Crop cultivation support device, crop cultivation support method, and crop cultivation support program

Abstract

In order to assist the growth of crops, the crop growth assistance device (1) is provided with: a receiving unit (11) for receiving a request comprising any of the size, taste, harvest time, and yield of the crop; a generating unit (12) for generating response information including a method for growing the crop on the basis of the request and a trained model obtained by learning relationships between methods for growing multiple crops and growth results comprising any of the size, taste, harvest time, and yield of the multiple crops; and an output unit (13) for outputting said growth method.

Description

Technical Field

[0001] The present invention relates to a crop growth support device that generates information related to the growth of crops and the like.

Background Art

[0002] In crop cultivation, a variety of operations occur, and the content and timing of each operation affect the cultivation results. In general, the determination of such operation content and timing is often based on experience and intuition. Also, as in Patent Document 1 below, a technique for acquiring plant growth management information using sensors is also known.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, stable cultivation results may not be obtained by experience and intuition alone, and it is not easy to pass on experience and intuition to new cultivators. Also, even if the cultivation state can be grasped by sensors, it is not easy to determine appropriate operation content according to the cultivation state. Therefore, there is a need for a technique to support crop cultivation so that experienced cultivators can grow crops more stably and so that even inexperienced cultivators can grow crops appropriately.

[0005] One aspect of the present invention has been made in view of the above problems, and an example of its object is to provide a technique for supporting crop cultivation.

Means for Solving the Problems

[0006] A crop cultivation support device according to one aspect of the present invention comprises: a receiving means for receiving a request that includes any of the size, taste, harvest time, or yield of a crop to be cultivated; a trained model that has learned the relationship between a plurality of crop cultivation methods and cultivation results that include any of the size, taste, harvest time, or yield of the plurality of crops; a generating means for generating response information that includes a cultivation method for the crop to be cultivated based on the request; and an output means for outputting the cultivation method.

[0007] In a method for supporting crop cultivation according to one aspect of the present invention, a computer receives a request that includes the size, taste, harvest time, and yield of the crop to be cultivated. Based on the request, the computer generates response information that includes the cultivation method of the crop to be cultivated, a trained model that has learned the relationship between cultivation methods for multiple crops and cultivation results including the size, taste, harvest time, and yield of the multiple crops, and outputs the cultivation method.

[0008] A program relating to one aspect of the present invention causes a computer to perform the following processes: receiving a request that includes the size, taste, harvest time, and yield of a crop to be grown; generating response information that includes the cultivation method of the crop to be grown based on a trained model that has learned the relationship between a plurality of crop cultivation methods and cultivation results that include the size, taste, harvest time, and yield of the plurality of crops; and outputting the cultivation method. [Effects of the Invention]

[0009] According to one aspect of the present invention, it is possible to support the cultivation of plants and animals. [Brief explanation of the drawing]

[0010] [Figure 1] This is a block diagram showing the configuration of a crop cultivation support device according to a first exemplary embodiment of the present invention. [Figure 2] This is a flowchart showing the flow of a crop cultivation support method according to the first exemplary embodiment of the present invention. [Figure 3] This diagram illustrates feature learning in graph-based relational learning. [Figure 4] This figure shows an overview of a crop cultivation support method according to a second exemplary embodiment of the present invention. [Figure 5] This is a block diagram showing the configuration of a crop cultivation support device according to a second exemplary embodiment of the present invention. [Figure 6] This is a flowchart showing the processing flow performed by a crop cultivation support device according to a second exemplary embodiment of the present invention. [Figure 7] This figure shows an example of response information. [Figure 8] This figure shows an overview of a crop cultivation support method according to a third exemplary embodiment of the present invention. [Figure 9] This is a block diagram showing the configuration of a crop cultivation support device according to a third exemplary embodiment of the present invention. [Figure 10] This is a flowchart showing the processing flow performed by a crop cultivation support device according to a third exemplary embodiment of the present invention. [Figure 11] This figure shows an overview of a crop cultivation support method according to a fourth exemplary embodiment of the present invention. [Figure 12] This is a block diagram showing the configuration of a crop cultivation support device according to a fourth exemplary embodiment of the present invention. [Figure 13] This is a flowchart showing the processing flow performed by the crop cultivation support device according to the fourth exemplary embodiment of the present invention. [Figure 14] This figure shows an overview of a crop cultivation support method according to a fifth exemplary embodiment of the present invention. [Figure 15] This is a block diagram showing the configuration of a crop cultivation support device according to a fifth exemplary embodiment of the present invention. [Figure 16] This is a flowchart showing the processing flow performed by a crop cultivation support device according to a fifth exemplary embodiment of the present invention. [Figure 17] This diagram illustrates an example of predicting the training results of a target based on features calculated from the training target graph and the training completion graph. [Figure 18]It is a configuration diagram for realizing a crop cultivation support device by software.

Embodiments for Carrying out the Invention

[0011] 〔Exemplary Embodiment 1〕 The first exemplary embodiment of the present invention will be described in detail with reference to the drawings. This exemplary embodiment is a basic form of the exemplary embodiments described later.

[0012] (Crop Cultivation Support Device) The configuration of the crop cultivation support device 1 according to this exemplary embodiment will be described with reference to FIG. 1. FIG. 1 is a block diagram showing the configuration of the crop cultivation support device 1. As shown in the figure, the crop cultivation support device 1 includes a reception unit (reception means) 11, a generation unit (generation means) 12, and an output unit (output means) 13.

[0013] The reception unit 11 receives a request including any one of the size, taste, harvest time, and harvest amount of the crop to be cultivated. The generation unit 12 generates response information including the cultivation method of the crop to be cultivated based on a learned model that has learned the relationship between the cultivation methods of a plurality of crops and the cultivation results including any one of the size, taste, harvest time, and harvest amount of the plurality of crops, and the request. The output unit 13 outputs the above response information.

[0014] According to the crop cultivation support device 1 having the above configuration, a request regarding the crop to be cultivated is received. Then, based on a learned model that has learned the relationship between the cultivation methods of a plurality of crops and the cultivation results including any one of the size, taste, harvest time, and harvest amount of the plurality of crops, and the request, response information including the cultivation method of the crop to be cultivated is generated.

[0015] Thereby, various information regarding the crops cultivated in the past can be considered, and response information useful for cultivating the crop to be cultivated can be generated. Therefore, according to the above configuration, an effect that the cultivation of crops can be supported can be obtained.

[0016] (program) The functions of the crop cultivation support device 1 described above can also be realized by a program. The program according to this exemplary embodiment causes a computer to perform the following processes: receiving a request that includes the size, taste, harvest time, and yield of the crop to be cultivated; generating response information that includes the cultivation method of the crop to be cultivated based on a trained model that has learned the relationship between cultivation methods for multiple crops and cultivation results that include the size, taste, harvest time, and yield of the multiple crops, and the request; and outputting the cultivation method. This program provides the effect of supporting crop cultivation.

[0017] (Methods for supporting crop cultivation) The crop cultivation support method according to this exemplary embodiment will be described with reference to Figure 2. Figure 2 is a flowchart showing the flow of the crop cultivation support method according to the first exemplary embodiment of the present invention.

[0018] In S11, the computer receives a request that includes one of the following: the size, taste, harvest time, or yield of the crop to be grown. The request can be received via any input device. For example, it may be received via a mouse, keyboard, touch panel, or voice input device.

[0019] In S12, the computer generates response information, including the cultivation method for the target crop, based on a trained model that has learned the relationship between cultivation methods for multiple crops and cultivation results, including size, taste, harvest time, and yield of multiple crops, and the request.

[0020] In S13, the computer outputs the response information generated in S12. The output device is arbitrary; for example, it may be output to a display device to show the information, or it may be output to an audio output device to output the information as audio.

[0021] As described above, in the crop cultivation support method according to this exemplary embodiment, the computer receives a request that includes the size, taste, harvest time, and yield of the crop to be cultivated (S11), generates response information that includes the cultivation method of the crop to be cultivated based on a trained model that has learned the relationship between cultivation methods for multiple crops and cultivation results that include the size, taste, harvest time, and yield of multiple crops, and the request received in S11 (S12), and outputs the response information generated in S12. This crop cultivation support method has the effect of being able to support crop cultivation.

[0022] Furthermore, the entity executing each step in the above-described crop cultivation support method may be a single computer (for example, crop cultivation support device 1), or each step may be executed by a different computer. This also applies to the flow described in Exemplary Embodiment 2 and subsequent examples.

[0023] [Graphs and Learning] In the following, we will describe graphs, which are an example of information that can be used to support crop cultivation in Exemplary Embodiment 1 and the subsequent Exemplary Embodiments (hereinafter referred to as each Exemplary Embodiment). We will also explain how to learn these graphs and how to make predictions using them.

[0024] (graph) In this context, a graph refers to data with a structure consisting of multiple nodes and links connecting those nodes. The type of link that represents the relationship between nodes is also called a "relation." Links are sometimes also called edges. Graphs can be broadly classified into directed graphs, where each link has a direction, and undirected graphs, where each link does not have a direction. It is possible to use either directed or undirected graphs, and it is also possible to use them in combination.

[0025] In each exemplary embodiment, when using a graph, the nodes may represent tangible or intangible elements relating to the crop being cultivated or the crop that has been cultivated. For example, • Identification information for agricultural products (e.g., name, ID, variety, etc.) • Growth status ·Growth environment • Types and content of work • Harvest time and yield A graph containing nodes representing various elements such as these can be used. Here, the growth state may include, but is not limited to, the condition of the leaves (color, size, shape) and the condition of the fruit (color, size, shape), etc. Also, the growth state of a certain crop may include the quality of that crop (aroma, taste (sugar content and acidity)), etc. Also, the growth environment may include, but is not limited to, temperature (room temperature in the case of greenhouse cultivation), humidity, solar radiation intensity, etc.

[0026] Furthermore, the types and content of work may include, for example, fertilization, watering, intertillage, and fruit thinning, as well as the amount of fertilizer applied, the timing of fertilization, and the frequency of watering. The types and content of work may also include operations to adjust solar radiation intensity, sunlight duration, room temperature, and humidity. Such operations constitute a part of the cultivation method.

[0027] Note that a graph may contain multiple nodes corresponding to a single element. For example, a node representing the growing environment of a certain crop may be represented by two separate nodes (e.g., "high temperature" and "high humidity"). The same applies to other elements.

[0028] When nodes exist as elements as described above, the relationships represented by links are: • The relationship between certain elements and growth conditions • The relationship between certain elements and the growing environment • The relationship between a certain element and the type or content of the work. This would represent things like that. For example, a link connecting a node representing the growing environment and a node representing the growing state may represent a relationship where the growing environment is a factor in the growing state.

[0029] For example, the graphs in each exemplary embodiment are: A graph including a crop ID node showing a crop ID, and nodes connected to that crop ID node showing the growth status or growing environment, and • A work ID node indicating a work ID, a crop ID node connected to the work ID node, and nodes indicating various work details connected to the work ID node. The structure may have a hierarchical structure, as shown above, but this does not limit the examples of the embodiments.

[0030] Furthermore, the harvesting period and the timing of various tasks mentioned above may or may not be represented by a single node. For example, the harvesting period may be represented hierarchically by multiple nodes indicating tasks and fruit thinning, and links connecting these nodes that indicate the temporal relationships between them. Similarly, the harvest yield may be represented hierarchically by using nodes and links indicating the first harvest yield at the first timing, and nodes and links indicating the second harvest yield at the second timing, for example.

[0031] (Learning and prediction) For graphs like those described above, machine learning techniques can be applied to perform graph-based relationship learning. This type of learning makes it possible to perform classification and prediction processes using graphs. In each exemplary embodiment, such learning may be performed as part of supporting crop cultivation, or a pre-trained graph that has already undergone such learning may be used.

[0032] In graph-based relational learning, the first step is to calculate the features of each node. These features can be represented, for example, as vectors. By representing the features of each node as feature vectors, it becomes possible to learn even graphs containing a mixture of nodes in various forms. For example, graph-based relational learning can be performed on graphs that include images, numerical values, and other elements representing the various elements mentioned above.

[0033] Next, the features of each node are updated based on the links connected to each node and the nodes to which those links are connected. This process is similar to the convolution process in a convolutional neural network. This will be explained with reference to Figure 3. Figure 3 is a diagram illustrating feature learning in graph-based relational learning.

[0034] The graph shown in Figure 3 contains four nodes, A through D. Node A is connected to nodes B and C, and node C is connected to node D. After calculating the initial features of these four nodes, the features of each node are updated by performing multiple convolutions as described below.

[0035] In the first convolution, the initial features of node A are added to the features of nodes B and C, which are connected to node A, after multiplying them by predetermined weights. Similarly, for node C, the initial features of node C are added to the features of node D, after multiplying them by predetermined weights. If the graph is valid, the weights are adjusted according to the direction of the links.

[0036] In the second convolution, just like in the first convolution, the features of each node are added to the features of the nodes linked to that node, multiplied by a predetermined weight. Here, the features of node C reflect the features of node D from the first convolution. Therefore, after the second convolution, node A reflects not only the features of node C but also the features of node D.

[0037] By repeating the above process a number of times corresponding to the node hierarchy, the features of each node connected directly or indirectly by links are reflected in each other. In graph-based relational learning, the weight values ​​used for the weighting described above are optimized based on the known relationships between nodes. By using such a trained graph (which can also be called a trained model), it becomes possible to predict inter-node relationships and linked nodes, as described below.

[0038] (Predicting inter-node relationships) By performing the learning described above, it becomes possible to predict relationships between nodes that are not explicitly shown in the original graph. To perform node relationship prediction, the user simply needs to specify two nodes and request that the relationship between those nodes be returned. For example, if a user requests the relationship between the node for "Crop A" and the node for "Crop B," node relationship prediction can predict whether the relationship, or link, connecting these nodes is "similar." Furthermore, node relationship prediction can also calculate the probability (likelihood) of the prediction result. The same applies to node prediction, which will be explained below.

[0039] (Node prediction) Furthermore, by performing the learning described above, it becomes possible to predict which nodes are connected to a given node by a predetermined link. To perform node prediction, the user simply needs to specify one node and the link that starts from that node, and request that the system return the linked node. For example, suppose a user submits a request asking for the node connected to the "Work History" node by the "September" link, that is, the node whose work should be done in September. In this case, node prediction can predict whether the node connected to the "Work History" node by the "September" link, i.e., the work to be done in September, is "Task a3" or "Task a4," etc.

[0040] [Exemplary Embodiment 2] (overview) Figure 4 shows an overview of the crop cultivation support method according to this exemplary embodiment. In this exemplary embodiment, an example of supporting crop cultivation using a target cultivation graph and a cultivated crop graph is described.

[0041] A cultivated graph is a graph and model that has been trained on the relationships between nodes, containing multiple nodes related to crops that have been cultivated in the past and links indicating the relationships between those nodes. A cultivated graph can also be called a knowledge graph. Note that a set of nodes and links corresponding to one cultivation cycle may be called a cultivated graph, or a set of nodes and links corresponding to multiple cultivation cycles may be called a cultivated graph as a whole.

[0042] For example, in Figure 4, the graph containing the node for "Crop A" is the cultivated graph. The cultivated graph for Crop A includes nodes and links indicating that the quality of the harvested crop at the time of cultivation was "high sugar content" and "large size." The cultivated graph for Crop A also includes nodes and links indicating that the growing environment at the time of cultivation was "average" and that the variety of Crop A was "a1." Furthermore, the cultivated graph for Crop A includes nodes and links showing the work history for each month from August to October.

[0043] Note that "a1" is the variety name, and "a3" to "a5" indicate the work. Here, "work" can include the type and content of the work. As mentioned above, for example, fertilization, watering, intertillage, fruit thinning, etc. are included in the category of "work," as are the details of the work, such as the amount of fertilizer to be applied, the timing of fertilization, and the frequency of watering. Such work constitutes a part of the cultivation method.

[0044] By learning the relationship between the cultivation method and the cultivation results of crop A, such a cultivated graph can be generated. Although not shown in Figure 4, a cultivated graph including nodes and links related to the cultivation method and cultivation results for crop B has also been generated, similar to crop A. In this way, multiple cultivated graphs are pre-generated.

[0045] Furthermore, growth status and growing environment may be represented by the difference from a predetermined standard state. For example, the difference or ratio between the total number of leaves of a crop during a certain growing period and the standard total number of leaves during that period may be represented as a node indicating the growth status of the crop during that period. Alternatively, the growth status may be classified into categories such as good, normal, poor, etc., based on the difference or ratio calculated in this way, and this classification may be represented as a node. Similarly, regarding the growing environment, the difference or ratio between the amount of sunlight during a certain growing period of a crop and the standard amount of sunlight during that period may be represented as a node indicating the growing environment during that period. Alternatively, the growing environment may be classified into categories such as dry, average, wet, etc., based on the difference or ratio calculated in this way, and this classification may be represented as a node.

[0046] The target crop graph is a graph containing multiple nodes related to the target crop. The target crop can be a crop to be cultivated in the future, or a crop that is currently being cultivated. In Figure 4, the graph containing the node "Cultivated Crops" is the target crop graph. This target crop graph includes nodes and links indicating that the crop's growth status up to this point is "Standard," the growing environment is "High Temperature," and the variety is "x1," as well as nodes and links indicating that the crop's work history for August is "x2." Such a target crop graph can be generated, for example, by receiving input of necessary information from the crop's grower.

[0047] By using the cultivated crop graph and the crop to be cultivated graph as described above, it becomes possible to predict which cultivation method is suitable for the crop. In other words, in the crop cultivation support method according to this exemplary embodiment, the cultivation method of the crop is predicted by link prediction, and response information is generated and output based on the prediction result.

[0048] For example, in the example in Figure 4, we can predict which of the various nodes representing the work content included in the cultivated graph are likely to be linked to the node included in the target cultivation graph (more specifically, the "September" node that links to "Work History"). Then, we can generate and output response information indicating that the predicted work is a suitable cultivation method for the crop.

[0049] (Device configuration) The configuration of the crop cultivation support device 2 according to a second exemplary embodiment of the present invention will be described with reference to Figure 5. Figure 5 is a block diagram showing the configuration of the crop cultivation support device 2 according to this exemplary embodiment.

[0050] As shown in the figure, the crop cultivation support device 2 includes a reception unit 201, a graph generation unit 202, a learning unit 203, a link prediction unit 204, an evaluation unit 205, a generation unit 206, a basis generation unit 207, and an output unit 208.

[0051] In addition to these components, the crop cultivation support device 2 may also include an input device for receiving user input, an output device for data output by the crop cultivation support device 2, a communication device for the crop cultivation support device 2 to communicate with other devices, etc. The output mode of the output device is arbitrary and may be, for example, a display output or an audio output.

[0052] The reception unit 201 accepts requests regarding the crop to be cultivated (corresponding to the cultivated crops mentioned above; hereinafter sometimes simply referred to as the "cultivated crop"). For example, the reception unit 201 accepts requests that include any of the following: size, taste, harvest time, and yield of the cultivated crop (more precisely, the harvest obtained from cultivating the cultivated crop). The reception unit 201 may also accept requests for information indicating the characteristics of the cultivated crop, such as the variety of the cultivated crop. Furthermore, if the cultivated crop is still being cultivated, the reception unit 201 may accept requests for information such as the growth status, growing environment, and work history of the cultivated crop.

[0053] The graph generation unit 202 generates a crop target graph that represents the crops based on information about the crops to be grown. For example, the graph generation unit 202 may represent the crops as nodes and link nodes indicating the properties, growth status, growth environment, and work history of the crops to these nodes to generate the crop target graph. Information regarding the properties of the crops may be included in the requests received by the reception unit 201, or it may be obtained from a database that aggregates the properties of various crops. The growth environment may be identified by analyzing a database that aggregates weather and weather information for the cultivation area.

[0054] The learning unit 203 learns the relationships between each node in the cultivated graph, in other words, the relationship between crop cultivation methods and cultivation results, based on various information about crops that have been cultivated in the past, and generates a trained cultivated graph. Cultivation results include the size, taste, harvest time, and yield of the crop (more precisely, the harvested product). Unless otherwise specified, the cultivated graph refers to one that has been trained by the learning unit 203. Alternatively, the trained cultivated graph may be loaded into the crop cultivation support device 2, in which case the learning unit 203 may be omitted.

[0055] The link prediction unit 204 uses the aforementioned target crop graph and cultivated crop graph to predict the relationships between nodes that are not connected by links in the target crop graph and cultivated crop graph. It then predicts nodes that link to nodes in the target crop graph from among the nodes in the cultivated crop graph that relate to tasks performed during the cultivation of crops that were previously cultivated. The tasks indicated by these nodes are candidates for cultivation methods. For example, in the example in Figure 4, the link prediction unit 204 predicts tasks that link to the "Task History" node in the target crop graph with a "September" link, and these predicted tasks become candidates for cultivation methods.

[0056] The evaluation unit 205 evaluates the recommendation level of the node predicted by the link prediction unit 204, i.e., the candidate cultivation method, based on other nodes included in the cultivated graph that include the node predicted by the link prediction unit 204. For example, in the example in Figure 4, suppose the task linked to the "Work History" node by the "September" link is predicted to be "a4". In this case, the evaluation unit 205 evaluates the recommendation level of task "a4" based on other nodes included in the cultivated graph of crop A (e.g., "High Sugar Content"). The evaluation method will be described later.

[0057] The generation unit 206 generates response information, including the cultivation method for the target crop, based on a trained model that has learned the relationship between cultivation methods and cultivation results for multiple crops, and the request received by the reception unit 201. More specifically, the generation unit 206 generates response information corresponding to the node predicted by the link prediction unit 204, i.e., the candidate cultivation method. This node indicates the work that should be applied to the cultivation of the target crop. As described above, the link prediction unit 204 performs link prediction using the cultivated graph, so the generation unit 206 generates response information based on the trained model by generating response information based on the link prediction results from the link prediction unit 204.

[0058] The evidence generation unit 207 generates evidence information that demonstrates the validity of the response information generated by the generation unit 206. The method for generating the evidence information will be described later.

[0059] The output unit 208 outputs various types of information generated by the crop cultivation support device 2. For example, the output unit 208 outputs response information generated by the generation unit 206 and rationale information indicated by the rationale generation unit 207. The destination of the information output is arbitrary; for example, if the crop cultivation support device 2 has an output device as described above, the information can be output to that output device. Alternatively, for example, the information may be output to an external output device of the crop cultivation support device 2.

[0060] As described above, the trained model used by the crop cultivation support device 2 may include multiple nodes related to crops cultivated in the past and links indicating the relationships between those nodes, and may also be a trained graph that has learned about the relationships between nodes. With this configuration, it is possible to generate and output response information indicating an appropriate cultivation method for the target crop, taking into account the cultivation methods and results of crops cultivated in the past and their interrelationships.

[0061] As described above, the crop cultivation support device 2 may include a link prediction unit 204 that uses the cultivation target graph and the cultivated graph to predict the relationships between nodes that are not connected by links in the cultivation target graph and the cultivated graph, and predicts nodes that link to nodes in the cultivation target graph from among the nodes related to operations performed during the cultivation of crops that have been cultivated in the past, which are included in the cultivated graph. The generation unit 206 may then generate response information corresponding to the nodes predicted by the link prediction unit 204.

[0062] Nodes that link to nodes in the target crop graph, related to tasks performed during the cultivation of previously grown crops, may also be related to tasks performed during the cultivation of the target crop. For example, suppose the cultivation graph of a previously grown crop with good results includes a node indicating that the frequency of watering was reduced, and this node is predicted to link to a node in the target crop graph. In this case, it can be expected that reducing the frequency of watering will also lead to good results for the target crop. Therefore, the above configuration can provide useful information regarding tasks performed on the target crop.

[0063] (Regarding link prediction) The link prediction unit 204 can also predict nodes that meet specified conditions. The conditions may be specified in advance, or they may be made available to the user. In the latter case, the reception unit 201 may accept the input of conditions as a request.

[0064] For example, if the reception unit 201 receives input for conditions regarding a cultivated graph, the link prediction unit 204 may predict nodes that link to nodes included in the target cultivated graph from among the nodes related to operations performed during the cultivation of crops that have been cultivated in the past, which are included in the cultivated graph that satisfies those conditions.

[0065] This allows for the prediction of nodes that align with the user's intent. For example, if a user inputs the condition "a predetermined training result was obtained," the system will predict nodes from the trained graph that indicate such a training result was achieved, linking to nodes included in the target training graph related to the tasks performed during training. In this case, the system can provide useful information about tasks that are highly likely to lead to the predetermined training result.

[0066] In addition to the above, you may also set conditions such as those listed below. To determine whether or not a condition is met, a trained graph that has learned information related to the fulfillment of the condition is used. For example, when applying the condition that the work cost is low, a trained graph that includes nodes and links indicating the work cost is used. • The user has achieved their set growth goals (which can be final or intermediate goals). • The prerequisites, such as crops, growing environment, and available materials and equipment, match (exact or partial match). • The growing environment after the work shown in the cultivation history matches the weather forecast (for example, if high temperatures are forecast for September, extract the work from the graph of crops with a cultivation history of high temperatures in September). • The cost of the work (working time, labor intensity, expenses, etc.) is low. • It is organically grown, or the use of pesticides is kept to a minimum. • The yield is within the specified range at the specified time.

[0067] (Node evaluation) As described above, the evaluation unit 205 evaluates the recommendation level of the node predicted by the link prediction unit 204 based on other nodes included in the cultivated graph that include the node predicted by the link prediction unit 204. The evaluation by the evaluation unit 205 is described below.

[0068] Nodes that link to nodes in the target crop graph may provide users with useful suggestions regarding appropriate tasks for the target crop. Furthermore, other nodes in the completed crop graph that contain those nodes may be relevant to the cultivation of the target crop. For example, suppose the completed crop graph contains nodes and links indicating that the cultivated crop had a high yield. In this case, the nodes in the completed crop graph that link to nodes in the target crop graph may contain information that contributes to improving the yield of the target crop, and therefore, those nodes can be said to have a high recommendation level.

[0069] Therefore, with the above configuration, the recommendation level for the node predicted by the link prediction unit 204 is evaluated based on other nodes included in the cultivated graph that include the node predicted by the link prediction unit 204. The user can then decide on the work to be done on the target of cultivation based on this evaluation. This can contribute to the determination of work that is expected to yield more favorable cultivation results.

[0070] The evaluation method can be predetermined based on the target node, and various criteria can be applied to the evaluation. For example, the evaluation may be based on the degree of suitability to the request. For example, suppose the request includes the quality required for the harvested product to be cultivated. In this case, the evaluation unit 205 may ensure that the evaluation of the node corresponding to the cultivated graph that includes the node indicating that quality is higher than the evaluation of the node corresponding to the cultivated graph that does not include the node indicating that quality.

[0071] Furthermore, the evaluation unit 205 may express the evaluation results numerically. In this exemplary embodiment, an example is described in which the evaluation unit 205 calculates the recommendation score of the node predicted by the link prediction unit 204. In this case, if the relationship between the nodes included in the cultivated graph and the recommendation score is predetermined as a rule, the evaluation unit 205 can calculate the recommendation score of each node according to that rule.

[0072] For example, the evaluation unit 205 may calculate the recommendation score for the cultivated graph that includes the nodes predicted by the link prediction unit 204, using at least one of the following criteria. • Points are added if the node and links demonstrate that the user has achieved their set growth goals; points are deducted if they are not. Points are added for nodes and links that show a high degree of agreement in prerequisites such as crop species, varieties, growing environment, and available materials and equipment; points are deducted for nodes and links that show a low degree of agreement. • Points are added for nodes and links that indicate similar growing environments, and points are deducted for nodes and links that indicate similar growing environments (Note: For crops currently being grown, weather forecasts may be used to determine the similarity of growing environments). - Points will be deducted if the node or link indicates a high cost of work (work time, labor intensity, expenses, etc.), and points will be added if there are fewer such nodes or links. • Points will be deducted if nodes and links indicating high pesticide use are included, and points will be added if such nodes and links are included. Points are added if the node and links indicate that the user-specified yield was within the specified range at the specified time; points are deducted if the node and links indicate a discrepancy between the time and yield.

[0073] (Method for generating supporting information) As described above, the evidence generation unit 207 generates evidence information that indicates the validity of the response information generated by the generation unit 206. For example, the evidence generation unit 207 may generate evidence information that includes past cases similar to the training method of the target of training. This allows the user to refer to the response information based on the evidence information and accurately judge the validity of the response information.

[0074] For example, the basis generation unit 207 may use all or part of a cultivated graph that includes the nodes predicted by the link prediction unit 204 as basis information. Alternatively, for example, the basis generation unit 207 may search for a cultivated graph that includes the nodes predicted by the link prediction unit 204 and contains a predetermined number or more nodes that are common to the cultivation method to be cultivated, and use all or part of the cultivated graph detected by this search as basis information.

[0075] (Regarding the generation of justification for the results of link prediction) The evidence generation unit 207 can also generate evidence information by analyzing the training target graph and the completed training graph. The method for generating evidence information by analyzing the training target graph and the completed training graph will be described below.

[0076] For example, the rationale generation unit 207 may use PCA (Principal Component Analysis) confidence scores based on the OWA (Open-world assumption) to mine one or more rules from the target graph and the cultivated graph. The rationale generation unit 207 may then generate rationale information using the one or more rules that it has mined. For rule mining, methods described in the following literature, for example, can also be applied.

[0077] Luis Galarraga et. al, “Fast rule mining in ontological knowledge bases with AMIE +”, The VLDB Journal(2015)24:707-730 As an example, the rules to be processed by the basis generation unit 207 are defined using Head r(x, y) and Body { B1 , . . . , Bn},

number

number

[0078] The evidence generation unit 207 sets the mining process conditions as follows: • Connected: All values ​​(variables, entities) within the rule are shared across different atoms. • Closed: All variables within the rule must appear at least twice. • Not reflexive: Rules containing recursive (reflective) atoms, such as r(x, x), will not be mined. The mining process is performed under the following conditions.

[0079] Furthermore, the basis generation unit 207 is

number

number

[0080] For example, suppose the rationale generation unit 207 has mined the rule that "if the growing environment immediately before harvest is normal, reducing the frequency of watering immediately before harvest will result in the harvest of produce with a high sugar content." In this case, when the link prediction unit 204 predicts that the task of reducing the frequency of watering should be performed, the rationale generation unit 207 should generate rationale information that shows the above rule as the basis for this prediction.

[0081] (Process flow) The flow of processing (crop cultivation support method) performed by the crop cultivation support device 2 will be explained based on Figure 6. Figure 6 is a flowchart showing the flow of processing performed by the crop cultivation support device 2.

[0082] In S201, the reception unit 201 receives requests regarding the crop to be grown. For example, in S201, the reception unit 201 receives requests that include one of the following: size, taste, harvest time, or yield of the crop to be grown (more precisely, the harvest obtained from growing the crop). Subsequently, in S202, the graph generation unit 202 generates a graph of the crop to be grown based on the information entered in S201.

[0083] In S203, the link prediction unit 204 determines candidate cultivation methods. Specifically, the link prediction unit 204 uses the cultivation target graph generated in S202 and the cultivated graph to predict nodes that link to nodes in the cultivation target graph from among the nodes in the cultivated graph that relate to operations performed during the cultivation of crops that have been cultivated in the past. The operations shown in these nodes are candidate cultivation methods. In conjunction with the processing in S203, the rationale generation unit 207 may generate rationale information that shows the basis for the prediction results of the link prediction unit 204 by analyzing the cultivation target graph and the cultivated graph.

[0084] In S204, the evaluation unit 205 evaluates the candidate breeding methods determined in S203. Specifically, the evaluation unit 205 evaluates the recommendation level of the candidate breeding method based on the nodes included in the breeding graph related to the candidate breeding method. If multiple candidates were determined in S203, the evaluation unit 205 evaluates each of the determined candidates.

[0085] In S205, the generation unit 206 generates response information based on the candidates determined in S203 and the request received in S201. As mentioned above, the trained graph is a trained model. The candidate training method is determined by link prediction using the trained graph. Therefore, it can be said that in S205, response information is generated based on the trained model and the request received in S201.

[0086] For example, the generation unit 206 may generate response information indicating candidates up to a predetermined rank from among the candidates determined in S203, based on the evaluation result in S204. Alternatively, for example, the generation unit 206 may generate response information indicating a suitable alternative from among the candidates determined in S203 that matches the request received in S201. In addition, for example, the generation unit 206 may generate response information indicating the candidates determined in S203 and the evaluation result in S204.

[0087] In S206, the evidence generation unit 207 generates evidence information that demonstrates the validity of the response information generated in S205. For example, the evidence generation unit 207 may detect past breeding cases similar to the breeding method of the target breeding from the breeding graph and generate evidence information that includes the detected breeding cases.

[0088] In S207, the output unit 208 outputs the response information generated in S206. The output unit 208 may also output the rationale information generated in S206 at this time. This completes the process shown in Figure 6.

[0089] (Example of response information) S207 may output response information as shown in Figure 7, for example. Figure 7 is a diagram showing an example of response information. The response information shown in Figure 7 includes a total of seven items: the "candidate" breeding method to be cultivated, the "target variety" to which the breeding method is applied, the "growing environment," "breeding results," "breeding cost," "harvest time," and "recommendation level" of the crop grown using the breeding method.

[0090] "Candidates" are those predicted by the link prediction unit 204. In the example in Figure 7, tasks a to c are included in the candidates. Tasks a to c may indicate the type of task, the content of the task, or both the type and content of the task.

[0091] The "Target Variety" to "Harvest Time" is identified from the candidate cultivation graph predicted by the link prediction unit 204. In the example in Figure 7, "Cultivation Cost" is shown as labor time and material costs. In addition, other factors such as the labor intensity of the "candidate" tasks may also be included in "Cultivation Cost". The basis generation unit 207 may generate basis information that includes such information.

[0092] The "Recommendation Level" indicates the evaluation result of the evaluation unit 205 for the candidate. The evaluation unit 205 only needs to calculate the recommendation level based on various information identified from the cultivated graph. In the example in Figure 7, the recommendation levels for tasks a to c are 15, 5, and 0, respectively. For example, rules may be predetermined such as: a recommendation level of +5 if the "Target Variety" and "Growing Environment" match the cultivation target; a recommendation level of +5 if one of the requested cultivation results is obtained; a recommendation level of +5 if the cultivation cost is below the requested upper limit; and a recommendation level of +5 if the harvest time matches the requested time. This allows the evaluation unit 205 to calculate the recommendation level for each candidate as shown in Figure 7 according to those rules.

[0093] [Exemplary Embodiment 3] (overview) Figure 8 shows an overview of the crop cultivation support method according to this exemplary embodiment. In this exemplary embodiment, an example is described in which a cultivation method for a target crop that matches a request is searched for while updating a target crop graph that includes multiple nodes related to the target crop.

[0094] In this exemplary embodiment, link prediction is performed using the crop to be grown graph and the grown-up graph, similar to exemplary embodiment 2. The crop to be grown graph shown in the upper left of Figure 8 includes nodes and links indicating that "task y1" was performed in September, regarding the management status of the cultivated crop up to this point.

[0095] Furthermore, Figure 8 shows cultivated graphs that include the node for "Crop A" and cultivated graphs that include the node for "Crop B," but other nodes and links are omitted from the illustration.

[0096] By learning from various cultivated graphs as described above, it becomes possible to predict which cultivation methods are likely to yield which cultivation results. In other words, in the crop cultivation support method according to this exemplary embodiment, a provisional cultivation target graph is generated, and the probability that the cultivation target shown in that graph will have the requested cultivation result is predicted.

[0097] For example, in the example shown in Figure 8, the probability that the node for "Cultivated Crop" in the cultivation target graph shown in the upper left corner will be linked to the node for "High Sugar Content" via a "Quality" link is predicted to be 30%. This probability is not considered sufficiently high.

[0098] Therefore, as shown in the lower part of the figure, the node connected to the "Work History" node linked to the "Cultivated Crops" node in the growth target graph is changed from "Work y1" to "Work y2" via the "September" link, and the link prediction is performed again. As a result, the predicted probability of the "High Sugar Content" node being connected to the "Cultivated Crops" node via the "Quality" link changes to 80%.

[0099] According to the crop cultivation support method of this exemplary embodiment, based on the results of the above processing, "task y2" can be recommended as a September task to give the target crop the quality of "high sugar content".

[0100] (Device configuration) The configuration of the crop cultivation support device 3 according to the third exemplary embodiment of the present invention will be described with reference to Figure 9. Figure 9 is a block diagram showing the configuration of the crop cultivation support device 3 according to this exemplary embodiment.

[0101] As shown in the figure, the crop cultivation support device 3 includes a reception unit 301, a graph generation unit 302, a link prediction unit 303, a graph update unit 304, a generation unit 305, a basis generation unit 306, and an output unit 307. In addition to these components, the crop cultivation support device 3 may also include a learning unit, an input device, an output device, a communication device, etc., similar to the crop cultivation support device 2 of the exemplary embodiment 2.

[0102] The reception unit 301 accepts requests regarding the crop to be grown. For example, the reception unit 301 accepts requests that include any of the following: size, taste, harvest time, and yield of the crop to be grown (more precisely, the harvest obtained from growing the crop). The reception unit 301 may also accept input of at least one of the details and timing of the work to be performed on the crop. This input may be accepted as part of the request or as a separate input from the request. The details and timing of the work to be performed on the crop may include the details and timing of work that has been performed on the crop in the past, or the details and timing of work that will be performed on the crop in the future.

[0103] The graph generation unit 302 generates a cultivation target graph based on the above request. The cultivation target graph generated by the graph generation unit 302 includes nodes that indicate at least one of the content and timing of work, and nodes that indicate the content and timing of work to be performed on the cultivation target in the future. For example, as shown in Figure 8, the graph generation unit 302 may generate a cultivation target graph in which a node indicating "cultivated crop" is connected to a node indicating "work history" by a link indicating "management status", and "work y1" is connected to the node indicating "work history" by a link indicating "September". Here, the node for "work y1" connected to the node indicating "work history" by a link indicating "September" indicates the content of work to be performed on the cultivation target in the future. The cultivation target graph may also include nodes and links that indicate the size, taste, harvest time, and yield of the cultivation target (more precisely, the harvest obtained from cultivating the cultivation target).

[0104] The link prediction unit 303 uses the target graph generated by the graph generation unit 302 and the trained cultivated graph to perform link prediction, which predicts the relationships between nodes that are not connected by links in the target graph and the cultivated graph. This calculates the probability that a node indicating a predetermined cultivation result will be linked to the target graph. The predetermined cultivation result is identified based on a request, for example. For example, if the requested cultivation result for the target is "high sugar content," the link prediction unit 303 calculates the probability that a node indicating that quality will be linked to a node included in the target graph (for example, the "cultivated crop" node in the example in Figure 8).

[0105] The graph update unit 304 updates the training target graph. For example, the graph update unit 304 performs at least one of the following: replacing nodes in the training target graph that indicate future work to be performed on the training target with nodes of other work contents, and adding nodes of new work contents.

[0106] The update of the graph to be trained may be performed according to user input or automatically. In the former case, the graph update unit 304 may output a list of tasks extracted from the trained graph to the output unit 307 and allow the user to select a new task from that list. In the latter case, the graph update unit 304 can select a new task from the tasks extracted from the trained graph.

[0107] The generation unit 305 generates response information, including the cultivation method for the target crop, based on a trained model that has learned the relationship between cultivation methods for multiple crops and cultivation results, including size, taste, harvest time, and yield of the multiple crops, and the request. More specifically, the generation unit 305 generates response information based on the probability calculated by the link prediction unit 303. A specific example of the generation of response information will be described later based on Figure 10.

[0108] As described above, the link prediction unit 303 performs link prediction using the trained model graph and the training target graph generated based on the request. Therefore, the generation unit 305 generates response information based on the results of the link prediction by the link prediction unit 303, thereby generating response information based on the trained model and the above request.

[0109] The evidence generation unit 306 generates evidence information that demonstrates the validity of the response information generated by the generation unit 305. Specifically, the evidence generation unit 306 generates evidence information that includes past cases similar to the cultivation method of the target crop. The evidence generation unit 306 may also generate evidence information regarding the link prediction results by the link prediction unit 303 by analyzing the target crop graph and the cultivated crop graph.

[0110] The output unit 307 outputs various types of information generated by the crop cultivation support device 3. For example, the output unit 307 outputs response information generated by the generation unit 305 and rationale information indicated by the rationale generation unit 306. Similar to the output unit 208 of the exemplary embodiment 2, the destination of the information output is not particularly limited.

[0111] As described above, in the crop cultivation support device 3, the reception unit 301 receives input of at least one of the content and timing of the work to be performed on the crop to be cultivated, and the link prediction unit 303 uses the crop to be cultivated graph, which includes nodes that indicate at least one of the input content and timing of the work, and the cultivated graph to calculate the probability that a node indicating a predetermined cultivation result will be linked to the crop to be cultivated graph by link prediction, which predicts the relationship between nodes that are not connected by links in the crop to be cultivated graph and the cultivated graph.

[0112] A training target graph that includes nodes indicating at least one of the content and timing of the work performed on the training target can be said to represent the state of the training target after the work has been performed. Therefore, the probability that a node indicating a predetermined training result is linked to this training target graph can be said to represent the likelihood that the predetermined training result will be obtained by the work in question. In other words, with the above configuration, it is possible to predict whether a predetermined training result is likely to be obtained by the work that the trainer is about to perform, before the work is actually performed.

[0113] (Process flow) The flow of processing (crop cultivation support method) performed by the crop cultivation support device 3 will be explained based on Figure 10. Figure 10 is a flowchart showing the flow of processing performed by the crop cultivation support device 3.

[0114] In S301, the reception unit 301 receives requests regarding the crop to be grown. For example, the reception unit 301 receives requests that include any of the following: size, taste, harvest time, and yield of the crop to be grown (more precisely, the harvest obtained from growing the crop). The reception unit also receives input regarding at least one of the tasks to be performed on the crop and the timing of those tasks.

[0115] In S302, the graph generation unit 302 generates a training target graph based on the information entered in S301. For example, if S301 receives input for at least one of the content and timing of the work to be performed on the training target, the graph generation unit 302 should generate a training target graph that includes at least one node for the content and timing of the work to be performed on the training target.

[0116] In S303, the link prediction unit 303 calculates the probability that a node representing a predetermined training result that matches the request received in S301 is linked to a node included in the training target graph generated in S302. As described above, this probability is calculated by link prediction using the trained training graph and the training target graph. In conjunction with the processing in S303, the basis generation unit 306 may generate basis information that shows the basis for the calculation result of the link prediction unit 303 by analyzing the training target graph and the training graph.

[0117] In S304, the graph update unit 304 determines whether the probability calculated in S303 is above a threshold. If it is determined to be above the threshold (YES in S304), the process proceeds to S306; if it is determined to be below the threshold (NO in S304), the process proceeds to S305.

[0118] If the request received in S301 includes multiple training results, S303 will make a prediction for each training result, and in S304, if the probability of all training results is above a threshold, the result should be YES, and if even one training result is below the threshold, the result should be NO. This allows for the estimation of a training method that can satisfy all of the requested training results.

[0119] In S305, the graph update unit 304 updates the target graph. For example, the graph update unit 304 replaces nodes in the current target graph that represent tasks to be performed on the target in the future with nodes representing different tasks. As mentioned above, the update content may be determined according to user input or determined by the graph update unit 304.

[0120] When the training target graph is updated, the process returns to S303, and the probability is calculated again. In other words, in the process shown in Figure 10, the probability calculation in S303 and the updating of the training target graph in S305 are repeated until YES is determined in S304.

[0121] In S306, the generation unit 305 estimates a breeding method to obtain a predetermined breeding result that conforms to the request received in S301, and generates response information including the estimated breeding method. Specifically, the generation unit 305 estimates that the breeding method shown in the breeding target graph when it was determined to be YES in S304 is a breeding result that conforms to the request, and generates response information indicating that breeding method.

[0122] In S307, the evidence generation unit 306 generates evidence information that demonstrates the validity of the response information generated in S306. Specifically, the evidence generation unit 306 generates evidence information that includes past cases similar to the cultivation method of the target crop.

[0123] In S308, the output unit 307 outputs the response information generated in S306. The output unit 307 may also output the rationale information generated in S307 at this time. This completes the process shown in Figure 10.

[0124] (Additional notes regarding exemplary embodiment 3) The method of generating the cultivation target graph according to this exemplary embodiment is not limited to the example described above. For example, the crop cultivation support device 3 according to this exemplary embodiment may have a configuration similar to that of the link prediction unit 204 according to exemplary embodiment 2. In this case, the graph generation unit 302 according to this exemplary embodiment may generate a cultivation target graph that includes nodes indicating the content and timing of future operations to be performed on the cultivation target, which is at least a part of the cultivation method predicted by the link prediction unit 204.

[0125] Furthermore, in the above configuration, the graph update unit 304 according to this exemplary embodiment may replace nodes in the current target graph that represent tasks to be performed on the target in the future with nodes representing different tasks predicted by the link prediction unit 204. In the flow of Figure 10, before processing S305, the link prediction predicts the tasks or timing of tasks for the target, and the update in S305 is performed according to the prediction result.

[0126] [Exemplary Embodiment 4] (overview) Figure 11 shows an overview of the crop cultivation support method according to this exemplary embodiment. In this exemplary embodiment, an example is described in which a cultivation target graph including nodes indicating the desired cultivation result is used to predict the cultivation method for obtaining the desired cultivation result.

[0127] In this exemplary embodiment, link prediction is performed using a target growth graph and a grown-up graph, similar to exemplary embodiments 2 and 3. The target growth graph shown in the upper left of Figure 11 includes nodes and links indicating "high sugar content" as an example of a desired growth result for the cultivated crop.

[0128] Figure 8 also shows a completed crop graph for crop A, which includes a node labeled "Crop A," a node indicating "High Sugar Content" linked to that node via a "Quality" link, and a node for "Task A1" linked to the "Crop A" node via a "Task History" node and a "September" link. Similarly, a completed crop graph for crop B is shown, which includes a node labeled "Crop B," a node indicating "High Yield" linked to that node via a "Quality" link, and a node for "Task B1" linked to the "Crop B" node via a "Task History" node and a "September" link.

[0129] By learning from various cultivated graphs as described above, it becomes possible to predict what kind of cultivation results are likely to be obtained from what kind of cultivation method. As an example, in the crop cultivation support method according to this exemplary embodiment, as shown in Figure 11, a cultivation target graph is generated that includes a node indicating "high sugar content" as the quality desired by the user, and the work content (the "work" connected to the work history node with a "September" link in Figure 11) shown in the cultivation target graph is predicted to be linked to obtain the desired quality.

[0130] Thus, in the crop cultivation support method according to this exemplary embodiment, a cultivation target graph is generated that includes nodes indicating the cultivation results desired by the user, and the cultivation method necessary for the cultivation target shown in the cultivation target graph to obtain the desired cultivation results is linked and predicted. This provides the user with the information necessary to determine the tasks required to obtain the desired cultivation results.

[0131] (Device configuration) The configuration of the crop cultivation support device 4 according to the fourth exemplary embodiment of the present invention will be described with reference to Figure 12. Figure 12 is a block diagram showing the configuration of the crop cultivation support device 4 according to this exemplary embodiment.

[0132] As shown in the figure, the crop cultivation support device 4 includes a reception unit 401, a graph generation unit 402, a link prediction unit 403, an evaluation unit 404, a generation unit 405, a basis generation unit 406, and an output unit 407. Furthermore, similar to the crop cultivation support device 2 or crop cultivation support device 3 of the exemplary embodiment described above, the crop cultivation support device 4 may also include a learning unit, an input device, an output device, a communication device, etc., in addition to these components.

[0133] The reception unit 401 receives requests regarding the crop to be grown. For example, the reception unit 401 receives requests that include any of the following: size, taste, harvest time, and yield of the crop to be grown (more precisely, the harvest obtained from growing the crop). The reception unit 401 also receives input of desired growing results for the crop. Here, the desired growing results may constitute part of the request or may be received as input separate from the request. The reception unit 401 may also further receive work history related to the crop.

[0134] The graph generation unit 402 generates a growth target graph based on the above request. The growth target graph generated by the graph generation unit 402 includes nodes that indicate the desired growth result for the growth target. For example, as shown in Figure 11, the graph generation unit 302 may generate a growth target graph in which a node indicating "Cultivated Crops" is connected to a node indicating "Quality" by a link indicating "High Sugar Content". The growth target graph may also include nodes indicating "Work History".

[0135] The link prediction unit 403 uses the aforementioned target crop graph and cultivated crop graph to predict the relationships between nodes that are not connected by links in the target crop graph and cultivated crop graph. It then predicts nodes that link to nodes in the target crop graph from among the nodes in the cultivated crop graph that relate to the work performed during the cultivation of crops that were cultivated in the past. For example, in the example in Figure 11, the link prediction unit 403 predicts the work that links to the "Work History" node in the target crop graph with a "September" link, and the predicted work becomes a candidate for cultivation method.

[0136] The evaluation unit 404 evaluates the recommendation level of the node predicted by the link prediction unit 403, i.e., the candidate cultivation method, based on other nodes included in the cultivated graph that include the node predicted by the link prediction unit 403. For example, in the example in Figure 11, suppose the task linked to the "Work History" node by the "September" link is predicted to be "A1". Since "A1" is a node included in the cultivated graph of crop A, in this case, the evaluation unit 404 evaluates the recommendation level of task "A1" based on other nodes included in the cultivated graph of crop A (e.g., "High Sugar Content").

[0137] The generation unit 405 generates response information, including the cultivation method for the target crop, based on a trained model that has learned the relationship between cultivation methods and cultivation results for multiple crops, and the request received by the reception unit 401. More specifically, the generation unit 405 generates response information corresponding to the node predicted by the link prediction unit 403, i.e., the candidate cultivation method. This node indicates the work to be applied to the cultivation of the target crop. As described above, the link prediction unit 403 performs link prediction using the cultivated graph, so the generation unit 405 generates response information based on the trained model by generating response information based on the link prediction results from the link prediction unit 403. The generation unit 405 may also generate response information that includes cultivation methods among the candidate cultivation methods predicted by the link prediction unit 403, the recommendation level evaluated by the evaluation unit 404 is above a predetermined threshold.

[0138] The evidence generation unit 406 generates evidence information that demonstrates the validity of the response information generated by the generation unit 405. Specifically, the evidence generation unit 406 generates evidence information that includes past cases similar to the cultivation method of the target crop. The evidence generation unit 406 may also generate evidence information regarding the link prediction results by the link prediction unit 403 by analyzing the target crop graph and the cultivated crop graph.

[0139] The output unit 407 outputs various information generated by the crop cultivation support device 4. For example, the output unit 407 outputs response information generated by the generation unit 405 and rationale information indicated by the rationale generation unit 406. The output unit 407 may also output the recommendation level evaluated by the evaluation unit 404. Similar to the output unit 208 or output unit 308 of the exemplary embodiment described above, the destination of the information output is not particularly limited.

[0140] As described above, in the crop cultivation support device 4, the reception unit 401 receives input of desired cultivation results for the crop to be cultivated, and the link prediction unit 403 uses the cultivation target graph, which includes nodes indicating the input cultivation results, and the cultivated graph to predict the relationships between nodes that are not connected by links in the cultivation target graph and the cultivated graph, and predicts nodes that link to the nodes included in the cultivation target graph from among the nodes related to the work performed during the cultivation of crops that have been cultivated in the past, which are included in the cultivated graph.

[0141] A crop target graph containing nodes indicating desired cultivation results can be said to represent the state of the crop after various cultivation tasks have been performed. Therefore, nodes linked to the nodes in this crop target graph, which relate to tasks performed during the cultivation of previously grown crops, may indicate the factors that contributed to obtaining the desired cultivation results. Thus, the above configuration can provide information for determining the tasks necessary to obtain the desired cultivation results.

[0142] (Process flow) The flow of processing (crop cultivation support method) performed by the crop cultivation support device 4 will be explained based on Figure 13. Figure 13 is a flowchart showing the flow of processing performed by the crop cultivation support device 4.

[0143] In S401, the reception unit 401 receives requests regarding the crop to be grown. For example, the reception unit 401 receives requests that include any of the following: size, taste, harvest time, and yield of the crop to be grown (more precisely, the harvest obtained from growing the crop). The reception unit 401 also receives input of desired growing results for the crop. Here, the desired growing results may constitute part of the request or may be received as input separate from the request. The reception unit 401 may also further receive work history related to the crop.

[0144] In S402, the graph generation unit 402 generates a training target graph based on the information input in S401. The training target graph generated by the graph generation unit 402 includes nodes that indicate the desired training results for the training target.

[0145] In S403, the link prediction unit 403 predicts nodes that link to nodes in the target crop graph generated in S402, from among the nodes related to operations performed during the cultivation of crops previously cultivated, which are included in the cultivated graph. As described above, this node prediction is performed by link prediction using the trained cultivated graph and the target crop graph. In conjunction with the processing in S403, the rationale generation unit 406 may generate rationale information that shows the basis for the calculation results of the link prediction unit 403 by analyzing the target crop graph and the cultivated graph.

[0146] In S404, the evaluation unit 404 evaluates the recommendation level of the node predicted by the link prediction unit 403 in S403, i.e., the candidate for the cultivation method, based on other nodes included in the cultivated graph that include the node predicted by the link prediction unit 403 in S403.

[0147] In S405, the generation unit 405 estimates a cultivation method to obtain a desired cultivation result that matches the request received in S401, and generates response information including the estimated cultivation method. As an example, in S403, the generation unit 405 generates response information including cultivation methods from among the candidate cultivation methods predicted by the link prediction unit 403, the recommendation level evaluated by the evaluation unit 404 in S404 is above a predetermined threshold.

[0148] In S406, the evidence generation unit 405 generates evidence information that demonstrates the validity of the response information generated in S404. Specifically, the evidence generation unit 405 generates evidence information that includes past cases similar to the cultivation method of the target crop.

[0149] In S407, the output unit 407 outputs the response information generated in S406. The output unit 407 may also output the rationale information generated in S407 and the recommendation score evaluated in S404. This completes the process shown in Figure 13.

[0150] [Exemplary Embodiment 5] (overview) Figure 14 shows an overview of the crop cultivation support method according to this exemplary embodiment. In this exemplary embodiment, an example of supporting the cultivation of a target crop is described using a target crop graph that includes multiple nodes related to the target crop, and multiple cultivated graphs that have been generated for each of the multiple crops that have been cultivated in the past.

[0151] In this exemplary embodiment of the crop cultivation support method, a request is accepted that includes one of the following: the size, taste, harvest time, or yield of the crop to be cultivated.

[0152] Next, in the crop cultivation support method according to this exemplary embodiment, a cultivation target graph is generated based on the above request. In the example in Figure 14, the cultivation target graph is a graph that includes a node called "Cultivated Crops". This cultivation target graph includes nodes and links that indicate the growth status of the cultivated crop up to this point is "Standard", the growing environment is "High Temperature", and the variety is "x1", as well as nodes and links that indicate the work history of the cultivated crop in August is "x2".

[0153] Furthermore, in the crop cultivation support method according to this exemplary embodiment, link prediction is performed using the cultivation target graph generated as described above and multiple cultivated graphs generated for each of the multiple crops cultivated in the past to identify the crops that have a predetermined relationship with the cultivation target among the multiple crops cultivated in the past. The cultivated graphs used are generated for multiple crops and have been learned about the predetermined relationships between the multiple crops.

[0154] In the example in Figure 14, link prediction is performed using a cultivated graph that has been trained on the similarity of previously cultivated crops A, B, ... to predict previously cultivated crops similar to the target crop (hereinafter also referred to as similar crops). Here, "similar" means that the graphs are similar. In this example, dissimilar crops are not connected with "similar" links (dissimilar crops are treated as negative examples), but it is also possible to train "dissimilar" links. In addition, crop B also includes nodes and links that show the work history and cultivation environment, but these are omitted from the illustration in Figure 14.

[0155] Information on crops similar to the target crop is useful in the cultivation of the target crop. Therefore, by generating and outputting response information on similar crops identified as described above, it is possible to provide information useful for cultivating crops, thereby effectively supporting the cultivation of the target crop.

[0156] Furthermore, it is possible to evaluate the similar crops identified as described above and determine which similar crops to include in the response information based on the evaluation results. For example, suppose a request is received stating that the desired quality for the target crop is "high sugar content." In this case, if the identified similar crop possesses the quality of "high sugar content," response information including information about that similar crop may be generated. On the other hand, if the identified similar crop does not possess the quality of "high sugar content," information about that similar crop does not need to be included in the response information.

[0157] (Device configuration) The configuration of the crop cultivation support device 5 according to the fifth exemplary embodiment of the present invention will be described with reference to Figure 15. Figure 15 is a block diagram showing the configuration of the crop cultivation support device 5 according to this exemplary embodiment.

[0158] As shown in the figure, the crop cultivation support device 5 includes a reception unit 501, a graph generation unit 502, a link prediction unit 503, an evaluation unit 504, a generation unit 505, a basis generation unit 506, and an output unit 507. In addition to these components, the crop cultivation support device 5 may also include a learning unit, an input device, an output device, a communication device, etc., similar to the crop cultivation support devices 2 to 4 of the exemplary embodiments described above.

[0159] The reception unit 501 receives requests regarding the crop to be grown. For example, the reception unit 501 receives requests that include any of the following: size, taste, harvest time, and yield of the crop to be grown (more precisely, the harvest obtained from growing the crop). The reception unit 501 may also receive input of desired growing results for the crop. Furthermore, the reception unit 501 may also receive work history related to the crop.

[0160] The graph generation unit 502 generates a training target graph based on the above request. The training target graph generated by the graph generation unit 502 includes multiple nodes related to the training target, as shown in Figure 14.

[0161] The link prediction unit 503 uses the aforementioned target crop graph and multiple cultivated crop graphs generated for each of the multiple crops cultivated in the past to perform link prediction, which predicts the relationships between nodes that are not connected by links in the target crop graph and the cultivated crop graphs. This allows it to identify crops from among the multiple crops cultivated in the past that have a predetermined relationship with the target crop. The predetermined relationship may be similarity, as in the example in Figure 14, or it may be any other relationship. For example, the link prediction unit 503 may identify crops cultivated in the past that are dissimilar to the target crop, or it may identify crops cultivated in the past that belong to the same classification as the target crop, or crops cultivated in the past that have commonalities in quality with the target crop.

[0162] Furthermore, the link prediction unit 503 may perform link prediction without considering the work history at a later stage in the development process than when the request was received. For example, when predicting the work to be done in September based on a request received in August, the link prediction may be performed without considering the nodes related to work from September onwards in the developed graph.

[0163] The evaluation unit 504 evaluates the crop predicted by the link prediction unit 503. For example, the evaluation unit 504 may evaluate the crop predicted by the link prediction unit 503 depending on whether it matches the request. For example, suppose a request is received stating that the desired quality for the crop to be grown is "high sugar content". In this case, the recommendation level for similar crops that have the quality "high sugar content" may be higher than the recommendation level for similar crops that do not have the quality "high sugar content".

[0164] The evaluation unit 504 may also express the evaluation results numerically, similar to the evaluation unit 205 in the exemplary embodiment 2. In this case, if the relationship between the nodes included in the cultivated graph and the recommendation level is defined in advance, the evaluation unit 504 can calculate the recommendation level of the crop predicted by the link prediction unit 503 according to that rule.

[0165] For example, the evaluation unit 504 may calculate a recommendation score based on at least one of the following criteria for the crop growth graph predicted by the link prediction unit 503. • Points are added if the node and links demonstrate that the user has achieved their set growth goals; points are deducted if they are not. Points are added for nodes and links that show a high degree of agreement in prerequisites such as crop species, varieties, growing environment, and available materials and equipment; points are deducted for nodes and links that show a low degree of agreement. - Points will be deducted if the node or link indicates a high cost of work (work time, labor intensity, expenses, etc.), and points will be added if there are fewer such nodes or links. • Points will be deducted if nodes and links indicating high pesticide use are included, and points will be added if such nodes and links are included. Points are added if the node and links indicate that the user-specified yield was within the specified range at the specified time; points are deducted if the node and links indicate a discrepancy between the time and yield.

[0166] The generation unit 505 generates response information including the cultivation method for the target crop based on a trained model that has learned the relationship between cultivation methods and cultivation results for multiple crops, and a request received by the reception unit 501. As an example, the generation unit 505 generates response information corresponding to the crop predicted by the link prediction unit 503, which includes the cultivation method for the target crop. More specifically, the generation unit 505 may generate response information that includes at least a portion of the cultivation methods included in the cultivation graph of similar crops predicted by the link prediction unit 503, and which includes the content and timing of tasks that can be performed on the target crop in the future.

[0167] As described above, the link prediction unit 503 performs link prediction using the cultivated graph, so the generation unit 505 generates response information based on the link prediction results from the link prediction unit 503, thereby generating response information based on the trained model. The generation unit 505 may also generate response information that includes similar crops among the similar crops predicted by the link prediction unit 503 whose recommendation level evaluated by the evaluation unit 504 is above a predetermined threshold.

[0168] The evidence generation unit 506 generates evidence information that demonstrates the validity of the response information generated by the generation unit 505. Specifically, the evidence generation unit 506 generates evidence information that includes past cases similar to the cultivation method of the target crop. The evidence generation unit 506 may also generate evidence information regarding the link prediction results by the link prediction unit 503 by analyzing the target crop graph and the cultivated crop graph.

[0169] The output unit 507 outputs various types of information generated by the crop cultivation support device 5. For example, the output unit 507 outputs response information generated by the generation unit 505 and rationale information indicated by the rationale generation unit 506. The output unit 507 may also output the recommendation level evaluated by the evaluation unit 504. Similar to the output units 208, 308, and 408 of the exemplary embodiment described above, the destination of the information output is not particularly limited.

[0170] As described above, the crop cultivation support device 5 uses a cultivation target graph containing multiple nodes related to the cultivation target and multiple cultivated graphs generated for each of the multiple crops cultivated in the past to identify the crops that have a predetermined relationship with the cultivation target from among the multiple crops that have been cultivated in the past, by performing link prediction to predict the relationships between nodes that are not connected by links in the cultivation target graph and the cultivated graph.

[0171] Since information about crops that have a predetermined relationship with the target crop is useful in cultivating the target crop, the above configuration can provide information useful for cultivating crops.

[0172] (Process flow) The flow of processing (crop cultivation support method) performed by the crop cultivation support device 5 will be explained based on Figure 16. Figure 16 is a flowchart showing the flow of processing performed by the crop cultivation support device 4.

[0173] In S501, the reception unit 501 receives requests regarding the crop to be grown. For example, the reception unit 501 receives requests that include any of the following: size, taste, harvest time, and yield of the crop to be grown (more precisely, the harvest obtained from growing the crop). The reception unit may also receive input of desired growing results for the crop. Furthermore, the reception unit 501 may also receive work history related to the crop.

[0174] In S502, the graph generation unit 502 generates a training target graph based on the above request. The training target graph generated by the graph generation unit 502 includes multiple nodes related to the training target.

[0175] In S503, the link prediction unit 503 uses the aforementioned target crop graph and multiple cultivated graphs generated for each of the multiple crops cultivated in the past to perform link prediction, which predicts the relationships between nodes that are not connected by links in the target crop graph and the cultivated graphs, thereby identifying crops similar to the target crop (similar crops) from among the multiple crops cultivated in the past.

[0176] In S504, the evaluation unit 504 evaluates the crop predicted by the link prediction unit 503 in S503. For example, the evaluation unit 504 evaluates the crop depending on whether the crop predicted by the link prediction unit 503 in S503 matches the request.

[0177] In S505, the generation unit 505 determines the cultivation method to be cultivated based on a trained model that has learned the relationship between cultivation methods and cultivation results for multiple crops, and the request received by the reception unit 501 in S501. More specifically, in S503, the generation unit 505 determines the cultivation method to be cultivated as the cultivation method that links to the nodes of the crops predicted by the link prediction unit 503. Here, the generation unit 505 may also determine the cultivation method to be cultivated as the cultivation method that links to the nodes of crops predicted by the link prediction unit 503 in S503, and whose recommendation level evaluated by the evaluation unit 504 in S504 is above a predetermined threshold.

[0178] In S506, the generation unit 505 generates response information including the cultivation method determined in S505.

[0179] In S507, the evidence generation unit 505 generates evidence information that demonstrates the validity of the response information generated in S506. Specifically, the evidence generation unit 505 generates evidence information that includes past cases similar to the cultivation method of the target crop.

[0180] In S508, the output unit 507 outputs the response information generated in S506. At this time, the output unit 507 may also output the rationale information generated in S507 and the recommendation score evaluated in S504. This completes the process shown in Figure 16.

[0181] (Additional information regarding exemplary embodiment 5) The processing performed by the crop cultivation support device 5 is not limited to the examples described above. For example, the crop cultivation support device 5 may have a configuration similar to that of the graph generation unit 302, link prediction unit 303, and graph update unit 304 according to exemplary embodiment 3. In this configuration, for example, in S508, the output unit 507 may output response information and then receive information from the user who has referred to the response information regarding future cultivation content for the target crop. Then, based on the received cultivation content, the graph generation unit 302 may generate a target crop graph, the link prediction unit 303 may perform link prediction using the generated target crop graph, and the graph update unit 304 may update the target crop graph, thereby generating the response information described in exemplary embodiment 3.

[0182] [Variation] As explained in Exemplary Embodiment 4, by using the target crop graph and the cultivated crop graph, the cultivation results of the target crop (more precisely, the harvest obtained from cultivating the target crop) can be predicted by link prediction. Furthermore, the cultivation results of the target crop can also be predicted by methods other than link prediction. This will be explained with reference to Figure 17. Figure 17 is a diagram illustrating an example of predicting the cultivation results of a target crop based on features calculated from the target crop graph and the cultivated crop graph. Figure 17 shows the cultivated crop graphs of crops A1 to A3 that have been cultivated in the past, and the target crop graph for the cultivated crop. Note that the nodes and links included in these graphs are omitted from the illustration.

[0183] Here, by multiplying the features of each node in the cultivated graph by a weight corresponding to the link connected to that node and adding them together, we can calculate the features of each crop that has been cultivated in the past. Therefore, if we perform learning to update the weights so that the calculated features correspond to the cultivation results of that crop, it becomes possible to predict the cultivation results of a target crop from the features of the target crop graph calculated by applying those weights.

[0184] For example, in the example in Figure 17, the features calculated from the cultivated graph of crop A1, which is known to have high sugar content, are trained to fall within the range corresponding to the cultivation result "high sugar content" in the feature space. Similarly, the features calculated from the cultivated graph of crop A2, which is known to have high yield, are trained to fall within the range corresponding to the cultivation result "high yield" in the feature space. Likewise, the features calculated from the cultivated graph of crop A3, which is known to have large size, are trained to fall within the range corresponding to the cultivation result "large size" in the feature space.

[0185] In this case, as shown in the figure, if the feature quantities calculated from the growth target graph fall within the range corresponding to the growth result of "high sugar content," it can be predicted that the growth target will have the growth result of "high sugar content." Such a method for predicting growth results can be applied as an alternative method to the method for predicting growth results in the exemplary embodiment described above.

[0186] [Examples of implementation using software] Some or all of the functions of the crop cultivation support devices 1 to 5 (hereinafter referred to as "the devices") may be implemented by hardware such as integrated circuits (IC chips) or by software.

[0187] In the latter case, the device is implemented, for example, by a computer that executes instructions for a program, which is software that implements each function. An example of such a computer (hereinafter referred to as computer C) is shown in Figure 18. Computer C comprises at least one processor C1 and at least one memory C2. Memory C2 stores a program (crop cultivation support program) P that causes computer C to operate as the device. In computer C, the processor C1 reads program P from memory C2 and executes it, thereby implementing each function of the device.

[0188] Processor C1 can include, for example, a CPU (Central Processing Unit), GPU (Graphic Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating Point Number Processing Unit), PPU (Physics Processing Unit), microcontroller, or a combination thereof. Memory C2 can include, for example, flash memory, HDD (Hard Disk Drive), SSD (Solid State Drive), or a combination thereof.

[0189] Computer C may also be equipped with RAM (Random Access Memory) for loading program P at runtime and for temporarily storing various data. Furthermore, computer C may be equipped with communication interfaces for sending and receiving data with other devices. Additionally, computer C may be equipped with input / output interfaces for connecting input / output devices such as keyboards, mice, displays, and printers.

[0190] Furthermore, program P can be recorded on a non-temporary, tangible recording medium M that is readable by computer C. Such a recording medium M could be, for example, tape, disk, card, semiconductor memory, or programmable logic circuitry. Computer C can acquire program P via such a recording medium M. Program P can also be transmitted via a transmission medium. Such a transmission medium could be, for example, a communication network or broadcast waves. Computer C can also acquire program P via such a transmission medium.

[0191] [Additional Note 1] The present invention is not limited to the embodiments described above, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technical means disclosed in the embodiments described above are also included in the technical scope of the present invention.

[0192] [Additional Note 2] Some or all of the embodiments described above may also be described as follows. However, the present invention is not limited to the embodiments described below.

[0193] (Note 1) A crop cultivation support device comprising: a receiving means for receiving requests that include the size, taste, harvest time, and yield of a crop to be cultivated; a trained model that has learned the relationship between cultivation methods for multiple crops and cultivation results that include the size, taste, harvest time, and yield of the multiple crops; a generating means for generating response information that includes the cultivation method of the crop to be cultivated based on the request; and an output means for outputting the cultivation method. This configuration provides the effect of supporting crop cultivation.

[0194] (Note 2) The crop cultivation support device described in Appendix 1 further comprises a basis generation means for generating basis information including past cases similar to the cultivation method of the crop to be cultivated, and the output means further outputs the basis information. This allows the user to refer to response information based on that basis.

[0195] (Note 3) The crop cultivation support device described in Appendix 1 or 2 is a cultivated graph in which the trained model includes multiple nodes relating to crops cultivated in the past and links indicating the relationships between those nodes, and the relationships between the nodes have been learned. This configuration provides the effect of supporting crop cultivation.

[0196] (Note 4) The crop cultivation support device according to Appendix 3, comprising: a cultivation target graph including a plurality of nodes relating to the cultivation target and the cultivated graph, and a link prediction means for predicting the relationships between nodes that are not linked in the cultivation target graph and the cultivated graph, thereby predicting nodes that link to nodes in the cultivation target graph from among the nodes relating to work performed during the cultivation of crops that have been cultivated in the past, which are included in the cultivated graph, and the generation means generates the response information corresponding to the nodes predicted by the link prediction means. With this configuration, useful information relating to work on the cultivation target can be provided.

[0197] (Note 5) The crop cultivation support device according to Appendix 4, wherein the receiving means further accepts input of conditions for the cultivated graph, and the link prediction means predicts nodes that link to nodes included in the cultivation target graph from among the nodes related to operations performed during the cultivation of crops that have been cultivated in the past and that satisfy the conditions. With this configuration, it is possible to predict nodes that are in line with the user's intentions.

[0198] (Note 6) The crop cultivation support device according to Appendix 4, further comprising an evaluation means for evaluating the degree of recommendation of the node predicted by the link prediction means, based on other nodes included in the cultivated graph that include the node predicted by the link prediction means. This configuration can contribute to determining the work that can be expected to yield more favorable cultivation results.

[0199] (Note 7) The crop cultivation support device according to Appendix 3, wherein the receiving means further receives input of at least one of the content and timing of the work to be performed on the crop to be cultivated, and the receiving means calculates the probability that a node indicating a predetermined cultivation result will be linked to the crop to be cultivated by link prediction, which uses a crop to be cultivated graph that includes a node indicating at least one of the content and timing of the input work, and the cultivated graph to predict the relationship between nodes that are not connected by links in the crop to be cultivated graph and the cultivated graph, and the generating means generates the response information based on the probability calculated by the link prediction means. With this configuration, it is possible to predict whether a predetermined cultivation result is likely to be obtained by the work that the grower is about to perform, before the work is performed.

[0200] (Note 8) The crop cultivation support device according to Appendix 3, wherein the receiving means receives input of desired cultivation results for the crop to be cultivated, and uses a cultivation target graph including nodes indicating the input cultivation results and the cultivated graph to predict the relationships between nodes that are not connected by links in the cultivation target graph and the cultivated graph, and uses link prediction to predict nodes that link to nodes in the cultivation target graph from among the nodes related to work performed during the cultivation of crops that have been cultivated in the past, which are included in the cultivated graph, and the generating means generates the response information based on the nodes calculated by the link prediction means.

[0201] (Note 9) The crop cultivation support device described in Appendix 3, comprising a link prediction means for predicting the relationships between nodes that are not linked in the crop cultivation graph and the cultivated graph, using a crop cultivation graph that includes multiple nodes relating to the crop to be cultivated and multiple cultivated graphs generated for each of the multiple crops cultivated in the past, and the generation means for identifying among the multiple crops cultivated in the past that have a predetermined relationship with the crop to be cultivated, wherein the generation means generates the response information relating to the crop identified by the link prediction means. With this configuration, it is possible to provide information useful for cultivating crops.

[0202] (Note 10) A method for supporting crop cultivation in which a computer receives a request that includes one of the following: size, taste, harvest time, and yield of a crop to be cultivated; a trained model that has learned the relationship between cultivation methods for multiple crops and cultivation results including one of the following: size, taste, harvest time, and yield of the multiple crops; and the request, generates response information including a cultivation method for the crop to be cultivated and outputs the cultivation method. This configuration provides the effect of supporting crop cultivation.

[0203] (Note 11) A crop cultivation support program that causes a computer to perform the following actions: receiving a request that includes one of the following: the size, taste, harvest time, and yield of the crop to be cultivated; a trained model that has learned the relationship between cultivation methods for multiple crops and cultivation results that include one of the following: the size, taste, harvest time, and yield of the multiple crops; generating response information that includes the cultivation method of the crop to be cultivated based on the request; and outputting the cultivation method. This configuration provides the effect of supporting crop cultivation.

[0204] [Additional Note 3] Some or all of the embodiments described above can also be expressed as follows:

[0205] A crop cultivation support device comprising: a process for receiving a request that includes the size, taste, harvest time, and yield of a crop to be cultivated; a trained model that has learned the relationship between a plurality of crop cultivation methods and cultivation results that include the size, taste, harvest time, and yield of the plurality of crops; a process for generating response information that includes the cultivation method of the crop to be cultivated based on the request; and a process for outputting the cultivation method.

[0206] Furthermore, this crop cultivation support device may also be equipped with memory, which may store a program (crop cultivation support program) that causes the processor to execute the following: a process for receiving a request that includes one of the size, taste, harvest time, and yield of the crop to be cultivated; a trained model that has learned the relationship between cultivation methods for multiple crops and cultivation results that include one of the size, taste, harvest time, and yield of the multiple crops; a process for generating response information that includes the cultivation method of the crop to be cultivated based on the request; and a process for outputting the cultivation method. This program may also be recorded on a computer-readable, non-temporary, tangible recording medium. [Explanation of Symbols]

[0207] 1. Crop cultivation support device 11 Reception Department 12 Generation part 13 Output section 2. Crop cultivation support device 201 Reception Department 204 Link Prediction Section 205 Evaluation Department 206 Generation part 207 Basis Generation Department 208 Output section 3. Crop cultivation support device Room 301 Reception 303 Link Prediction Unit 304 Graph Update Section 305 Generation part 306 Basis Generation Department 307 Output section 4. Crop cultivation support device 401 Reception Desk 403 Link Prediction Section 405 Generator 406 Basis Generation Section 407 Output section 5. Crop cultivation support device 501 Reception Department 502 Graph Generation Unit 503 Link Prediction Unit 504 Evaluation Department 505 Generation part 506 Basis Generation Department

Claims

1. A means of receiving requests that include any of the following: size, taste, harvest time, and yield of the crop to be cultivated. A trained model that has learned the relationship between methods for growing multiple crops and the results of growing the multiple crops, including the size, taste, harvest time, and yield of the crops, and a generation means that generates response information including the method for growing the target crop based on the request, An output means for outputting the aforementioned cultivation method, Equipped with, The aforementioned trained model is a trained graph that includes multiple nodes related to the cultivation of crops that have been cultivated in the past, and links that show the relationships between those nodes, and the relationships between the nodes have been learned. The system further comprises a link prediction means that uses a target graph containing multiple nodes relating to the target to be trained and the trained graph to perform link prediction for predicting the relationships between nodes that are not connected by links in the target graph and the trained graph, The generation means generates the response information based on the link prediction in the growth target graph. A device to support crop growth.

2. The system further comprises evidence generation means for generating evidence information including past cases similar to the cultivation methods of the crops to be cultivated, The output means further outputs the basis information. The crop cultivation support device according to claim 1.

3. The link prediction means uses the target growth graph, which includes a plurality of nodes relating to the target growth, and the grown-up graph to predict the relationships between nodes that are not connected by links in the target growth graph and the grown-up graph, and predicts nodes that link to nodes included in the target growth graph from among the nodes in the grown-up graph relating to operations performed during the growth of crops that have been grown in the past, The crop cultivation support apparatus according to claim 1 or 2, wherein the generation means generates the response information corresponding to the node predicted by the link prediction means.

4. The aforementioned receiving means further accepts input of conditions for the cultivated graph, The crop cultivation support device according to claim 3, wherein the link prediction means predicts a node that links to a node included in the cultivation target graph from among the nodes relating to operations performed during the cultivation of crops that have been cultivated in the past and are included in the cultivated graph that satisfies the above conditions.

5. The crop cultivation support device according to claim 3, further comprising an evaluation means for evaluating the degree of recommendation of the node predicted by the link prediction means based on other nodes included in the cultivated graph that include the node predicted by the link prediction means.

6. The aforementioned receiving means further receives input of at least one of the content and timing of the work to be performed on the training target, The link prediction means uses the target graph, which includes nodes indicating at least one of the content and timing of the input work, and the completed graph to calculate the probability that a node indicating a predetermined training result will be linked to the target graph by the link prediction, which predicts the relationship between nodes that are not connected by links in the target graph and the completed graph. The crop cultivation support apparatus according to claim 1 or 2, wherein the generation means generates the response information based on the probability calculated by the link prediction means.

7. The receiving means receives input of desired cultivation results for the crop to be cultivated. The link prediction means uses the cultivation target graph, which includes nodes showing the input cultivation results, and the cultivated graph to predict the relationships between nodes that are not connected by links in the cultivation target graph and the cultivated graph, and predicts nodes that link to nodes in the cultivation target graph from among the nodes in the cultivated graph that relate to operations performed during the cultivation of crops that have been cultivated in the past. The crop cultivation support apparatus according to claim 1 or 2, wherein the generation means generates the response information based on the node calculated by the link prediction means.

8. The link prediction means uses the target crop graph which includes a plurality of nodes relating to the target crop and a plurality of cultivated graphs which have been generated for each of the plurality of crops that have been cultivated in the past to predict the relationships between nodes that are not connected by links in the target crop graph and the cultivated graph, thereby identifying the plurality of crops that have been cultivated in the past that have a predetermined relationship with the target crop. The crop cultivation support apparatus according to claim 1 or 2, wherein the generation means generates the response information relating to the crop identified by the link prediction means.

9. Computers We accept requests that include any of the following: size, taste, harvest time, and yield of the crop to be cultivated. Based on a trained model that has learned the relationship between cultivation methods for multiple crops and cultivation results including size, taste, harvest time, and yield of the multiple crops, and the request, response information including the cultivation method for the target crop is generated. Output the aforementioned breeding method, The aforementioned trained model is a trained graph that includes multiple nodes related to the cultivation of crops that have been cultivated in the past, and links that show the relationships between those nodes, and the relationships between the nodes have been learned. Using the target graph containing multiple nodes related to the target to be trained and the trained graph, link prediction is performed to predict the relationships between nodes that are not connected by links in the target graph and the trained graph. Based on the link prediction in the growth target graph, the response information is generated. Methods for supporting crop cultivation.

10. For computers, A process that accepts requests that include one of the following: size, taste, harvest time, and yield of the crop to be cultivated. A trained model that has learned the relationship between methods for growing multiple crops and the results of growing the crops, including their size, taste, harvest time, and yield, and a request, generates response information including the method for growing the target crop. A process to output the aforementioned cultivation method, Make it run, The aforementioned trained model is a trained graph that includes multiple nodes related to the cultivation of crops that have been cultivated in the past, and links that show the relationships between those nodes, and the relationships between the nodes have been learned. A process for predicting links to predict the relationships between nodes in the target graph and the developed graph that are not connected by links, using a target graph containing multiple nodes related to the target to be trained and the developed graph. Based on the link prediction in the aforementioned growth target graph, the response information is generated. Agricultural crop cultivation support program.

Images