricultural product price node N may include the historical fluctuation value of the root influence factor node m directly connected thereto, the output of the neural network Q-1 corresponding to the non-root influence factor node N-1 directly connected thereto, and the historical fluctuation of the agricultural product price node N. The output of the agricultural product price node N is a predicted value of the agricultural product price fluctuation.

It should be noted that the above description with respect to fig. 5 is by way of example only and not by way of limitation. The entries of the non-root node may not be limited as shown in fig. 5. In some embodiments, the number of nodes directly connected to the non-root node may determine the non-root node's entry.

In some embodiments, the input to the agricultural product price fluctuation prediction model 421 may include at least one of a bid matrix, a deal vector. For a description of the relevance of the bid matrix, the deal vector, see the description of FIG. 4.

The agricultural product price fluctuation prediction model 421 may be trained in at least one of joint training and partial training.

Joint training may refer to training together a plurality of neural networks included in the agricultural product price fluctuation prediction model 421. Local training may refer to training of a portion of the plurality of neural networks included in the agricultural product price fluctuation prediction model 421. The prediction module 220 may select a training mode according to actual requirements.

In some embodiments, prediction module 220 may train a price fluctuation prediction model for the agricultural product based on the historical data. The historical data may include, among other things, influence graph data and data that affects sample nodes in the graph. The prediction module 220 may build training samples based on the influence graph data and labels of sample nodes in the influence graph. For example, the effect graph data may include node features of sample nodes, directed edges connecting sample nodes, and effect graph structures. The node characteristics of the sample nodes are the node characteristics of the influence factor nodes and the node characteristics of the price nodes of the agricultural products. The node characteristics of the sample nodes and the related description of the acquisition modes thereof can be referred to in fig. 8, and are not repeated here. The directed edges between the connected sample nodes and their associated description refer to step 410, and are not described in detail herein. The influence graph structure is data describing connection relations between nodes. It will be appreciated that based on the influence graph structure, the prediction module 220 may obtain nodes that are one hop away from the sample node. The relevant description of the structure of the influence diagram is referred to in fig. 4, and will not be repeated here.

In some embodiments, the labels of the sample nodes may be historical fluctuations of the influence factor nodes in the historical data. In some embodiments, the labels of the sample nodes may be obtained by manual entry, reading of stored data, invoking a related interface, or otherwise.

In some embodiments, the prediction module 220 may input the training samples into a machine learning model corresponding to the sample nodes, update model parameters of the machine learning model based on an output of the machine learning model, and obtain the machine learning model corresponding to the trained sample nodes. The machine learning model may be any feasible model including, but not limited to, GNN, LSTM, etc.

In some embodiments, the prediction module 220 may input the training samples into the initial neural network corresponding to the sample node, and output the prediction result of the sample node after the training samples are processed by the initial neural network corresponding to the sample node. The prediction module 220 may construct a loss function based on the prediction result of the sample node and the sample label, and update model parameters of the initial neural network corresponding to the sample node based on the loss iteration. When the loss function of the initial neural network corresponding to the sample node meets the preset condition, training the neural network corresponding to the sample node is completed, and the trained neural network corresponding to the sample node is obtained. The neural network corresponding to the trained sample nodes can be used to construct the agricultural product price fluctuation prediction model 421. The preset conditions may include convergence of the loss function, reaching of the iteration number to a threshold value, and the like. Training methods include, but are not limited to, gradient descent methods, regularization and conjugate gradient methods, and the like.

In some embodiments of the present description, prediction module 220 may obtain the agricultural product price fluctuation prediction model through joint and/or local training. Under some conditions, the method is beneficial to solving the problem that labels are difficult to obtain when the price fluctuation prediction model of the agricultural product is independently trained, and can enable the price fluctuation prediction model of the agricultural product to better reflect the price fluctuation condition of the agricultural product.

The loss function of the agricultural product price fluctuation prediction model may be constructed based on labels of the sample nodes and predicted values of the sample nodes.

In some embodiments, during training of the agricultural product price fluctuation prediction model 421, the loss term of the loss function may include at least one of a loss term of agricultural product price node fluctuation, a loss term of influence factor node fluctuation.

The loss term of the agricultural product price node fluctuation may reflect a relationship between the label of the sample agricultural product price node and the predicted agricultural product price fluctuation. The loss term of the agricultural product price node fluctuation reflects the accuracy of the initial agricultural product price fluctuation prediction model to the agricultural product price fluctuation prediction.

The loss term for the price node fluctuation of the agricultural product can be determined in various ways. For example, the loss term for the agricultural product price node fluctuation may be a label of the sample agricultural product price node and a variance, an average of absolute differences, etc. of the predicted agricultural product price node fluctuation.

The penalty term for the influence factor node fluctuation may reflect the relationship between the labels of the different sample influence factor nodes and the predicted values of the sample influence factor nodes. The loss term of the influence factor node fluctuation reflects the accuracy degree of the initial agricultural product price fluctuation prediction model on the influence factor node fluctuation prediction.

The penalty term affecting factor node fluctuations may be determined in various ways. For example, the penalty term for the influence factor node fluctuation may be the label of the sample influence factor node and the variance of the predicted influence factor node fluctuation, the average of absolute differences, etc.

The loss term of the price node fluctuation of the agricultural product and the loss term of the influence factor node fluctuation can be combined in various ways. Such as summation, weighted summation, etc. The loss function may also include other terms, etc.

In some embodiments, prediction module 220 may determine weight data for the loss function and determine the loss function based on the loss term for the price node fluctuation of the agricultural product, the loss term for the impact factor node fluctuation, and the weight data. The weight data can be used to reflect the extent to which a loss term affecting factor node fluctuation affects in the agricultural product price fluctuation prediction process. The larger the weight data is, the greater the influence degree of the loss items of the corresponding influence factor node fluctuation on the agricultural product price fluctuation prediction is.

In some embodiments, the prediction module 220 may determine weight data corresponding to a penalty term that affects factor node fluctuations based on the node distance. In some embodiments, the node distance may be a distance between the impact factor node and the agricultural price node. For example, the influence factor node a is directly connected with the agricultural product price node (i.e., the distance between the influence factor node a and the agricultural product price node is 1 hop), and the influence factor node b is separated from the agricultural product price node by one influence factor node (i.e., the distance between the influence factor node b and the agricultural product price node is 2 hops), the weight data corresponding to the loss term of the influence factor node a fluctuation is greater than the weight data corresponding to the loss term of the influence factor node b fluctuation.

In some embodiments of the present disclosure, the prediction module 220 may assign different weight data to the loss term affecting the factor node fluctuation based on the extent of the influence of the loss term affecting the factor node fluctuation in the agricultural product price fluctuation prediction process, so that the agricultural product price fluctuation prediction may be more accurate, and further, the adjustment of the quote may be more accurate.

In some embodiments, the penalty term for the influence factor node fluctuation is related to the edge feature of the directed edge corresponding to the influence factor node in the influence graph. Wherein edge features can be used to represent features of causal relationships between two connected nodes. For example, an edge feature may include a degree of reliability that may be used to represent the strength of a causal relationship between two connected nodes.

For a description of the directed edges and their edge features in the influence graph, see step 410, which is not repeated here.

In some embodiments of the present disclosure, the prediction module 220 may influence the loss term of the factor node fluctuation, and consider the edge feature of the directed edge of the connection node in the influence graph, so that the loss term of the factor node fluctuation contains more abundant information, thereby improving the accuracy of the prediction result, and further improving the accuracy of the adjustment of the quotation.

In some embodiments of the present disclosure, an agricultural product price fluctuation prediction model is constructed based on a plurality of neural networks, so that a complex relationship between factors affecting agricultural product price fluctuation can be better simulated, and the agricultural product price fluctuation prediction model can predict fluctuation of agricultural product price based on a multi-level neural network after information exchange. The agricultural product price fluctuation prediction model comprises a plurality of neural networks, node characteristics of influence factor nodes in the image map and causal relations among the influence factor nodes are fully considered, so that efficiency and accuracy of a prediction result are improved, and accuracy of quotation adjustment is improved.

FIG. 6 is an exemplary flow chart for determining price-affecting parameters of agricultural products according to some embodiments of the present description. As shown in fig. 6, the process 600 includes the following steps. In some embodiments, the process 600 may be performed by the processing device 110.

In step 610, an impact factor is obtained. In some embodiments, step 610 may be performed by the acquisition module 210.

In some embodiments, the influencing factors may include a primary influencing factor, a secondary influencing factor, a P-1 influencing factor, a P-influencing factor (P is a positive integer). The first-order influencing factor is an influencing factor directly connected with the price node of the agricultural product. The secondary influencing factor is a influencing factor directly connected with the primary influencing factor. By such a push, the P-level influence factor is an influence factor directly connected with the P-1 level influence factor, etc.

In some embodiments, the acquisition module 210 may acquire the impact factor in a variety of ways, including, but not limited to, acquisition from the network 120, input from the storage device 130, and/or direct user input, among others. In some embodiments, the acquisition module 210 may store the acquired impact factors. For example, in the storage device 130.

And step 620, constructing an influence graph for the directed edge based on the relation of the influence factors as nodes and the influence factors. The content of step 620 is the same as that of step 410, so more details regarding step 620 are described in relation to step 410 and are not repeated here.

Step 630, determining agricultural product price impact parameters based on the impact graph. In some embodiments, step 630 may be performed by determination module 221.

The agricultural product price influencing parameter refers to a parameter influencing the price of the agricultural product. For example, the weight of each influencing factor, etc.

In some embodiments, the agricultural product price impact parameters may be learned based on various algorithms. For example, the algorithms may include linear regression algorithms, logistic regression algorithms, decision trees, artificial neural networks, and the like.

In some embodiments, determination module 221 may construct an agricultural product price fluctuation prediction model based on the impact graph, may train the agricultural product price fluctuation prediction model based on the historical data, and may determine the agricultural product price impact parameters based on model parameters of the agricultural product price fluctuation prediction model. For more details on the construction and training of the agricultural product price fluctuation prediction model, see the relevant description of fig. 5.

In some embodiments, the model parameters of the agricultural product price fluctuation prediction model determine the weight of the agricultural product price fluctuation (i.e., the agricultural product price influence parameter) for each influence factor, so the model parameters are agricultural product price influence parameters.

In some embodiments of the present disclosure, a price fluctuation prediction model is constructed through an influence graph, the price fluctuation prediction model is trained through historical data, and agricultural product price influence parameters are determined through model parameters of the trained price fluctuation prediction model, so that influence of each influence factor on agricultural product price fluctuation can be clarified, accuracy of predicting agricultural product price fluctuation can be improved, and accuracy of adjusting quotation is improved.

In some embodiments of the present disclosure, an impact factor is used to construct an impact graph, and an agricultural product price fluctuation prediction model is constructed to determine an agricultural product price impact parameter, so that the impact of each impact factor on the agricultural product price is more intuitive, the agricultural product price fluctuation can be predicted more accurately, and the quotation can be adjusted more accurately.

FIG. 7 is an exemplary schematic diagram of a model structure for dynamically adjusting a price fluctuation prediction model for agricultural products, according to some embodiments of the present description. In some embodiments, the flow 700 may be performed by the dynamic adjustment module 222.

And step 710, obtaining the prediction effect of the agricultural product price fluctuation prediction model.

The prediction effect 711 refers to a comparison result of the prediction result of the agricultural product price fluctuation model with the true value. For example, the prediction effect may be represented by an absolute error, a relative error, a mean square error, a root mean square error, or the like between the prediction result of the agricultural product price fluctuation prediction model and the true value.

In some embodiments, dynamic adjustment module 222 may obtain a prediction of price fluctuation of the agricultural product by using a price fluctuation prediction model of the agricultural product, and calculate the prediction with a true value to obtain prediction effect 711. The operations may include absolute error, relative error, mean square error, root mean square error, etc.

Step 720, dynamically adjusting a model structure of the agricultural product price fluctuation prediction model based on the prediction effect.

In some embodiments, the agricultural product price fluctuation prediction model may include a plurality of neural networks that are a plurality of linear regression layers. For example, each of the plurality of neural networks is a linear regression layer corresponding to a node of the influence graph.

In some embodiments, dynamic adjustment module 222 may dynamically adjust the model structure of the agricultural product price fluctuation prediction model based on the eigenvalues that affect the agricultural product price fluctuation prediction result. In some embodiments, the characteristic values that affect the outcome of the agricultural product price fluctuation prediction may include a confidence 721 or a contribution 722, or the like. For more on the trustworthiness 721, see the relevant description of fig. 4.

In some embodiments, the adjustment of the confidence level by the dynamic adjustment module 222 may include a single factor chain adjustment, an overall adjustment, or the like.

In some embodiments, the dynamic adjustment module 222 adjusts the confidence level by a single factor chain. The dynamic adjustment module 222 may obtain the primary impact factor K _1max corresponding to the greatest weight (in the linear regression layer) from the primary impact factors K ₁ (where K represents some impact factor that meets the requirements and 1 represents one). The dynamic adjustment module 222 may reduce the confidence level of the impact factor K _1max corresponding to the directed edge connected to the price node of the agricultural product, retrain the price fluctuation prediction model based on the historical data, and evaluate the prediction effect. If the prediction effect does not reach the standard, the dynamic adjustment module 222 obtains all the secondary influence factors K ₂ connected to the primary influence factor K ₁, and obtains the secondary influence factor K _2max corresponding to the maximum weight (in the linear regression layer). The dynamic adjustment module 222 may reduce the confidence level of the directed edge of the impact factor K _2max that is connected to the primary impact factor K ₁, retrain the agricultural product price fluctuation prediction model based on the historical data, and evaluate the prediction effect. If the prediction effect reaches the standard (the standard may be default or manually input), the adjustment reliability is stopped, and the dynamic adjustment module 222 may obtain an adjusted agricultural product price fluctuation prediction model. If the prediction effect does not reach the standard, the dynamic adjustment module 222 may adjust the credibility (such as three-level influence factor K ₃, four-level influence factor K ₄, etc.) corresponding to the directional edge where the influence factor corresponding to the maximum weight is connected with the last-level influence factor step by step until the last-level influence factor (such as P-level influence factor), so as to obtain the adjusted agricultural product price fluctuation prediction model.

In some embodiments, the dynamic adjustment module 222 adjusts the confidence level by a whole. The dynamic adjustment module 222 may obtain weights (in the linear regression layer) corresponding to all the influence factors in the influence graph, rank the weights from large to small, use the influence factors corresponding to the first n weights as the target influence factors K _target, reduce the reliability of the directional edge corresponding to the last-stage influence factor connected with the target influence factor K _target (for example, if the target influence factor is a 3-stage influence factor and the last-stage influence factor is a 2-stage influence factor), reduce the amplitude by a preset value, retrain the price fluctuation prediction model of the agricultural product after reduction, and evaluate the prediction effect. If the prediction effect reaches the standard, the dynamic adjustment module 222 stops adjusting the reliability, and obtains an adjusted agricultural product price fluctuation prediction model. And if the prediction effect does not reach the standard, repeating the previous step to continuously adjust the credibility until the standard is reached, and stopping adjusting, thereby obtaining an adjusted agricultural product price fluctuation prediction model.

In some embodiments, the dynamic adjustment module 222 may obtain the reliability of the edge feature of the directed edge based on the prediction effect, determine whether the reliability is less than a reliability threshold, and delete the directed edge corresponding to the reliability when the reliability is less than the reliability threshold. For more details on the confidence level, see the relevant description of fig. 4.

In some embodiments, the confidence threshold may be a pre-set value. In some embodiments, the confidence threshold may include a primary confidence threshold, a secondary confidence threshold, a tertiary confidence threshold, a P-level confidence threshold, and the like. The confidence threshold value of each stage may be the same or different.

In some embodiments, the prediction module 220 may determine the trustworthiness of the directed edge correspondence by historical data between two impact factor nodes of the directed edge correspondence. For more details on the acquisition of trustworthiness, see the relevant description of fig. 4.

In some embodiments, the dynamic adjustment module 222 may obtain an adjusted agricultural product price fluctuation prediction model when the prediction effect meets a criterion; when the predicted effect does not reach the standard, the dynamic adjustment module 222 may obtain the credibility of the directed edge. The dynamic adjustment module 222 may compare whether each obtained reliability is less than the reliability threshold, delete the directed edge corresponding to the reliability when the reliability is less than the reliability threshold, retrain the agricultural product price fluctuation prediction model using training samples (such as historical data) to obtain a new prediction effect, and determine whether the new prediction effect meets the standard. If the criteria are met, the dynamic adjustment module 222 may directly obtain the adjusted agricultural product price fluctuation model. If the prediction effect does not reach the standard, the dynamic adjustment module 222 may repeat the foregoing step of adjusting the reliability until the prediction effect reaches the standard, thereby obtaining the adjusted agricultural product price fluctuation prediction model.

In some embodiments, dynamic adjustment module 222 may obtain the trustworthiness of the directed edge between each level one impact factor node and the agricultural price node in the impact graph. The dynamic adjustment module 222 may compare the obtained reliability with the first-level reliability threshold, and delete the directed edge between the first-level influence factor node corresponding to the reliability and the price node of the agricultural product if the reliability between the first-level influence factor node and the price node of the agricultural product is less than the first-level reliability threshold.

In some embodiments, the dynamic adjustment module 222 may obtain the trustworthiness of the directed edge between each secondary and primary impact factor node in the impact graph. The dynamic adjustment module 222 may compare the obtained reliability with the second-level reliability threshold, and delete the directed edge between the second-level influence factor node and the first-level influence factor node corresponding to the reliability if the reliability between the second-level influence factor node and the first-level influence factor node is smaller than the second-level reliability threshold.

Similarly, the dynamic adjustment module 222 may repeat the foregoing steps to determine the degree of reliability of each level of influence factor nodes and the previous level of influence factor nodes and the magnitude of the reliability threshold of each level, thereby adjusting the agricultural product price fluctuation prediction model. It should be noted that the step of adjusting the reliability of each stage may be performed step by step, or may be performed simultaneously.

In some embodiments of the present disclosure, the credibility of the directed edges between the influence factor nodes is obtained through the prediction effect, and the preset credibility threshold is set to screen the directed edges between the influence factor nodes corresponding to the lower credibility, so that the constructed agricultural product price fluctuation prediction model is more reasonable, the prediction accuracy is improved, and the accuracy of adjusting the quotation is further improved.

In some embodiments, the dynamic adjustment module 222 may obtain a contribution value for each impact factor node based on the predicted effect, and the dynamic adjustment module 222 may determine whether the contribution value is less than a contribution threshold. When the contribution value is less than the contribution threshold, the dynamic adjustment module 222 may delete the impact factor node to which the contribution value corresponds.

The contribution value may be represented by a weight corresponding to the influence factor node (in the linear regression layer), and the smaller the absolute value of the weight, the smaller the contribution value to the influence factor node of the previous stage. For example, if the absolute value of the weight of the secondary influence factor node is smaller than the secondary contribution threshold, the contribution value of the secondary influence factor node to the primary influence factor node is lower and can be ignored. In some embodiments, the contribution threshold may include a primary contribution threshold, a secondary contribution threshold, a tertiary contribution threshold, a P-stage contribution threshold, etc., the contribution threshold of each stage may be the same or different.

In some embodiments, the dynamic adjustment module 222 may obtain an adjusted agricultural product price fluctuation prediction model when the prediction effect meets a criterion; when the predicted effect does not reach the standard, the dynamic adjustment module 222 may obtain a contribution value of each influence factor node. The dynamic adjustment module 222 compares whether the contribution value of each obtained influence factor node is smaller than the contribution value threshold, when the contribution value is smaller than the contribution value threshold, the dynamic adjustment module 222 can delete the influence factor node corresponding to the contribution value, correspondingly, also delete the directed edge corresponding to the influence factor node, retrain the agricultural product price fluctuation prediction model by using a training sample (such as historical data), obtain a new prediction effect, determine whether the new prediction effect reaches the standard, and if so, directly obtain the adjusted agricultural product price fluctuation model. If the prediction effect does not reach the standard, the dynamic adjustment module 222 may repeat the foregoing step of adjusting the contribution value until the prediction effect reaches the standard, thereby obtaining the adjusted agricultural product price fluctuation prediction model.

In some embodiments, the dynamic adjustment module 222 may obtain weights corresponding to all the primary influence factor nodes (linear regression layer), and the dynamic adjustment module 222 may compare the absolute value of the weight corresponding to the primary influence factor node with the magnitude of the primary contribution threshold, and if the absolute value of the weight corresponding to a certain primary influence factor node is smaller than the primary contribution threshold, delete the primary influence factor node, and correspondingly delete the directed edge of the primary influence factor node reaching the price node of the agricultural product.

Similarly, the dynamic adjustment module 222 may repeat the foregoing steps to determine the absolute value of the weight corresponding to each level of the impact factor node and the magnitude of the contribution threshold corresponding to each level, thereby adjusting the agricultural product price fluctuation prediction model. It should be noted that the step of adjusting the contribution value may be performed step by step or may be performed simultaneously.

In some embodiments of the present disclosure, by comparing the contribution value of each level of the influence factor node with the corresponding contribution threshold of each level, so as to screen out the influence factor node with a lower contribution value, a more reasonable agricultural product price fluctuation prediction model can be constructed, and the accuracy of prediction is improved, so that the accuracy of price adjustment is improved.

In some embodiments of the present disclosure, the model structure of the prediction model is dynamically adjusted by obtaining the prediction effect of the prediction model of the price fluctuation of the agricultural product in real time, so that the prediction result is more accurate, the adjustment of the price is more accurate, and the application range is wider.

Fig. 8 is an exemplary flow chart of a method of automatically quoting agricultural products according to some embodiments of the present description. In some embodiments, the process 800 may be performed by the processing device 110. As shown in fig. 8, the process 800 may include the steps of:

Step 810, acquiring quotation information.

The offer information may refer to a selling price of the agricultural product offered by the seller of the agricultural product. In some embodiments, the commodity seller may include one or more of a commodity producer, a commodity provider, a commodity distributor, a commodity retailer, and the like.

In some embodiments, the acquisition module 210 may acquire the offer information through a pricing system. The pricing system may refer to an information system that collects price quotation information of agricultural products across the country through the network 120, and aggregates and uniformly provides the price quotation information. Wherein the offer information may include a plurality of offers, each of the plurality of offers may be a price of the agricultural commodity offered for one of the plurality of agricultural commodity sellers.

In some embodiments, the offer information may include agricultural product provider offer information and a credibility factor. Acquisition module 210 may acquire agricultural product provider pricing information and/or a reliability factor.

The credibility factor may be used to describe the credibility of the thing. In some embodiments, the credibility factor may be used to describe the degree of credibility of the offer information provided by the agricultural product provider. For example, the credibility factor may be a degree of deviation of the agricultural product price of the agricultural product provider from the agricultural product price quote. In some embodiments, each agricultural product provider may correspond to a trust factor.

In some embodiments, acquisition module 210 may acquire agricultural product provider pricing information via a pricing system. In some embodiments, the acquisition module 210 may determine the confidence factor by calculating a degree of deviation of the commodity price of the commodity provider from the commodity price. For example, the reliability factor is K, and the calculation formula of K is k= (a-B)/a, where a is the price of a certain agricultural product provider; b is the price of the commodity provider.

In some embodiments, the acquisition module 210 may construct a quotation matrix for the quotation information based on the commodity provider pricing information and the credibility factor.

The quotation matrix is a matrix formed by taking the corresponding volume of the different quotation intervals and the credibility intervals as matrix elements. The amount of the agricultural product may be an amount of the agricultural product supplied by the agricultural product provider satisfying the above-described conditions of the quotation interval and the credibility interval. An interval of offers refers to a collection of one or more offers. For example, the interval [50, 60] may be a bid interval representing a set of bids equal to or greater than 50 and equal to or less than 60. A reliability interval refers to a set of one or more reliability factors. For example, the interval [90%,100% ] may be a reliability interval, which represents a set of reliability factors of 90% or more and 100% or less.

In some embodiments, the obtaining module 210 may obtain a plurality of offer information and a plurality of credibility factors corresponding to a plurality of agricultural product suppliers, thereby constructing an offer matrix of offer information. The bid for each row of elements in the bid matrix may correspond to a bid interval, and the confidence factors for each column of elements in the bid matrix may correspond to a confidence interval. For example, quotation matrixThe bid matrix A indicates that the bid is divided into four bid intervals and the confidence factors are divided into four confidence intervals. Matrix element C _ij (i, j=1, 2,3, 4) represents the agricultural product yield of the agricultural product provider for which the bid belongs to the ith bid interval and for which the credibility factor belongs to the jth credibility interval.

In some embodiments, the interval corresponding to each column element in the offer matrix may be referred to as a degree of influence interval. For more information on the degree of influence see the relevant description of step 830.

In some embodiments of the present description, the acquisition module 210 may construct a quotation matrix based on the quotation information of the agricultural product suppliers and the credibility factor, and may reflect the reliability degree of the quotation information of different agricultural product suppliers, so as to ensure the reasonability of the final agricultural product price prediction and the quotation of the agricultural product.

Step 820, obtain the transaction vector.

The trading vector is a vector formed by taking the trading volume corresponding to different trading price intervals as vector elements. The amount of the agricultural product may be an amount of the agricultural product offered by the agricultural product provider satisfying the condition of the price section of the above-mentioned amount of the agricultural product offered. An exchange price interval refers to a collection of one or more exchange prices. For example, the section [50, 60] may be a transaction price section representing a set of transaction prices of 50 or more and 60 or less. For example, the transaction vector a= (D ₁,D₂,D₃,D₄). The transaction vector a indicates that the transaction price is divided into four transaction price sections. Vector element D _i (i=1, 2,3, 4) represents the agricultural product volume of the agricultural product provider whose price of the deal belongs to the i-th price interval of the deal.

In some embodiments, the acquisition module 210 may acquire the deal vector through historical agricultural product sales data. The agricultural product sales data may include a price of the deal, an amount of the deal, a date of the deal, and other information. In some embodiments, the obtaining module 210 may count the amount of the historical agricultural product sales data that the price of the transaction is in the same price interval, and use the amount of the transaction as the element value of the corresponding price interval in the transaction vector.

In step 830, public opinion information is obtained.

The public opinion information is information which is generated through consciousness of people and can influence the price of agricultural products. In some embodiments, the public opinion information may include the price information of agricultural products predicted by the institution and their impact. The influence degree may refer to the degree of influence of information on the outside world. In some embodiments, the impact level may be used to measure the impact of a piece of commodity price information on the mass. In some embodiments, the degree of influence may be expressed as a numerical value. For example, the influence degree may be a value of 0 or more, and the larger the value is, the greater the influence degree of the agricultural product price information on the public is.

In some embodiments, the public opinion information may include a public opinion matrix. The public opinion matrix may be constructed based on different degrees of influence, which may be determined based on the propagation relationship of the predicted price.

The public opinion matrix is a matrix formed by taking the number of nodes of the propagation mechanism corresponding to different predicted price intervals and different influence intervals as matrix elements. In some embodiments, the predicted price of each row element in the public opinion matrix may correspond to a predicted price interval, and the influence of each column element in the public opinion matrix may correspond to an influence interval. For example, public opinion matrixThe public opinion matrix B indicates that the predicted price is divided into four predicted price intervals, and the influence is divided into four influence intervals. Matrix element E _ij (i, j=1, 2,3, 4) represents the number of propagation mechanisms whose predicted price belongs to the i-th predicted price interval and whose influence belongs to the j-th influence interval.

The forecast price may refer to a price of the agricultural product forecast by the forecast authority. A forecasting agency may refer to a agency that forecasts and/or publishes prices for agricultural products. In some embodiments, the predicted price may be represented as a numerical value. For example, the predicted price may be a value greater than 0. A predicted price interval refers to a set of one or more predicted prices. For example, the section [50, 60] may be a predicted price section representing a set of predicted prices of 50 or more and 60 or less. An influence region refers to a set of one or more influence values. For example, the interval [0.1,0.5] may be an influence interval representing a set of influence degrees of 0.1 or more and 0.5 or less.

The acquisition module 210 may determine the fixed loudness based on the propagation relationship of the predicted price. The propagation relationship may refer to the propagation of information outside. In some embodiments, the propagation relationship may be described based on at least one propagation characteristic. For example, the propagation characteristics include propagation breadth, propagation depth, and the like. The propagation breadth may refer to the number of times the predicted price is propagated. In some embodiments, the acquisition module 210 may determine the propagation breadth of the predicted price as the impact of the predicted price. For example, the prediction mechanism may send out a piece of predicted price information, where the predicted price information is published on 5 intermediate nodes (such as websites, microblog account numbers, etc.), and the obtaining module 210 may determine that the influence of the predicted price information is 5.

The propagation depth may refer to the audience size of a node when the predicted price is propagated to that node. The audience size may refer to the audience's attention to the predicted price information, such as the amount of browsing, the number of reviews, the number of praise, etc. In some embodiments, the acquisition module 210 may determine the impact of the predicted price by the propagation depth of the predicted price. For example, the influence of the predicted price may be equal to the audience's attention to the predicted price information. For example, the prediction mechanism sends a piece of predicted price information, the predicted price information is transmitted to 2 intermediate nodes (such as websites, microblog account numbers, etc.), the total browsing amount of the audience of the 2 intermediate nodes on the predicted price information is 10000, and the obtaining module 210 can determine that the influence degree of the predicted price information is 10000. In some embodiments, the impact of the predicted price may also be determined based on the propagation breadth and the propagation depth weighting of the predicted price.

In some embodiments of the present disclosure, a public opinion matrix is constructed through nodes with different influence degrees, so that the influence degree of different predicted price information on the market can be reflected, and the influence degree of different public opinion information on the price of the agricultural product is quantized, thereby improving the accuracy of predicting the price of the agricultural product and further improving the quotation accuracy of the agricultural product.

In some embodiments, the obtaining module 210 may obtain public opinion information through the internet. For example, the acquisition module 210 may capture price forecast information for each forecast authority on the public website via a web crawler; for another example, the obtaining module 210 may capture one or more of browsing amount, praise number, comment number, forwarding number, etc. of each piece of price prediction information while capturing the price prediction information by the web crawler, and calculate the influence degree of each piece of price prediction information.

In some embodiments, the acquisition module 210 may acquire a propagation path of the predicted price.

A propagation path may refer to a path formed by the transfer of information in one or more propagation media. For example, one propagation path may be: price prediction mechanism- > propagation medium 1- > propagation medium 2- > propagation medium 3. A propagation medium may refer to a carrier that conveys the information. The propagation medium may include newspapers, magazines, radio, television, networks, etc.

In some embodiments, the acquisition module 210 may construct a knowledge-graph based on the propagation path.

The knowledge-graph may include nodes, edges between nodes. In some embodiments, a node may correspond to a price prediction mechanism or propagation medium on a propagation path, and an edge between nodes may represent a propagation direction of the propagation path.

In some embodiments, the nodes in the knowledge-graph may include propagation mechanism nodes, each of which may correspond to one of the propagation mechanism entities. Each propagation entity may publish its predicted price to the outside. In some embodiments, the propagation entity may generate a predicted price and issue it externally, e.g., the propagation entity is a price prediction mechanism. In some embodiments, the propagation entity may obtain the predicted price from other propagation entities and issue it to the outside, e.g., a newspaper company, a radio station, a television station, etc. In some embodiments, an edge of the knowledge-graph may include a propagation relationship of the predicted price, the edge may be a directed edge, and a direction of the edge may represent a direction of the propagation relationship.

In some embodiments, nodes in the knowledge-graph may have node features. In some embodiments, the node characteristics of the propagation facility nodes may include predicted price, attention, and credibility.

Attention may refer to the degree of attention of the masses to the spreading mechanism. In some embodiments, the degree of interest of a propagation facility node may be determined based on the number of people who are interested in the propagation facility. For example, the average number of listeners to each program in a certain station for a period of time (for example, one month) is counted and used as the attention of the station.

The credibility may refer to the credibility of the predicted price published by the propagation facility node. In some embodiments, the trustworthiness of the propagation facility node may be determined based on a plurality of historical published predicted prices. For example, the reliability of the propagation facility node may be equal to an average of the accuracy of the plurality of historically published predicted prices, which may be determined based on a difference ratio of the predicted prices to the actual prices.

In some embodiments, edges in the knowledge-graph may have edge features. In some embodiments, the propagation mechanism node and the edge feature of the propagation mechanism node may include a propagation force.

Propagation forces may refer to the ability of information to achieve efficient propagation. In some embodiments, the propagation force of an edge may be determined based on node characteristics of two propagation mechanism nodes to which the edge is connected. For example, the propagation force of an edge feature may be determined by the ratio of the attention of the end node to the attention of the start node of the directed edge.

In some embodiments, the node characteristics of the propagation mechanism node may also include propagation forces and effects.

In some embodiments, the propagation force of a propagation mechanism node may be determined based on the propagation force of a propagation path that originates at the propagation mechanism node. For example, the propagation force of a propagation mechanism node may be equal to the sum of the propagation forces of propagation paths starting from the propagation mechanism node. The propagation force of a propagation path may be determined based on the propagation force of a directed edge contained in the propagation path. For example, the propagation force of the propagation path "a→b→c" may be equal to the propagation force of the side "a→b" x the propagation force of the side "b→c".

In some embodiments, the impact of a propagation mechanism node may be determined based on the trustworthiness of the propagation mechanism node and the propagation force. For example, the impact of a propagation mechanism node may be equal to the product of the reliability of the propagation mechanism node and the propagation force. In some embodiments, knowledge maps may be constructed based on expert experience.

In some embodiments, the obtaining module 210 may perform influence calculation on the knowledge graph to obtain a public opinion matrix of public opinion information. For example, the obtaining module 210 may obtain the predicted price and the influence degree of each propagation mechanism node in the knowledge graph, and based on the predicted price and the influence degree, correspond each propagation mechanism node to one predicted price interval and one influence degree interval, and count the number of propagation mechanism nodes under each predicted price interval and each influence degree interval, and use the number of propagation mechanism nodes as element values of the corresponding predicted price interval and the influence degree interval in the public opinion matrix.

In some embodiments, the nodes of the knowledge-graph may include network media nodes, and the node characteristics of the network media nodes may include a farmer browsing volume and a user total browsing volume.

In some embodiments, there may be a directed edge between the network medium node and the propagation mechanism node, the network medium node and the network medium node, representing a propagation relationship of the predicted price from a start node to an end node of the directed edge. In some embodiments, the edge characteristics of the network medium nodes and network medium nodes, network medium nodes and propagation mechanism nodes may include propagation forces.

The farmer view amount may refer to a farmer user access amount of a web page containing predicted price information. In some embodiments, the amount of farmer browsing may be obtained by processing device background data. The user total browsing amount may refer to a user total access amount of a web page containing predicted price information. In some embodiments, the total amount of user browsing may be obtained by processing device background data.

In some embodiments, the acquisition module 210 may acquire a ratio of the amount of user browsing to the total amount of user browsing, and determine the degree of influence of the propagation mechanism node based on the ratio. For example, the acquisition module 210 may determine the propagation forces of propagation paths that include the network media node based on the farmer's browsing volume ratio of the network media node, and thus determine the impact of the propagation mechanism node. Illustratively, the propagation forces of the propagation path "propagation mechanism node a→network medium node b→propagation mechanism node C" including network medium node B are: the propagation force of the side 'A-B' is multiplied by the propagation force of the side 'B-C', wherein alpha represents the ratio of the browsing amount of a peasant household of the network medium node B on the propagation path. In some embodiments, the acquisition module 210 may determine the propagation forces of the propagation paths containing the plurality of network media nodes based on the farmer's browsing volume duty cycle of the plurality of network media nodes. For example, the acquisition module 210 may determine the propagation force of a propagation path containing a plurality of network media nodes based on an average farmer browsing volume ratio of the network media nodes. Illustratively, for one propagation path: propagation mechanism node A- & gt network medium node B- & gt propagation mechanism node C- & gt network medium node D, wherein the propagation force of the propagation path is as follows: the propagation force of (1+β) x side "a→b" x the propagation force of side "b→c" x the propagation force of side "c→d", where β represents the average farmer view ratio of network medium node B and network medium node D on the propagation path.

In some embodiments of the present disclosure, by introducing network media nodes and features of the browsing amount of the farmers, different users may obtain the predicted price information and then determine the influence degree on the price of the agricultural product, and based on the ratio of the browsing amount of the farmers in the total browsing amount of the users, the influence degree of different public opinion information on the price of the agricultural product may be quantified more accurately, so that the quotation of the agricultural product may be determined more accurately.

In some embodiments, the nodes of the knowledge-graph may include audience nodes, and the node characteristics of the audience nodes may include a farmer type vector.

An audience may refer to a person who is paying attention to price information of agricultural products through some way. Each audience node may correspond to an audience entity. In some embodiments, the audience may include groups of general users, farmers, and the like. The farmer type vector may be used to represent a plurality of farmer type intervals with a maximum ratio among the audience nodes. In some embodiments, the farmer type interval may include one or more of an age interval, an annual revenue interval, an annual production interval, an annual consumption interval, and the like. For example, one farmer type vector is as follows: (1,3,2,1) the vector contains four elements, representing respectively an age interval, an annual income interval, an annual production interval, an annual consumption interval to which the farmer belongs, the first element being 1 representing that the farmer is located in the first age interval (e.g. 1 representing 0-20 years old, 2 representing 20-40 years old, 3 representing 40-60 years old, etc.), the second element being 3 representing that the farmer is located in the third annual income interval (e.g. 1 representing 0-5 ten thousand, 2 representing 5-10 ten thousand, 3 representing 10-15 ten thousand, etc.), etc. In some embodiments, the farmer type interval to which each farmer belongs may be obtained through account registration information of the farmer.

In some embodiments, the network medium node and the audience node, the propagation facility node and the audience node may have directed edges that may represent propagation relationships of the predicted price. In some embodiments, the edge characteristics of the network media nodes and the audience nodes, the propagation facility nodes and the audience nodes may include propagation forces.

In some embodiments, the acquisition module 210 may determine the impact level of the propagation facility node based on the farmer type vector. For example, the acquisition module 210 may determine the propagation force of the propagation path containing the audience node based on the farmer type vector of the audience node, and thus determine the impact level of the propagation facility node. In some embodiments, the acquisition module 210 may determine the preset factor based on a farmer type vector of an audience node, and determine the propagation force of a propagation path containing the audience node based on the preset factor. The preset factors may be empirically preset, e.g., for each different farmer type vector, there is a preset factor corresponding to it. Illustratively, the propagation forces of the propagation path including the audience node C, namely the propagation mechanism node A, the network medium node B and the audience node C, are as follows: the propagation force of the side 'A- & gt B' is multiplied by the propagation force of the side 'B- & gt C', wherein k represents a preset factor corresponding to the peasant family type vector of the network medium node B on the propagation path.

In some embodiments of the present disclosure, by introducing audience nodes and farmer type vector features, different farmers can be represented to obtain the influence degree of predicted price information on the price of the agricultural product, and the fixed loudness is determined based on the farmer type vector, so that the influence degree of different public opinion information on the price of the agricultural product can be more accurately quantified, and further the quotation of the agricultural product can be more accurately determined.

In some embodiments of the present disclosure, the influence degree of the knowledge graph is calculated and the public opinion matrix is obtained by constructing the knowledge graph, and the influence degree of different public opinion information on the price of the agricultural product can be more accurately quantified by the influence degree, so that the quotation of the agricultural product can be more accurately determined.

And step 840, predicting the price of the agricultural product through the agricultural product price prediction model based on at least one of the quotation information, the bargain and the public opinion information. The agricultural product price prediction model is a machine learning model. In some embodiments, step 840 may be performed by prediction module 220.

The agricultural product price prediction model refers to a model that can predict the price of agricultural products. In some embodiments, the types of agricultural product price prediction models may include deep neural networks, recurrent neural networks, etc., and the selection of model types may be contingent on the circumstances.

In some embodiments, the input of the agricultural product price prediction model may include at least one of quotation information, a deal vector, and public opinion information. The output of the agricultural product price prediction model may include the agricultural product price at a future time or time period. The future time or period may be arbitrary and is typically determined by the labels of the training samples in the model training. Illustratively, the sample label when training the agricultural product price prediction model is the agricultural product price (average value) for one week (period) in the future, and the trained agricultural product price prediction model can be used to predict the agricultural product price (average value) for one week (period) in the future.

In some embodiments, the agricultural product price prediction model may be a long-short term memory model.

In some embodiments, the agricultural product price prediction model may be derived based on a plurality of training samples and label training.

In some embodiments, the training samples include corresponding sample bid information, deal vectors, public opinion information, etc. for a plurality of different historical moments. The label is the price of the agricultural product at a corresponding sample future time or future time period of a plurality of different historical time instants, etc. The training data can be acquired based on historical data, and the label of the training data can be determined by means of manual labeling or automatic labeling. Inputting the training sample with the label into an agricultural product price prediction model, updating parameters of the agricultural product price prediction model through training, and obtaining a trained agricultural product price prediction model after training is finished when the trained agricultural product price prediction model meets preset conditions.

In some embodiments, the regularization term in the loss function may be inversely related to the product of the row space dimension and the column space dimension of the public opinion matrix. For example, the regularization term in the penalty function may be multiplied by a weight, which may be inversely related to the product of the row space and column space dimensions of the matrix. Illustratively, the loss function of the model is: loss=c ₀ +k×r, where C ₀ is a Loss function term, R is a regularization term, and k is a weight of the regularization term, and its value may be: k=1/W, where W is the product of the row space dimension and the column space dimension of the public opinion matrix.

In some embodiments, the prediction module 220 may process at least one characteristic of the offer information, the deal vector, the public opinion information based on the agricultural product price prediction model to predict the agricultural product price.

A price quote for the agricultural product is determined based on the predicted price of the agricultural product, step 850. In some embodiments, step 850 may be performed by the determination module 221.

In some embodiments, determination module 221 may determine a price for the agricultural product based on the predicted price for the agricultural product. For example, the determination module 221 may determine the predicted price of the agricultural product as a price quote for the agricultural product.

In some embodiments of the present disclosure, the price of the agricultural product is predicted based on the quotation information, the bargaining vector and the public opinion information, so that multiparty factors affecting the price of the agricultural product are effectively considered, and meanwhile, the price of the agricultural product is predicted by using a machine learning model, so that the accuracy of the price prediction of the agricultural product is greatly improved, and the accuracy of the quotation of the agricultural product is further improved.

It should be noted that the above description of the flow is only for the purpose of illustration and description, and does not limit the application scope of the present specification. Various modifications and changes to the flow may be made by those skilled in the art under the guidance of this specification. However, such modifications and variations are still within the scope of the present description.

While the basic concepts have been described above, it will be apparent to those skilled in the art that the foregoing detailed disclosure is by way of example only and is not intended to be limiting. Although not explicitly described herein, various modifications, improvements, and adaptations to the present disclosure may occur to one skilled in the art. Such modifications, improvements, and modifications are intended to be suggested within this specification, and therefore, such modifications, improvements, and modifications are intended to be included within the spirit and scope of the exemplary embodiments of the present invention.

Meanwhile, the specification uses specific words to describe the embodiments of the specification. Reference to "one embodiment," "an embodiment," and/or "some embodiments" means that a particular feature, structure, or characteristic is associated with at least one embodiment of the present description. Thus, it should be emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various positions in this specification are not necessarily referring to the same embodiment. Furthermore, certain features, structures, or characteristics of one or more embodiments of the present description may be combined as suitable.

Furthermore, the order in which the elements and sequences are processed, the use of numerical letters, or other designations in the description are not intended to limit the order in which the processes and methods of the description are performed unless explicitly recited in the claims. While certain presently useful inventive embodiments have been discussed in the foregoing disclosure, by way of various examples, it is to be understood that such details are merely illustrative and that the appended claims are not limited to the disclosed embodiments, but, on the contrary, are intended to cover all modifications and equivalent arrangements included within the spirit and scope of the embodiments of the present disclosure. For example, while the system components described above may be implemented by hardware devices, they may also be implemented solely by software solutions, such as installing the described system on an existing server or mobile device.

Likewise, it should be noted that in order to simplify the presentation disclosed in this specification and thereby aid in understanding one or more inventive embodiments, various features are sometimes grouped together in a single embodiment, figure, or description thereof. This method of disclosure does not imply that the subject matter of the present description requires more features than are set forth in the claims. Indeed, less than all of the features of a single embodiment disclosed above.

In some embodiments, numbers describing the components, number of attributes are used, it being understood that such numbers being used in the description of embodiments are modified in some examples by the modifier "about," approximately, "or" substantially. Unless otherwise indicated, "about," "approximately," or "substantially" indicate that the number allows for a 20% variation. Accordingly, in some embodiments, numerical parameters set forth in the specification and claims are approximations that may vary depending upon the desired properties sought to be obtained by the individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and employ a method for preserving the general number of digits. Although the numerical ranges and parameters set forth herein are approximations that may be employed in some embodiments to confirm the breadth of the range, in particular embodiments, the setting of such numerical values is as precise as possible.

Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments of this specification. Other variations are possible within the scope of this description. Thus, by way of example, and not limitation, alternative configurations of embodiments of the present specification may be considered as consistent with the teachings of the present specification. Accordingly, the embodiments of the present specification are not limited to only the embodiments explicitly described and depicted in the present specification.

Claims

1. The utility model provides a method for constructing public opinion information of agricultural products, which is characterized in that the method comprises the following steps:

Acquiring a propagation path of a predicted price;

Constructing a knowledge graph based on the propagation path, wherein the knowledge graph comprises nodes, edges between the nodes and the nodes, the nodes comprise propagation mechanism nodes, network medium nodes and audience nodes, node characteristics of the propagation mechanism nodes comprise the predicted price, attention, credibility, propagation force and influence degree, node characteristics of the network medium nodes comprise peasant household browsing amount and user total browsing amount, node characteristics of the audience nodes comprise peasant household type vectors, the peasant household type vectors are used for representing a plurality of peasant household type intervals with the largest proportion in the audience nodes, and the peasant household type intervals comprise one or more of age intervals, annual income intervals, annual interval and annual consumption intervals; the edge represents a propagation direction of the propagation path, and the edge feature of the edge comprises a propagation force;

And calculating influence degree of the knowledge graph to generate a public opinion matrix, wherein the public opinion matrix is a matrix formed by taking different predicted price intervals and the number of the transmission mechanism nodes corresponding to the different influence degree intervals as matrix elements, the predicted price of each row of elements in the public opinion matrix corresponds to one predicted price interval, and the influence degree of each column of elements in the public opinion matrix corresponds to one influence degree interval.

2. The method of claim 1, wherein the performing influence calculation on the knowledge graph, generating a public opinion matrix comprises:

acquiring the ratio of the browsing amount of the farmer in the total browsing amount of the user;

Determining the degree of influence based on the duty cycle;

And generating the public opinion matrix based on the influence degree.

3. The method of claim 2, wherein the performing influence calculation on the knowledge graph, generating a public opinion matrix comprises:

determining the influence degree based on the farmer type vector;

And generating the public opinion matrix based on the influence degree.

4. The method according to claim 1, wherein the method further comprises:

acquiring quotation information;

obtaining an intersection vector;

Predicting the price of the agricultural product through an agricultural product price prediction model based on at least one of the quotation information, the diagonalization vector and the public opinion matrix, wherein the agricultural product price prediction model is a machine learning model;

Based on the predicted price of the agricultural product, a price quote for the agricultural product is determined.

5. A system for constructing public opinion information of agricultural products, the system comprising:

An acquisition module for:

Acquiring a propagation path of a predicted price;

6. The system of claim 5, wherein the acquisition module is further to:

Determining the degree of influence based on the duty cycle;

And generating the public opinion matrix based on the influence degree.

7. The system of claim 6, wherein the acquisition module is further to:

determining the influence degree based on the farmer type vector;

And generating the public opinion matrix based on the influence degree.

8. The system of claim 5, wherein the system further comprises:

the acquisition module is further used for acquiring quotation information;

The acquisition module is further used for acquiring the transaction vector;

The prediction module is used for predicting the price of the agricultural product through an agricultural product price prediction model based on at least one of the quotation information, the diagonalization vector and the public opinion matrix, wherein the agricultural product price prediction model is a machine learning model;

A determination module for determining a price of the agricultural product based on the predicted price of the agricultural product.

9. An agricultural product public opinion information construction apparatus, the apparatus comprising:

at least one storage medium storing computer instructions;

At least one processor executing the computer instructions to implement the agricultural product public opinion information construction method of any of claims 1-4.

10. A computer-readable storage medium storing computer instructions that, when read by a computer, perform the agricultural product public opinion information construction method of any of claims 1-4.