Method, apparatus, and program for generating a prediction model for predicting crop production results

Abstract

To implement a technology for predicting production performance of crops.SOLUTION: An apparatus (20) for generating a prediction model for predicting production performance of crops includes a prediction model generation unit (23) which generates a prediction model (24) which predicts production performance of the year from at least a part of actual measurement values of production performance of previous years, the prediction model (24) referring to actual measurement value data of past annual production performance and representing interannual change of production performance.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method for generating a prediction model for predicting the production performance of agricultural crops, a generating apparatus, and a generation program.

Background Art

[0002] In various agricultural crop cultivation sites, in order to be utilized for cultivation management, shipping plans, sales negotiations, etc., attempts have been made to predict the component content of agricultural crops during the growth period and harvest period using meteorological data.

[0003] Patent Document 1 describes a technique for predicting solar radiation information and temperature information for making the sugar content of a crop the target sugar content, and controlling the environment in a greenhouse based on the predicted solar radiation information and temperature information. In the technique described in Patent Document 1, the difference between the reference sugar content specific to the crop and the target sugar content is related to the difference between the average value of solar radiation and temperature in the four weeks before harvest and the reference value, and is used for sugar content prediction.

[0004] Non-Patent Document 1 describes constructing a model learned using time-series data of meteorological variations and time-series data of the quality of mandarin oranges over the past few years. In the technique described in Non-Patent Document 1, using the constructed model, the quality of mandarin oranges at the harvest time of the current year is predicted from the time-series data of meteorology and the quality of mandarin oranges up to the harvest time of the current year.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Non-Patent Documents

[0006]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0007] As problems in predicting the component content of agricultural crops, the following three can be cited. The first is that the number of data measurement points is small. There are few examples of frequently measuring the component content of agricultural crops during the growth process to accumulate data, and in most cases, the component content is only measured at the harvest stage or at the time of shipment. The second is that the component content is also affected by factors other than meteorology. For example, in the case of perennial crops, annual fluctuations may affect the component content, and in the case of annual crops, the annual trend of cultivation management reflecting consumer preferences may affect the component content. The third is that in many cases, the types and times of meteorological data variables that affect the component content are unknown.

[0008] Therefore, if a prediction model that reflects the annual changes in the crop production environment can be realized using past measured values of component content and meteorological data, it will greatly contribute to the cultivation site.

[0009] In the technology described in Patent Document 1, it is difficult to apply it to sugar content prediction when the standard sugar content specific to the crop is unknown. Also, in the technology described in Patent Document 1, when the harvest date is undetermined, since the starting points of the solar radiation and air temperature data used for prediction are not determined, it is difficult to apply it to sugar content prediction.

[0010] In the technology described in Non-Patent Document 1, in order to create time-series data on the quality of mandarin oranges, it is necessary to measure the component content multiple times during the growth process. Also, the model described in Non-Patent Document 1 can be used for predicting the quality at the next stage in the cultivation calendar, but it is difficult to predict the quality earlier.

[0011] One aspect of the present invention has been made to solve the above-described problems, and its object is to realize a technology for predicting the production performance of agricultural crops.

Means for Solving the Problems

[0012] A method for generating a prediction model according to one aspect of the present invention is a method for generating a prediction model for predicting the production performance of agricultural crops. The method includes a step of generating a prediction model that represents the secular change in the production performance by referring to the measured value data of the production performance for each past year, and predicts the production performance of the current year from at least a part of the measured values of the production performance up to the previous year.

[0013] A prediction model generation device according to one aspect of the present invention is a prediction model generation device for predicting the production performance of agricultural crops. The device includes a generation unit that generates a prediction model that represents the secular change in the production performance by referring to the measured value data of the production performance for each past year, and predicts the production performance of the current year from at least a part of the measured values of the production performance up to the previous year.

[0014] The prediction model generation device according to each aspect of the present invention may be realized by a computer. In this case, a control program for the generation device that realizes the generation device by operating the computer as each part (software element) included in the generation device, and a computer-readable recording medium on which the program is recorded also fall within the scope of the present invention.

Effects of the Invention

[0015] According to one aspect of the present invention, a technique for predicting the production performance of agricultural crops can be realized.

Brief Description of the Drawings

[0016]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Embodiments for Carrying Out the Invention

[0017] One aspect of the present invention realizes a technology for predicting the production performance of agricultural crops. The production performance of agricultural crops predicted according to one aspect of the present invention includes the yield of agricultural crops; component contents such as sugar content, acidity, protein content, molecular components measured by mass spectrometry, etc., and pigment content; forms such as color, shape, size, length, etc.; lodging degree, growth indices of leaves, flowers, stems, fruits (length of stems, degree of opening of leaves, coloring of flowers and fruits), number of flowering, number of fruiting, and number of leaves, but is not limited thereto. That is, according to one aspect of the present invention, at least one of the evaluation items for evaluating the produced agricultural crops can be predicted. The agricultural crops whose production performance is predicted according to one aspect of the present invention are not particularly limited, and include perennial plants such as oranges, apples, peaches, grapes, loquats, plums, coffee, green tea, beets, turmeric, licorice, peonies, and rhubarb; annual plants such as rice, wheat, soybeans, peanuts, millet, tomatoes, spinach, eggplants, pumpkins, strawberries, perilla, ginseng, psyllium, potatoes, sweet potatoes, and ornamental flowers, etc. are given as examples.

[0018] 〔Prediction System 100〕 Based on FIG. 1, the prediction system 100 used for predicting the production performance of agricultural crops will be described. The prediction system 100 includes a prediction device 10, a prediction model generation device 20, and a weather data prediction model generation device 30. Further, the prediction system 100 further includes an input device 40, a storage device 50, and an output device 60. The prediction system 100 may include the prediction device 10, the prediction model generation device 20, and the prediction model generation device 30 as independent devices respectively, or may be integrally provided in one device.

[0019] The input device 40 receives input operations by the user for the prediction system 100. As an example, the input device 40 receives the input of data used to predict the production results of agricultural crops in the prediction device 10. Further, the input device 40 receives the input of data used to generate a prediction model in the prediction model generation device 20. Furthermore, the input device 40 receives the input of data used to generate a prediction model of weather data in the prediction model generation device 30. The input device 40 may be, for example, a keyboard, a mouse, a touch sensor, or the like.

[0020] The storage device 50 stores programs and data used in the prediction system 100. As an example, the storage device 50 stores various data input via the input device 40. Further, as an example, the storage device 50 stores a prediction model, input information, and output information used to predict the production results of agricultural crops in the prediction device 10. Furthermore, as an example, the storage device 50 stores learning data used for generating a prediction model and the generated prediction model in the prediction model generation device 20. Also, as an example, the storage device 50 stores learning data used for generating a prediction model and the generated prediction model in the prediction model generation device 30. The storage device 50 may have a database for storing various data on the cloud or a server.

[0021] The output device 60 outputs the result predicted by the prediction device 10. Further, the output device 60 may output cultivation management guidance information based on the result predicted by the prediction device 10. The cultivation management guidance information includes, for example, the timing of mulching, information on the maintenance of drainage channels, the number and timing of fruit picking, information on fertilization, information on the implementation of root pruning, and the like.

[0022] The mode of output by the output device 60 is not particularly limited. The output device 60 may be, for example, a display device that displays the information as an image, a printing device that prints the information, or an alarm device that outputs the information as sound. Further, the output device 60 may be a display of a mobile device such as a smartphone that displays the result predicted by the prediction device 10, cultivation management guidance information based on the result, and the like.

[0023] (Prediction device 10) The prediction device 10 is a prediction device that predicts the production results of agricultural crops. The prediction device 10 uses a prediction model representing the secular change in production results, which is generated by referring to the measured value data of production results for each past year, to predict the production results of the current year from at least a part of the measured values of production results up to the previous year.

[0024] The prediction device 10 includes a control unit 11. The control unit 11 comprehensively controls each part of the prediction device 10 and is realized by, for example, a processor and a memory. In this example, the processor accesses a storage (not shown), loads a program (not shown) stored in the storage into the memory, and executes a series of instructions included in the program. Thereby, each part of the control unit 11 is configured. As each of these parts, the control unit 11 includes an input data acquisition unit 12 and a prediction unit 13.

[0025] The input data acquisition unit 12 acquires input data that is input information for a prediction model for predicting production results. The input data acquisition unit 12 reads out measured value data including at least a part of the measured values of production results up to the previous year from the storage device 50 based on an input signal representing an instruction to start prediction from the input device 40. Further, the input data acquisition unit 12 may acquire the measured value data input via the input device 40. The input data acquisition unit 12 outputs the acquired input data to the prediction unit 13.

[0026] Here, the measured values of the production results up to the previous year are intended to be the measured values of the production results measured before the current year, which is the year for which the production results are to be predicted. The measured values of the production results only need to be measured at least once in each year, during at least the growth period, the harvest period, or at the time of shipment. The measured value data only needs to include at least the measured values of the production results of the previous year, but may also include the measured values of the production results of multiple years up to the previous year.

[0027] The prediction unit 13 predicts the production results of the current year from at least a part of the measured values of the production results up to the previous year, using a prediction model 24 that represents the secular change in the production results and is generated by referring to the measured value data of the production results for each past year. The prediction unit 13 reads the prediction model 24 from the storage device 50, inputs the measured value data sent from the input data acquisition unit 12 into the prediction model 24, and acquires the production results of the current year output from the prediction model 24. The prediction unit 13 outputs the data representing the production results of the current year acquired as the prediction result to the output device 60.

[0028] The prediction model 24 used by the prediction unit 13 to predict the production results can be the prediction model 24 generated by a prediction model generation device 20 described later. As an example, the prediction model 24 is generated by performing machine learning using the measured value data as learning data, and is a prediction model in which at least a part of the measured values of the production results up to the previous year is input information and the production results of the current year are output information. That is, by using the prediction model 24, it is possible to predict the production results while considering the annual fluctuations (fluctuations in the production results) of the production results of agricultural crops as secular changes.

[0029] As an example, it is known that perennial plants such as mandarin oranges tend to have a decrease in production results in the year following a year with good production results due to a decline in tree vigor such as a decrease in the stored nutrients of the tree. And perennial plants repeat good years (years with good production results) and bad years (years with bad production results) alternately. Thus, since there are annual fluctuations in the production results of agricultural crops, it is possible to predict the production results considering the annual fluctuations by using the prediction model 24 that predicts the production results of the current year from the measured values of the production results up to the previous year.

[0030] Also, as another example, in annual plants, fluctuations in cultivation management methods can occur on an annual basis, and these fluctuations in cultivation management methods lead to fluctuations in production results from year to year. Therefore, even in annual plants, by using the prediction model 24 that predicts the production results of the current year from the measured values of production results up to the previous year, it is possible to predict the production results considering the annual fluctuations caused by the fluctuations in cultivation management methods.

[0031] Here, the measured value data used for predicting production results will be described with reference to FIG. 2. FIG. 2 is a diagram showing an example of the measured value data of the production results for each year up to the previous year used in the prediction device 10. In the graph on the left side of FIG. 2, the X-axis represents the calendar year, the Y-axis represents the sugar content, and the Z-axis represents the year. This graph shows the annual change in sugar content for each year. By extracting the data of the sugar content change for each year on a specific day from the data of the annual sugar content change for each year like this, a graph representing the sugar content change for each year on a specific day can be obtained as shown in the graph on the right side of FIG. 2. Note that the graph on the left side of FIG. 2 is obtained from the results of measuring the sugar content multiple times in each year. FIG. 2 is shown for explaining the concept of the present invention and does not limit the present invention to the embodiment shown in FIG. 2.

[0032] As an example, in FIG. 2, a graph representing the annual sugar content change for each of the past 15 years (left graph) and a graph representing the sugar content change for the past 15 years on October 30 (right graph) are shown. The prediction device 10 can thus predict production results considering annual fluctuations by using data representing the sugar content change for each year on a specific day.

[0033] When actually measuring the annual changes in production performance, multiple measurements are repeated annually for a reference plant individual (such as a determined single reference tree). That is, for sampling, leaves and fruits are thinned out from the reference plant individual multiple times a year. Due to the removal of leaves and fruits by sampling, the translocation of photosynthetic products within the plant individual may change, or the wounds caused by sampling may become injury stresses, and thus the state of the plant individual may change. Therefore, the production performance will be measured in a state different from normal growth, and the data may not represent the actual growth process. If the production performance of agricultural crops is predicted based on the data obtained in this way, there is a risk of a decrease in prediction accuracy. In the prediction device 10, since the production performance is predicted considering the changes in the production performance for each year on a specific day instead of the annual changes in production performance, accurate prediction is possible.

[0034] In addition, the prediction unit 13 can predict the production performance of the current year using a prediction model 24 that represents the secular change in production performance and the influence of weather, which is generated by referring to measurement data including the measured value data of the production performance for each past year and the observed value data of the weather data of those years. The prediction unit 13 reads the prediction model 24 from the storage device 50, inputs the measured value data sent from the input data acquisition unit 12 and the forecast value of the weather data of the current year into the prediction model 24, and acquires the production performance of the current year output from the prediction model 24. As an example, the forecast value of the weather data can be obtained from a database publicly provided by an organization that provides weather information such as the Japan Meteorological Agency. The prediction device 10 may store the acquired forecast value of the weather data in the storage device 50 and read it from the storage device 50 when the prediction unit 13 predicts the production performance.

[0035] The prediction model 24 is generated by performing machine learning using measurement data including the measured value data of the production results for each past year and the observed value data of the meteorological data for those years as learning data. And such a prediction model 24 is a prediction model in which at least a part of the measured values of the production results up to the previous year and the forecast values of the meteorological data for the current year are input information, and the production results for the current year are output information. By using such a prediction model 24, it is possible to predict the production results in consideration of the meteorological data that affects the production results along with the fluctuations in the production results of agricultural crops for each year. By further considering the meteorological data closely related to agriculture, it is possible to predict with higher accuracy than when predicting the production results considering only the fluctuations in the production results for each year.

[0036] The influence of aging and weather on the production results will be described with reference to FIG. 3. FIG. 3 is a diagram for explaining the aging change of the production results of agricultural crops and the influence of weather on the production results. In the graph of FIG. 3, the X-axis represents the year and the Y-axis represents the sugar content, and the graph represents the change in sugar content for each year on a specific day. It is considered that this change in sugar content per year is affected not only by the aging change of the production results but also by the weather. Therefore, as in the graph on the right side of FIG. 3, the change in sugar content per year (data of the solid line) is decomposed into the change in sugar content due to the influence of aging change (data of the broken line) and the change in sugar content due to the influence of weather represented by the difference between the data of the broken line and the data of the solid line (range of the arrow).

[0037] The sugar content representing the influence of the aging change represented by the broken line in FIG. 3 is the sugar content predicted in consideration of the above-described fluctuations in the production results for each year, and this is used as the reference sugar content (reference production result). Since the reference sugar content is predicted considering only the fluctuations for each year, it may be different from the actual sugar content represented by the solid line in FIG. 3. The difference between this actual sugar content and the reference sugar content predicted considering the fluctuations for each year is considered to be the influence of the weather on the sugar content.

[0038] Since the prediction model 24 is a prediction model that predicts the production results of the current year from measurement data including the measured value data of the production results for each past year and the observed value data of the weather data for those years, it is possible to predict the production results considering not only the fluctuations for each year but also the influence of the weather. Therefore, in the graph on the right side of FIG. 3, the correction within the range of the arrow representing the influence of the weather is added to the reference production results represented by the broken line, enabling a highly accurate prediction of the production results that is closer to the actual production results represented by the solid line.

[0039] Also, the prediction model 24 is a prediction model generated for each prediction target location, and can be a model that predicts the production results of the current year at the prediction target location from at least a part of the measured values of the production results up to the previous year at the prediction target location and the forecast values of the weather data for the current year at the prediction target location.

[0040] The secular changes and the influence of the weather on the production results can vary depending on the region where the crops are grown. Therefore, by predicting the production results using the prediction model 24 created for each prediction target location, which is the region where the crops for predicting the production results are grown, the production results can be predicted with higher accuracy.

[0041] According to the prediction device 10, since the production results of the current year can be predicted based on at least a part of the measured values of the production results up to the previous year, it is possible to predict the production results at the harvest time or the shipping date of that year even at the initial stage of the cultivation period. As a result, by predicting the production results of that year at the initial stage of the cultivation period and changing the cultivation management such as fertilization, irrigation, and fruit picking based on the prediction results, the production results of that year can be improved. Also, based on the predicted production results, it is possible to determine the optimal working time and harvest time.

[0042] (Prediction Model Generation Device 20) The prediction model generation device 20 is a device that generates a prediction model for predicting the production performance of agricultural crops. The prediction model generation device 20 refers to the measured value data of the production performance for each past year, and generates a prediction model that represents the secular change in the production performance, and predicts the production performance of the current year from at least a part of the measured values of the production performance up to the previous year.

[0043] The prediction model generation device 20 includes a control unit 21. The control unit 21 comprehensively controls each part of the prediction model generation device 20, and is realized by, for example, a processor and a memory. In this example, the processor accesses a storage (not shown), loads a program (not shown) stored in the storage into the memory, and executes a series of instructions included in the program. Thereby, each part of the control unit 21 is configured. As each of these parts, the control unit 21 includes a learning data acquisition unit 22 and a prediction model generation unit (generation unit) 23.

[0044] The learning data acquisition unit 22 acquires learning data for learning the prediction model 24. The learning data acquisition unit 22 reads the learning data from the storage device 50 based on an input signal representing an instruction to start learning from the input device 40. Further, the learning data acquisition unit 22 may acquire the learning data input via the input device 40. The learning data acquisition unit 22 outputs the acquired learning data to the prediction model generation unit 23.

[0045] The prediction model generation unit 23 refers to the measured value data of the production performance for each past year, and generates a prediction model 24 that represents the secular change in the production performance, and predicts the production performance of the current year from at least a part of the measured values of the production performance up to the previous year.

[0046] The prediction model generation unit 23 generates a prediction model 24 by performing machine learning using the learning data acquired from the learning data acquisition unit 22. Here, the learning data is actual measurement value data. As an example, the prediction model generation unit 23 generates the prediction model 24 using known machine learning methods such as neural networks, decision trees, random forests, and support vector machines. The prediction model 24 generated by the prediction model generation device 20 is a prediction model that takes at least a part of the actual measurement values of the production results up to the previous year as input information and outputs the production results of the current year as output information.

[0047] Also, as an example, the prediction model generation unit 23 may generate a statistical model representing the correlation between the annual production results up to the previous year of a certain year and the production results of that year. That is, the prediction model 24 may be a statistical model such as a linear regression model, sparse modeling, generalized linear model, state space model, hierarchical Bayesian model, time series model, or clustering.

[0048] Since the prediction model 24 is a model representing the secular change in production results by referring to the actual measurement value data of the production results for each past year, it can be used for predicting production results considering the fluctuations in the annual production results of agricultural crops as secular changes. The prediction model 24 is the prediction model used for predicting production results in the above-described prediction device 10. The prediction model generation unit 23 may store the generated prediction model 24 in the storage device 50.

[0049] In addition, the prediction model generation unit 23 can generate a prediction model 24 that represents the secular change in production performance and the influence of weather by referring to measurement data including the measured value data of the production performance for each past year and the observed value data of the weather data for those years. Such a prediction model 24 is a prediction model that predicts the production performance of the current year from at least a part of the measured values of the production performance up to the previous year and the forecast values of the weather data of the current year. The observed value data of the weather data can be obtained, for example, from a database publicly provided by an organization that provides weather information such as the Japan Meteorological Agency. The prediction model generation device 20 may store the observed values of the acquired weather data in the storage device 50 and read them out from the storage device 50 when generating the prediction model 24 in the prediction model generation unit 23.

[0050] The prediction model generation unit 23 performs machine learning using the measurement data as learning data, and generates a prediction model in which at least a part of the measured values of the production performance up to the previous year and the forecast values of the weather data of the current year are input information, and the production performance of the current year is output information. Further, as an example, the prediction model generation unit 23 may generate a statistical model representing the correlation between the production performance for each year up to the previous year of a certain year, the observed values of the weather data of that year, and the production performance of that year as the prediction model 24.

[0051] The prediction model 24 can be used for predicting the production performance, taking into account the weather data that affects the production performance along with the fluctuations in the annual production performance of agricultural crops. By considering the weather data closely related to agriculture, it can be used to predict with higher accuracy than when predicting the production performance considering only the fluctuations in the annual production performance.

[0052] The prediction model generation unit 23 preferably generates the prediction model 24 for each prediction target location. The prediction model generation unit 23 generates the prediction model 24 for each prediction target location by referring to measurement data including the measured values of the production performance for each past year at the prediction target location and the observed values of the weather data for those years at the prediction target location.

[0053] The secular changes and the influence of weather on production results can vary from region to region where crops are grown. Therefore, by generating the prediction model 24 for each target point, which is the region where the crops to be predicted are grown, it can be used for more accurate prediction of production results.

[0054] The prediction model generation unit 23 may regenerate the prediction model 24 based on the prediction results in order to improve the prediction accuracy. When the prediction model 24 is a learning model generated by machine learning, the prediction model 24 may be relearned. As an example, the prediction model generation unit 23 regenerates the prediction model 24 so that the difference between the measured value of the production result in a certain past year and the predicted value of the production result in that year predicted using the prediction model 24 is minimized.

[0055] As an example, the prediction model generation unit 23 calculates the difference between the measured value of the production result for each year up to the previous year and the predicted value of the production result for those years predicted using the prediction model 24 for each year, and regenerates the prediction model 24 so that the difference is minimized for all years. Also, the prediction model generation unit 23 may regenerate the prediction model 24 so that the sum of the squares of the differences between the measured value of the production result in a certain past year and the predicted value of the production result in that year predicted using the prediction model 24 is minimized.

[0056] Furthermore, when regenerating the prediction model 24, the prediction model generation unit 23 may regenerate the prediction model using regression analysis that regresses the difference between the measured value of the production result in a certain past year and the predicted value of the production result in that year predicted by the prediction model 24 based on the observed value of the weather data for that year.

[0057] As described with respect to the prediction device 10, referring to FIG. 3, the change in production results for each year can be decomposed into the change in production results due to the influence of secular changes and the change in production results due to the influence of weather. The prediction model generation unit 23 decomposes the change in production results for each year into the change in production results due to the influence of secular changes and the change in production results due to the influence of weather, taking into account the balance between the two.

[0058] The prediction model generation unit 23 calculates, as an example, a reference production result that is a predicted value of the production result considering secular changes using regularization. First, the strength of the regularization is varied in various ways, and the reference production result is calculated for each regularization coefficient. Then, the difference between the measured value of the production result and the reference production result for each regularization coefficient is obtained, and this difference is calculated as the influence of the weather on the production result respectively. Then, the influence of the weather is regressed using the observed values of the weather data of that year. The prediction model generation unit 23 repeats such processing to determine the regularization coefficient, the reference production result, and the regression coefficient of the weather data at which the difference between the predicted value and the measured value of the production result becomes minimum, and regenerates the prediction model 24.

[0059] Note that an example of the regression analysis of the influence of the weather on the production result using the weather data, which is performed when the prediction model generation unit 23 regenerates the prediction model 24, can be the regression analysis performed by the prediction model generation unit 33 in the prediction model generation apparatus 30 described later.

[0060] In this way, by regenerating the prediction model 24, it is possible to generate a prediction model 24 that can predict the production result with higher accuracy.

[0061] According to the prediction model generation apparatus 20, since it is possible to generate a prediction model that predicts the production result of the current year based on at least a part of the measured values of the production results up to the previous year, it can also be used to predict the production result of the harvest period or the shipping date of that year at the initial stage of the cultivation period. Thereby, it can be used to predict the production result of that year at the initial stage of the cultivation period, and based on the prediction result, by changing cultivation management such as fertilization, irrigation, and fruit picking, it can be used to improve the production result of that year. Also, it can be used to determine the optimal working time and harvest time based on the predicted production result.

[0062] (Prediction model generation apparatus 30) The weather data prediction model generation device 30 is a device that generates a prediction model for predicting the production performance of agricultural crops. The prediction model generation device 30 refers to the observed value data of weather data before the prediction target date and generates a prediction model 34 that represents the production performance and the influence of weather, and predicts the production performance from the forecast value of the weather data on the prediction target date. The prediction model generation device 30 selects a combination of variables of weather data so that the difference between the measured value of the production performance in a certain past year and the predicted value of the production performance in that year predicted using the prediction model 34 is minimized, and generates the prediction model 34.

[0063] The prediction model generation device 30 includes a control unit 31. The control unit 31 comprehensively controls each part of the prediction model generation device 30 and is realized by, for example, a processor and a memory. In this example, the processor accesses a storage (not shown), loads a program (not shown) stored in the storage into the memory, and executes a series of instructions included in the program. Thereby, each part of the control unit 31 is configured. As each part, the control unit 31 includes a learning data acquisition unit 32 and a prediction model generation unit 33.

[0064] The learning data acquisition unit 32 acquires learning data for learning the prediction model 34. The learning data acquisition unit 32 reads the learning data from the storage device 50 based on an input signal representing an instruction to start learning from the input device 40. Further, the learning data acquisition unit 32 may acquire the learning data input via the input device 40. The learning data acquisition unit 32 outputs the acquired learning data to the prediction model generation unit 33.

[0065] The prediction model generation unit 33 refers to the observed values of weather data before the prediction target date and generates a prediction model 34 that represents the production performance and the influence of weather, and predicts the production performance from the forecast value of the weather data on the prediction target date.

[0066] The prediction model generation unit 33 generates a prediction model 34 that represents the production results and the influence of the weather, with reference to the learning data including the observed values of the daily weather data and the measured values of the production results on those days. The prediction model generation unit 33 generates the prediction model 34 by performing machine learning using the learning data acquired from the learning data acquisition unit 22. As an example, the prediction model generation unit 33 generates the prediction model 34 using known machine learning methods such as neural networks, decision trees, random forests, and support vector machines.

[0067] Also, as an example, the prediction model generation unit 33 may generate a statistical model that represents the correlation between the daily weather data and the production results on those days. That is, the prediction model 34 may be a statistical model such as a linear regression model, sparse modeling, generalized linear model, state space model, hierarchical Bayesian model, time series model, or clustering.

[0068] The prediction model generation device 30 selects a combination of variables of the weather data so that the difference between the measured value of the production results in a certain past year and the predicted value of the production results in that year predicted using the prediction model 34 is minimized, and generates the prediction model 34. The weather data that is expected to affect the production results of agricultural crops includes many variables such as precipitation, average temperature, maximum temperature, minimum temperature, total daily solar radiation, and sunshine hours. Among the variables of the weather data, there are variables that are interdependent. As an example, on days when it rains, the temperature is lower (the relationship between precipitation and temperature), and there are combinations of highly interdependent variables. Therefore, by selecting the combination of variables of the weather data that is optimal for predicting the production results, a prediction model that can accurately predict the production results can be generated.

[0069] Regarding the influence of combinations of variables in meteorological data on production performance, an explanation will be given with reference to Fig. 4. Fig. 4 is a diagram explaining the influence of variables in meteorological data on production performance. In Fig. 4, the graph on the upper left shows the influence of precipitation on sugar content, the graph in the center of the upper row shows the influence of average temperature on sugar content, and the graph on the upper right shows the influence of sunshine hours on sugar content. In these graphs, the X-axis represents the date, the Y-axis represents the increase or decrease value of sugar content, and the Z-axis represents the measurement location. That is, these graphs show the influence of variables in meteorological data on production performance for each measurement location.

[0070] In Fig. 4, the lower graph is a graph obtained by extracting and overlapping only the graphs of a specific measurement location in the three upper graphs. That is, this graph collectively shows the influence of each variable in meteorological data on sugar content. In the lower graph, the solid line indicates the influence of precipitation, the dotted line indicates the influence of average temperature, and the ○-marked plot indicates the influence of sunshine hours.

[0071] As shown in the upper graphs of Fig. 4, the degree to which production performance is affected by variables varies for each location and each day. As shown in the lower graph of Fig. 4, by overlapping the graphs of each variable, it is clear that even at the same location and on the same day, the types and degrees of variables affected are different.

[0072] In this way, by selecting the optimal combination of variables for predicting production performance based on the degree of influence of variables in meteorological data on production performance, a prediction model that can predict production performance with higher accuracy can be generated. Also, by selecting the optimal combination of variables on a daily or location-by-location basis, a prediction model that can predict production performance with higher accuracy can be generated.

[0073] The prediction model generation unit 33 selects a combination of variables of the meteorological data by performing a regression analysis based on the respective integrated values or average values for each variable of the meteorological data before the prediction target date. The prediction model generation unit 33 calculates, for each variable, the integrated value obtained by integrating each over a predetermined period, or the average value within each predetermined period, and performs a regression analysis based on the calculated integrated value or average value to select a combination of variables that minimizes the difference between the measured value and the predicted value of the production results. As an example, the prediction model generation unit 33 selects an optimal combination of variables by performing a regression analysis using machine learning that can analyze variables with high interdependence.

[0074] The prediction model generation unit 33 selects a combination of variables of the meteorological data by performing a regression analysis based on the start date and end date for calculating the integrated value or average value for each variable, and the number of days of calculation. The influence of the weather on production results can vary not only on an annual basis but also on a short-term basis such as the flowering period or the fruit-setting period. Therefore, by performing a regression analysis based on the start date for starting the integration and the end date for ending the integration for each variable, or the start date for starting the calculation of the average value and the end date for ending the calculation of the average value for each variable, an optimal combination of variables can be selected for each period. Similarly, by performing a regression analysis based on the number of days for integrating or averaging (number of calculation days) for each variable, an optimal combination of variables can be selected in shorter time units.

[0075] The prediction model generation unit 33 selects a combination of variables of the meteorological data by performing regression analysis based on the integrated value or average value of the number of calculation days for each day from the start date to the end date for each variable of the meteorological data. In this way, by calculating the integrated value or average value of each variable for the number of calculation days for each day within the calculation period from the start date to the end date, the window for the number of calculation days is shifted one day at a time, and a vector representing the influence of each variable on the production results for each day within the calculation period can be created. Then, by performing regression analysis using the created daily vectors and selecting the combination of variables that minimizes the difference between the measured value and the predicted value of the production results, the combination of variables that is optimal for predicting the production results can be selected. Also, by performing regression analysis using the daily vectors, it is possible to select a combination of variables while considering the influence of the continuity of daily variables on the production results, such as the influence of consecutive days of intense heat.

[0076] The prediction model generation unit 33 is a prediction model that represents the secular change in production results and the influence of weather, generated by referring to measurement data including the measured value data of the production results for each past year and the observed value data of the meteorological data for those years. A prediction model 34 for predicting the production results may be generated from at least a part of the measured values of the production results up to the previous year and the forecast values of the meteorological data for the prediction target day. Also, such a prediction model 34 may be generated by performing machine learning using the measurement data as learning data. As a result, a prediction model 34 may be generated in which at least a part of the measured values of the production results up to the previous year and the forecast values of the meteorological data for the prediction target day are input information, and the production results for the current year are output information. That is, the prediction model generation unit 33 can generate a prediction model 34 that represents the secular change in production results and the influence of weather by regressing the influence of weather when the change in production results for each year is decomposed into the influence of secular change and the influence of weather in the prediction model generation unit 23 of the prediction model generation device 20 described above.

[0077] Further, it is preferable that the prediction model generation unit 33 generates a prediction model 34 for each prediction target location. The prediction model generation unit 33 refers to measurement data including the measured values of the production results for each past year at the prediction target location and the observed values of the meteorological data for those years at the prediction target location, and generates a prediction model 24 for each prediction target location. By generating a prediction model 34 for each prediction target location, which is an area where the crops for predicting the production results grow, it can be used for predicting the production results with higher accuracy.

[0078] According to the prediction model generation device 30, since the production results of the current year can be predicted from the forecast values of the meteorological data for the prediction target day, it can also be used to predict the production results of the harvest period or the shipping date of that year at the initial stage of the cultivation period. Thereby, by predicting the production results of that year at the initial stage of the cultivation period and changing the cultivation management such as fertilization, irrigation, and fruit picking based on the prediction results, it can be used to improve the production results of that year. Also, based on the predicted production results, it can be used to determine the optimal working time and harvest time.

[0079] (Prediction of Production Results in the Prediction System 100) Regarding the concept of the flow of the production result prediction process using the prediction model in the prediction system 100, it will be described with reference to FIG. 5. FIG. 5 is a diagram for explaining the concept of predicting the production results of crops using the prediction model, which is executed by the prediction system according to one aspect of the present invention. Note that the variables of the meteorological data shown in FIG. 5 are examples, and the variables of the meteorological data are not limited to this.

[0080] As shown in FIG. 5, a learned prediction model is used with measurement data including the measured value data of the production results for each past year and the observed value data of the weather data for those years. The measured values of the production results up to the previous year (past) and the forecast values of the weather data for the current year are input into the prediction model. The prediction model divides the predicted value of the production result into the reference production result for the current year and the influence of the weather for the current year. The prediction model regresses the influence of the weather for the current year based on a combination of variables of the weather data and selects the optimal combination of variables. The prediction model regresses the influence of the weather for the current year based on the selected combination of variables. The prediction model outputs the predicted value of the production result for the current year from the reference production result for the current year and the influence of the weather for the current year.

[0081] (Flow of Prediction Processing) The flow of the prediction process (prediction method) by the prediction device 10 will be described with reference to FIG. 6. FIG. 6 is a flowchart showing an example of the prediction process executed by the prediction device according to one aspect of the present invention. As shown in FIG. 6, first, the input data acquisition unit 12 acquires the measured values of the production results up to the previous year (past) (step S1). Next, the input data acquisition unit 12 acquires the forecast values of the weather data for the current year (step S2). Then, the prediction unit 13 inputs the acquired measured values of the production results up to the previous year and the forecast values of the weather data for the current year into the prediction model 24, and acquires the predicted value of the production result for the current year output (step S3, prediction step). The prediction unit outputs the acquired predicted value as a prediction result to the output device 60 (step S4) and ends the process.

[0082] (Flow of Prediction Model Generation Processing) The process of generating the prediction model 24 by the prediction model generation device 20 (prediction model generation method) will be described with reference to FIG. 7. FIG. 7 is a flowchart showing an example of the prediction model generation process executed by the prediction model generation device according to an aspect of the present invention. As shown in FIG. 7, first, the learning data acquisition unit 22 acquires the measured values of the production results for each past year (step S11). Next, the learning data acquisition unit 22 acquires the observed values of the weather data for those years (step S12). Then, the learning data acquisition unit 22 associates the measured values with the production results for each past year and the observed values of the weather data for those years to generate learning data (measurement data) (step S13). The prediction model generation unit 23 uses the generated learning data to generate a prediction model 24 in which the measured values of the production results up to the previous year (past) and the forecast values of the weather data for the current year are input information and the production result for the current year is output information (steps S14, generation step). The prediction model generation unit 23 stores the generated prediction model 24 in the storage device 50.

[0083] (Flow of prediction model generation process for weather data) The process of generating the prediction model 34 by the prediction model generation device 30 (prediction model generation method) will be described with reference to FIG. 8. FIG. 8 is a flowchart showing an example of the prediction model generation process executed by the weather data prediction model generation device according to an aspect of the present invention. As shown in FIG. 8, first, the prediction model generation unit 33 sets a starting date and an ending date for calculating the integrated value or average value of each variable of the weather data, and the number of days for integration or averaging (calculation days) (step S21). Next, the prediction model generation unit 33 calculates the integrated value or average value of the calculation days for each day from the starting date to the ending date for each variable (step S22). Then, the prediction model generation unit 33 regresses the prediction model 34 using the calculated values and obtains the error between the predicted value and the measured value (step S23). The prediction model generation unit 33 obtains the combination of variables, the starting date and ending date of the integrated value or average value, and the calculation days when the error is minimized (step S24). The prediction model generation unit 33 generates the prediction model 34 based on the obtained combination of variables, the starting date and ending date of the integrated value or average value, and the calculation days (step S25, generation step). The prediction model generation unit 33 stores the generated prediction model 34 in the storage device 50 (step S26).

[0084] Example of Realization by Software The functions of the prediction device 10, the prediction model generation device 20, and the prediction model generation device 30 (hereinafter referred to as "devices") can be realized by programs for causing a computer to function as these devices, and programs for causing a computer to function as each control block (particularly each part included in the control unit 11, the control unit 21, and the control unit 31) of these devices.

[0085] In this case, these devices include a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the above programs. By executing the above programs with this control device and storage device, each function described in the above embodiment is realized.

[0086] The above program may be recorded on one or more computer-readable recording media, rather than temporarily. These recording media may or may not be provided in these devices. In the latter case, the above program may be supplied to these devices via any wired or wireless transmission medium.

[0087] Also, part or all of the functions of each of the above control blocks can also be realized by a logic circuit. For example, an integrated circuit in which a logic circuit functioning as each of the above control blocks is formed is also included in the scope of the present invention. In addition to this, for example, it is also possible to realize the functions of each of the above control blocks by a quantum computer.

[0088] Also, each of the processes described in the above embodiments may be executed by AI (Artificial Intelligence). In this case, the AI may operate in the above control device, or may operate in another device (for example, an edge computer or a cloud server, etc.).

[0089] 〔Example〕 In mandarin oranges, the sugar content and acidity were predicted using a prediction model generated using past sugar content and acidity data at a plurality of prediction target points, and the accuracy was verified. The accuracy of sugar content prediction and acidity prediction was verified by comparing the measured values in 2020 with the predicted values using the prediction model. The meteorological observation values in 2018 and 2019, and the weather forecast values as of March 22, 2020 were used.

[0090] For all prediction target points and all prediction target days over two years, the RMSE (root mean square error) between the predicted value and the measured value was obtained. In sugar content prediction, when using meteorological observation values, the RMSE was 0.6561 (Figure 9). In sugar content prediction, when using weather forecast values, the RMSE was 1.5267 (Figure 10). Next, in acidity prediction, when using meteorological observation values, the RMSE was 0.2517 (Figure 11). In acidity prediction, when using weather forecast values, the RMSE was 0.2636 (Figure 12).

[0091] Also, based on the weather forecast values as of March 22, Reiwa 2, the sugar content and acidity for the Reiwa 2 fiscal year were predicted, and the accuracy was verified using the measured values up to November 9, Reiwa 2. In the accuracy verification up to November 9, Reiwa 2, the weather forecast values were used. The RMSE in sugar content prediction was 0.7355 (Figure 13). The RMSE in acidity prediction was 0.1615 (Figure 14).

[0092] Thus, it was shown that by using the prediction model, it is possible to predict with sufficient accuracy to grasp the trend of the production results for the following year as of March.

[0093] Figure 15 shows the influence of three weather variables by time period in the prediction of selecting fruits of high sugar content varieties for the Reiwa 2 fiscal year at a certain location to be predicted. In Figure 15, the solid line indicates the influence of precipitation, the dotted line indicates the influence of the average temperature, and the ○-marked plots indicate the influence of sunshine hours. Note that since the weather forecast values as of March 22, Reiwa 2 are used, it is different from the influence of the actual weather in the Reiwa 2 fiscal year. When the weather from late April onwards, which is the normal value, is normal, among precipitation, average temperature, and sunshine hours, the influence of precipitation on sugar content is the greatest. As shown in Figure 15, the precipitation up to around the end of June, Reiwa 2 had the effect of correcting the sugar content at the time of fruit selection upwards by up to 0.01 per day. On the other hand, the precipitation in July and August, Reiwa 2 had the effect of correcting the sugar content at the time of fruit selection downwards by 0.01 to up to about 0.025 per day. Although multi-sheet covering was carried out during this period, it was suggested that the influence of the precipitation in July and August was not zero.

[0094] 〔Other Embodiments〕 (Aspect of Using the Average Value of Production Results) The prediction device 10 may predict the production results of the current year from at least a part of the measured value data of the production results up to the previous year, using a prediction model that represents the average value of the production results, which is generated by referring to the measured value data of the production results for each past year. As an example, when the number of accumulated years of the production results for each past year is as small as about two to five years, instead of a prediction model that represents the secular change in the production results, a prediction model that represents the average value of the production results may be used. Here, the average value of the production results may be a value obtained by averaging the production results for each year over a predetermined number of years. Further, the prediction model 24 generated by the prediction model generation device 20 and the prediction model 34 generated by the prediction model generation device 30 may be prediction models that represent the average value of the production results and the influence of the weather, which are generated by referring to measurement data including the measured value data of the production results for each past year and the observed value data of the weather data for those years. Such prediction models 24 and 34 can predict the production results by considering the annual fluctuations of agricultural crops as average values.

[0095] That is, in each of the above-described embodiments, embodiments in which all the explanations regarding the secular change in the production results are read as the average value of the production results are also included in the technical scope of the present invention.

[0096] (Aspect of using the measured value data of soil components as measurement data) The prediction device 10 uses at least one of the measured value data of the production results for each past year, the observed value data of the weather data for those years, and the measured value data of the soil components for those years side Including measurement data and The production results of the current year may be predicted using a prediction model that represents the secular change or the average value of the production results and the influence of the weather, which is generated by referring to the measurement data. Here, the measured value data of the soil components may be, for example, soil component values (soil moisture values, soil pH values, etc.) measured by a soil sensor such as a soil moisture meter provided in a field where agricultural crops are grown.

[0097] The prediction model 24 generated by the prediction model generation device 20 is based on at least one of the measured value data of the production results for each past year, the observed value data of the meteorological data for those years, and the measurement data of the soil components. side Including measurement data and It may be a prediction model that represents the secular change or average value of production results and the influence of at least one of meteorology and soil components, generated with reference to the above. Such a prediction model 24 is generated by performing machine learning using the measurement data as learning data. At least a part of the measured values of the production results up to the previous year and at least one of the forecast values of the meteorological data for the current year and the measured values of the soil components for the current year are input information, and the production results for the current year are output values. According to the prediction model 24, the production results can be predicted considering the soil components.

[0098] The prediction model 24 is a prediction model generated for each prediction target location, and predicts the production results for the current year at the prediction target location from at least a part of the measured values of the production results up to the previous year at the prediction target location and at least one of the forecast values of the meteorological data for the current year and the measured values of the soil components for the current year at the prediction target location. The prediction model 24 may be regenerated using regression analysis that regresses the difference between the measured value of the production results for a certain past year and the predicted value of the production results for that year predicted by the prediction model 24 based on at least one of the observed value of the meteorological data for that year and the measured value of the soil components for that year.

[0099] The prediction model 34 generated by the prediction model generation device 30 is a prediction model that represents the production results and the influence of at least one of meteorology and soil components, generated with reference to at least one of the observed value data of the meteorological data and the measured value data of the soil components before the prediction target date, and can be a prediction model that predicts the production results from at least one of the forecast value of the meteorological data for the prediction target date and the measured value of the soil components.

[0100] Such a prediction model 34 is generated by selecting combinations of variables of meteorological data and combinations of variables of soil components so that the difference between the measured value of the production performance in a certain past year and the predicted value of the production performance in that year predicted using the prediction model 34 is minimized. These combinations of variables may be selected by regression analysis based on the integrated value or average value of each variable for the meteorological data and soil components before the prediction target date. Also, these combinations of variables may be selected by regression analysis based on the start date and end date for calculating the integrated value or average value, and the number of days of calculation. Furthermore, these combinations of variables may be selected by regression analysis based on the integrated value or average value for the number of days of calculation for each day from the start date to the end date for each variable of meteorological data and soil components.

[0101] Also, the prediction model 34 is based on at least one of the measured value data of the production performance for each past year, the observed value data of the meteorological data for those years, and the measured value data of the soil components for those years side Including measurement data and A prediction model that represents the secular change or average value of the production performance and the influence of at least one of meteorology and soil components, generated with reference to the measurement data, and predicts the production performance from at least a part of the measured values of the production performance up to the previous year and at least one of the forecast value of the meteorological data and the measured value of the soil components on the prediction target date. Such a prediction model 34 can be generated by performing machine learning using the measurement data as learning data. Such a prediction model 34 is a prediction model in which at least a part of the measured values of the production performance up to the previous year and at least one of the forecast value of the meteorological data and the measured value of the soil components on the prediction target date are input information, and the production performance of the current year is output information.

[0102] The influence of soil components on production performance will be described with reference to FIG. 16. FIG. 16 is a diagram explaining the influence of soil moisture value on production performance. The X-axis in the graph shown in FIG. 16 represents the date, and the Y-axis represents the influence value of the soil moisture value in terms of sugar content conversion. The graph shown in FIG. 16 shows the influence of the soil moisture value on the sugar content at harvest.

[0103] As shown in Fig. 16, the degree to which the soil moisture value affects the sugar content at harvest varies daily. As an example, according to Fig. 16, it can be seen that the soil moisture values from July to September have a great influence on the sugar content at harvest. This finding is consistent with the existing knowledge in the fruit tree field that the sugar content increases due to the drought stress in July and August, and the effect saturates thereafter.

[0104] In this way, by predicting the production results in consideration of the soil components, the production results can be predicted more accurately. In addition, by predicting the production generator in consideration of both the weather and the soil components, a further improvement in the prediction accuracy can be expected.

[0105] That is, in each embodiment described in this specification, embodiments in which all the descriptions regarding the meteorological data are rewritten as (i) the measured value data of the soil components or (ii) the meteorological data and the measured value data of the soil components are also included in the technical scope of the present invention.

[0106] (Aspect of selecting variables based on the information amount criterion) The selection of the combination of variables of the meteorological data and the combination of variables of the soil components in the prediction model 34 generated by the prediction model generation device 30 may be executed based on the information amount criterion. That is, in the method for generating the prediction model, an aspect of selecting the combination of variables of the meteorological data and the combination of variables of the soil components based on the information amount criterion and generating the prediction model 34 is also included in the technical scope of the present invention. Here, as the information amount criterion, a conventionally known information amount criterion can be used, and as an example, the Kullback-Leibler information amount criterion, the Akaike information amount criterion, the Bayesian information amount criterion, etc. can be mentioned.

[0107] As an example, when the model selection criterion is not included in the estimation algorithm of each prediction model generated in the combination of meteorological data variables and soil component variables, the combination of meteorological data variables and soil component variables may be selected based on the information criterion. This way, the operator's knowledge, proficiency, etc. have no influence on the variable selection process, and the variable selection process can be executed uniformly.

[0108] (Aspect of predicting the growth stage) The prediction device 10 may predict the growth stage of the agricultural crop by comparing the predicted production result with a reference value representing the result for each growth stage of the agricultural crop. Here, the growth stage of the agricultural crop is intended to be the flowering stage, harvesting stage, transplantation stage, etc. of the agricultural crop. The prediction of the growth stage in the prediction device 10 can be executed in the prediction unit 13. As an example, when the predicted value of the production result such as the sugar content or acidity obtained by the prediction model 24 exceeds a preset reference value (threshold value), the prediction unit 13 of the prediction device 10 predicts that the time when the production result is predicted is the harvesting time, assuming that the harvesting standard is met.

[0109] Further, the prediction device 10 may calculate the influence amount of at least one of meteorology and soil components in the prediction model, and predict the growth stage of the agricultural crop with reference to the calculated influence amount.

[0110] Among agricultural crops, there are agricultural crops that are grown indoors as seedlings and then transplanted outdoors, and agricultural crops that are stored in the shaded area with a roof after harvesting. Regarding such agricultural crops, in the prediction model generation unit 23 and the prediction model generation unit 33, if the influence amount (influence factor) of meteorology or soil components on the production result is calculated, the date when the influence amount discretely switches can be detected. For example, in the case of an agricultural crop stored in the shaded area with a roof after harvesting, the date when the influence amount of the sunshine duration suddenly becomes almost zero corresponds to the date when it is transferred to the storage environment with a roof after harvesting. Therefore, the prediction device 10 can predict the growth stage of the agricultural crop by referring to such an influence amount.

[0111] (Prediction device for predicting the growth stage) Note that at least one of the data representing the transition of the actual growth stage of crops for each past year, the observed value data of the meteorological data of those years, and the measured value data of the soil components of those years side Using a prediction model generated with reference to the measurement data including the above, which represents the transition of the growth stage and the influence of at least one of meteorology and soil components, from at least a part of the transition of the actual growth stage up to the previous year and at least one of the forecast value of the meteorological data of the current year and the measured value of the soil components of the current year, predicting the transition of the growth stage of the current year, a growth stage prediction device can also be included in the technical scope of the present invention. Here, as an example, the transition of the growth stage is intended to be information regarding the dates of switching of each growth stage of crops such as the flowering date, the harvest date, the transplanting date, etc.

[0112] That is, the prediction model used by such a growth stage prediction device is a growth stage prediction model representing the secular change or average value of the transition of the growth stage of crops and the influence of at least one of meteorology and soil components. The growth stage prediction device inputs at least a part of the transition of the actual growth stage up to the previous year and at least one of the forecast value of the meteorological data of the current year and the measured value of the soil components of the current year into the growth stage prediction model, and obtains the transition of the growth stage output. As an example, the growth stage prediction device can obtain prediction results such as the date of the flowering date, the number of days from a predetermined date to the flowering date, the number of days from the sowing date to the flowering date, etc. by using the growth stage prediction model.

[0113] Note that in the embodiments described in this specification, embodiments in which all the explanations regarding the prediction of production results are read as the prediction of the transition of the growth stage are also included in the technical scope of the present invention.

[0114] (Aspect of calculating the potential value of production results) Another form of the prediction device will be described with reference to FIG. 17. For convenience of explanation, members having the same functions as those described in the above-described prediction system 100 and prediction device 10 are denoted by the same reference numerals, and the description thereof will not be repeated. FIG. 17 is a block diagram showing a main configuration of a prediction system 200 according to another aspect of the present invention. The prediction system 200 is different from the prediction system 100 shown in FIG. 1 in that the control unit 211 of the prediction device 210 includes an input value generation unit 212, an output value acquisition unit 213, a setting unit 214, and a calculation unit 215.

[0115] The input value generation unit 212 generates a simulation input value of at least one of the observed value of past weather data and the measured value of soil components based on at least one of the observed value of past weather data and the measured value of soil components (step of generating an input value). The input value generation unit 212 replaces the data on a daily basis in at least one of the observed value of the weather data and the measured value of the soil components acquired by the input data acquisition unit 12 to generate a simulation input value. The data replacement is performed by replacing with data on the same day of another year in the same location in the past. That is, the simulation input value generated by the input value generation unit 212 may be data of at least one of the observed value of the weather data and the measured value of the soil components obtained by splicing data for multiple years on a daily basis. The input value generation unit 212 outputs the generated simulation input value to the output value acquisition unit 213.

[0116] The output value acquisition unit 213 acquires a simulation output value of production results predicted from the simulation input value using the prediction model 24 or the prediction model 34 (step of acquiring an output value). The output value acquisition unit 213 inputs the simulation input value to the prediction model 24 or the prediction model 34 and acquires the predicted value of the production results output as the simulation output value. The output value acquisition unit 213 outputs the acquired simulation output value to the setting unit 214.

[0117] The setting unit 214 sets the simulation input value when the obtained simulation output value is the highest as the potential value (the setting step). The setting unit 214 compares the simulation output values obtained for various simulation input values, and sets the simulation input value when the simulation output value is the highest as the potential value. That is, the setting unit 214 compares the production results predicted using a plurality of meteorological data or soil component data with different patchwork modes, and sets the potential value. The setting unit 214 outputs the set potential value to the calculation unit 215.

[0118] The calculation unit 215 calculates the simulation input value when the potential value is obtained as the optimal environmental value (the calculation step). That is, the calculation unit 215 calculates the meteorological data or soil component data when the best production result is obtained, with at least one of the meteorological conditions and soil component conditions for obtaining the production result, as the optimal environmental value.

[0119] In the prediction device 210, for at least one of the observed value of meteorological data and the measured value of soil components, the simulation input value obtained by swapping with the data of the same day of another year in the same location in the past is used as the virtual environmental condition, and the production result predicted by the virtual environmental condition is used as the potential value. Such a simulation input value is generated, for example, by performing a process of swapping "the maximum temperature, sunshine hours, and precipitation on March 3, 2010" with "the maximum temperature, sunshine hours, and precipitation on March 3, 2011" for all dates during the growth period. In this way, by comparing the production results for each virtual environmental condition, the most ideal environmental condition can be predicted.

[0120] Note that the input value generation unit 212 may determine which weather conditions or soil component conditions at which time to swap based on the information regarding the influence of weather or soil components on the production results obtained from the prediction model generation unit 23 and the prediction model generation unit 33. Further, the input value generation unit 212 may generate simulation input values using an algorithm that determines the data to be swapped based on the information regarding the influence of weather or soil components on the production results.

[0121] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope indicated in the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

Description of Reference Numerals

[0122] 10, 210 Prediction device 13 Prediction unit 20 Prediction model generation device 23 Prediction model generation unit (generation unit) 30 Prediction model generation device 33 Prediction model generation unit 100, 200 Prediction system

Claims

1. A method for generating a prediction model for predicting the production performance of agricultural crops, wherein the generation method is executed by a computer under the control of software, referring to the measured value data of the production performance on a specific day of each past year, a prediction model representing the secular change of the production performance, and generating a prediction model for predicting the production performance of the current year from at least a part of the measured values of the production performance up to the previous year, the computer executes a step of generating, In the step of generating, by performing machine learning using the measured value data of the production performance on the specific day of each year as learning data, a prediction model in which at least a part of the measured values of the production performance up to the previous year is input information and the production performance of the current year is output information is generated. Generation method.

2. In the step of generating, by performing machine learning using the measured value data of the production performance on the specific day of each year and measurement data including at least one of the observed value data of the weather data of those years and the measured value data of the soil components of those years as learning data, a prediction model representing the secular change of the production performance and the influence of at least one of weather and soil components, wherein at least a part of the measured values of the production performance up to the previous year and at least one of the forecast values of the weather data of the current year and the measured values of the soil components of the current year are input information, and the production performance of the current year is output information. The generation method according to claim 1, wherein a prediction model is generated.

3. In the step of generating, by performing machine learning using the measured value data of the production performance on the specific day of each year at the prediction target location and measurement data including at least one of the observed value data of the weather data of those years and the measured values of the soil components of those years at the prediction target location as learning data, a prediction model for each prediction target location is generated. The generation method according to claim 2.

4. The computer further executes a step of re-learning the prediction model so that the difference between the measured value of the production performance in a certain past year and the predicted value of the production performance in that year predicted using the prediction model is minimized. The generation method according to claim 2 or 3.

5. In the step of re-learning, a regression analysis is used to regress the difference between the measured value of the production performance in a certain past year and the predicted value of the production performance in that year predicted by the prediction model, based on at least one of the observed value of the meteorological data in that year and the measured value of the soil components in that year, to re-learn the prediction model. The generation method according to claim 4.

6. A generation device for a prediction model that predicts the production performance of agricultural crops, Comprising a generation unit that generates a prediction model that represents the secular change of the production performance by referring to the measured value data of the production performance on a specific day for each past year, and predicts the production performance of the current year from at least a part of the measured values of the production performance up to the previous year. The generation unit generates a prediction model by performing machine learning using the measured value data of the production performance on a specific day for each year as learning data, where at least a part of the measured values of the production performance up to the previous year is input information and the production performance of the current year is output information. The generation device.

7. A prediction model generation program for predicting the production performance of agricultural crops, A program for causing a computer to execute a step of generating a prediction model that represents the secular change of the production performance by referring to the measured value data of the production performance on a specific day for each past year, and predicts the production performance of the current year from at least a part of the measured values of the production performance up to the previous year. In the step of generating, a prediction model is generated by performing machine learning using the measured value data of the production performance on a specific day for each year as learning data, where at least a part of the measured values of the production performance up to the previous year is input information and the production performance of the current year is output information. The generation program.

Images