Analytical device, analytical method, and analytical program

Abstract

Analyze the factors that cause fluctuations in livestock productivity when they occur. [Solution] The analysis device comprises an analysis request receiving unit that receives requests to analyze livestock production results; an instruction input unit that inputs into the LLM the prerequisites for analysis included in the analysis request and instructions specifying what to analyze; and an analysis result acquisition unit that acquires the analysis results of production results output from the LLM. The instruction input unit inputs at least the following instructions to the LLM: to consider, as prerequisite knowledge when analyzing production results, a data set that defines the relationship between patterns that combine livestock-related data and events related to production results; to use actual values ​​of data related to the patterns and events when analyzing production results; and, if the actual value differs from the value of a data item included in the event associated with the data item of the actual value, to analyze the factors that caused the difference.

Description

[Technical Field]

[0001] The present invention relates to an analysis device, an analysis method, and an analysis program. [Background technology]

[0002] The following Patent Document 1 discloses a system that constantly monitors the rearing conditions of livestock and provides timely feedback to livestock farmers to support their management. This system is disclosed to estimate the cause of poor rearing conditions of livestock based on fluctuations in the rearing conditions of livestock.

[0003] Specifically, this system monitors the rearing conditions of livestock, such as increases or decreases in feed consumption, differences from ideal weight, and increases or decreases in water intake, and if it determines that the rearing conditions are poor, it estimates the causes of the poor conditions, such as disease, lack of nutritional value, malfunctioning feeders, malfunctioning waterers, poor ventilation, or inappropriate temperature management. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2021-111069 Summary of the Invention [Problem to be solved by the invention]

[0005] In livestock farming, production performance is an important factor for the success of livestock management. Therefore, when production performance fluctuates, it is effective to have a system to identify the cause.

[0006] However, the factors that cause fluctuations in production performance are not limited to the above-mentioned rearing conditions, but also involve a complex intertwining of various conditions related to livestock farming, such as the condition of the livestock products, the environment, the feed, the reproductive status, and the financial status. It takes a great deal of effort to manage these various conditions and accurately grasp the factors that cause fluctuations in production performance.

[0007] Therefore, an object of the present invention is to provide an analytical device, an analytical method, and an analytical program that can analyze the factors behind fluctuations in productivity related to livestock farming. [Means for solving the problem]

[0008] An analysis system according to one embodiment of the present invention comprises an analysis request receiving unit that receives a request to analyze livestock production results, an instruction input unit that inputs into a language model the prerequisites for the analysis and instructions specifying what to analyze, which are included in the received analysis request, and an analysis result acquisition unit that acquires the analysis results of the production results output from the language model, and the instruction input unit inputs into the language model at least the following instructions: to consider, as prerequisite knowledge when analyzing the production results, a data set that defines the relationship between patterns that combine data related to livestock production and events related to the production results, which is stored in the memory unit; to use the actual values ​​of the data related to the patterns and events stored in the memory unit when analyzing the production results; and, if the actual values ​​differ from the values ​​of data items included in the events associated with the data items of the actual values, to analyze the factors that caused the difference.

[0009] In addition, an analysis method according to another aspect of the present invention is a method executed by a processor, and includes a first step of receiving a request to analyze livestock production results; a second step of inputting into a language model the prerequisites for the analysis and instructions specifying what to analyze, which are included in the received analysis request; and a third step of acquiring the analysis results of the production results output from the language model, in which the second step inputs into the language model at least the following instructions: to consider, as prerequisite knowledge when analyzing the production results, a data set that defines the relationship between patterns that combine data related to livestock production, which is stored in a memory unit, and events related to the production results; to use the actual values ​​of the data related to the patterns and events stored in the memory unit when analyzing the production results; and, if the actual value differs from the value of a data item included in the event associated with the data item of the actual value, to analyze the factors that caused the difference.

[0010] Furthermore, an analysis program according to another aspect of the present invention causes a computer to function as an analysis request receiving unit that receives a request to analyze livestock production results, an instruction input unit that inputs into a language model the prerequisites for the analysis and instructions specifying what to analyze, which are included in the received analysis request, and an analysis result acquisition unit that acquires the analysis results of the production results output from the language model, and the instruction input unit inputs into the language model at least the following instructions: to consider, as prerequisite knowledge when analyzing the production results, a data set that defines the relationship between patterns that combine data related to livestock production, which is stored in the memory unit, and events related to the production results; to use the actual values ​​of the data related to the patterns and events stored in the memory unit when analyzing the production results; and, if the actual value differs from the value of a data item included in the event that is associated with the data item of the actual value, to analyze the factors that caused the difference.

[0011] According to these aspects, when instructions specifying the prerequisites for analysis and what to analyze are input into a language model based on a request to analyze production results, the instructions can include at least considering a dataset that defines the relationship between a pattern combining data related to livestock farming and events related to production results as prerequisite knowledge when analyzing production results, using the actual values ​​of the data related to the patterns and events when analyzing production results, and, if the actual values ​​differ from the values ​​of the data items included in the events associated with the data items of the actual values, analyzing the factors that caused the difference, and the analysis results of production results can be obtained from the language model into which these instructions have been input.

[0012] This makes it possible to obtain analytical results for production performance in which the factors that caused the difference are analyzed, taking into account a data set that defines the relationship between the patterns of combined livestock data and events related to production performance, when the actual values ​​differ from the values ​​included in events related to production performance. [Effects of the Invention]

[0013] According to the present invention, it is possible to provide an analysis device, an analysis method, and an analysis program that can analyze the factors behind fluctuations in productivity related to livestock farming. [Brief explanation of the drawings]

[0014] [Figure 1] FIG. 2 is a block diagram illustrating a physical configuration of an analysis device according to an embodiment. [Figure 2] FIG. 2 is a block diagram illustrating an example of the functional configuration of the analysis device. [Figure 3] 10 is a flowchart illustrating the operation of the analysis device. DETAILED DESCRIPTION OF THE INVENTION

[0015] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A preferred embodiment of the present invention will be described with reference to the accompanying drawings. In the drawings, components with the same reference numerals have the same or similar configurations.

[0016] [Analytical equipment configuration] An example of the physical configuration of an analysis device 1 according to an embodiment will be described with reference to Fig. 1. The analysis device 1 has a function of inputting actual results into a large language model (hereinafter also referred to as "LLM") to which a data set that defines the relationship between patterns combining data related to livestock farming (including dairy farming) and events related to livestock production performance is added as prerequisite knowledge for analyzing livestock production performance, and, if the actual results differ from the values ​​of a data set registered in advance, having the LLM analyze the factors that caused the difference. Details of this function will be described later. The analysis device 1 is connected to one or more user terminals 2 via a network N.

[0017] The analysis device 1 includes, as its physical components, a processor 11, a storage device 12, and a communication interface 13, for example.

[0018] The processor 11 is, for example, a CPU (Central Processing Unit) or an MPU (Micro Processing Unit), etc. The processor 11 executes a program 121 stored in the storage device 12 to realize various functions described below.

[0019] The storage device 12 is a computer-readable recording medium such as a disk drive or a semiconductor memory. The storage device 12 functions as a storage unit that stores a program 121 for implementing various functions of the analysis device 1, various data 122 used by the program 121, an LLM 123, and the like. Note that the LLM 123 does not necessarily have to be provided in the analysis device 1, but may also be provided in an external device or system.

[0020] The various data 122 includes, for example, data related to livestock farming and data sets described below.

[0021] Data related to livestock farming includes, for example, data on individual livestock, the number of individuals comprising a livestock herd such as a cow herd, livestock product data, production performance data, environmental data such as livestock sheds, feed data, breeding data, and financial data.

[0022] Individual livestock data can be managed for each identification information for identifying an individual livestock, and can include data items related to the individual livestock, such as test results and pedigree. Test results can include data items such as weight, body condition score (BCS), blood, feed intake (including feed design), genome, feces, medication, disease, and treatment.

[0023] In this embodiment, the description will be made mainly using cows as an example of livestock, but the livestock is not limited to cows. For example, the present invention can be applied to other livestock including pigs, birds, fish, etc. in the same way as cows.

[0024] The production performance data may include, as data items, data related to production performance, such as the quality of livestock products, production volume of livestock products, grading of livestock products, weight of livestock products, and reproductive performance. In the case of cattle, livestock products include, for example, beef, milk, and dairy products. Here, production performance data for dairy cattle includes, for example, milk quality, milk volume, milk price, number of culled head, number of shipped head, diseases, and herd dynamics. In the case of beef cattle, production performance data includes, for example, number of shipped head, grading, carcass unit price, carcass weight, number of culled head, diseases, and herd dynamics.

[0025] LLM123 is a generative AI model that generates and outputs content such as responses to input instructions (hereinafter also referred to as "prompts"). Note that the model is not limited to a large-scale language model, and may be a language model including a small language model, or a generative AI model that handles images and sounds in addition to language.

[0026] The communication interface 13 functions as a communication unit that connects to the network N and communicates with other devices on the network N. In this embodiment, it controls communication with one or more user terminals 2 operated by a user.

[0027] The functional configuration of the analysis device 1 according to the embodiment will be described with reference to Fig. 2. The analysis device 1 includes, as functional components for realizing various functions, a data collection unit 111, a data set setting unit 112, a performance data acquisition unit 113, an analysis request reception unit 114, an instruction input unit 115, an analysis result acquisition unit 116, an analysis result presentation unit 117, and an alert output unit 118. Each unit will be described below.

[0028] [Data Collection Department] The data collection unit 111 collects data related to livestock farming and stores it in the storage device 12. The collected data may include data managed by various systems related to livestock farming, such as a farm management system, data acquired by an IoT sensor, data recognized by an OCR (Optical Character Recognition / Reader), etc. The IoT sensor may be provided in a livestock barn or attached to the livestock. The collected data may be directly input from a user terminal 2 by a data manager or the like and stored in the storage device 12.

[0029] [Dataset Settings] The data set setting unit 112 stores a data set that defines the relationship between patterns of combining data related to livestock farming and events related to livestock production performance in the storage device 12. The data set is set based on collected data and is a collection of data that organizes how fluctuations in data related to livestock farming affect production performance.

[0030] Patterns combining livestock data can be set by any combination of, for example, data representing cattle characteristics such as weight, body condition score, and withers height; data representing evaluations of genetic ability such as genomic breeding value (GEBV) and predicted breeding value (PTA); data representing blood test results such as AST, ALT, NEFA, ketone bodies, and white blood cell count; and data representing nutritional information such as feed intake, crude protein (CP), non-digestible fiber (NDF), and energy (NEL).

[0031] Events related to livestock production performance can be set by combining patterns of livestock data, such as the quality of livestock products such as milk quality, the production volume of livestock products such as milk volume, the grading of livestock products such as carcasses, the weight of livestock products such as carcasses, or the reproductive performance of dairy cows.

[0032] [Actual data acquisition section] The performance data acquisition unit 113 acquires data related to livestock farming, including data corresponding to the data items set in the dataset, as performance data, and stores it in the storage device 12. The performance data can be acquired from various systems, IoT sensors, OCR, etc., similar to the data related to livestock farming collected by the data collection unit 111 described above.

[0033] [Analysis Request Reception Department] The analysis request receiving unit 114 receives a request to analyze livestock production results. The request to analyze production results includes, for example, the prerequisites for analyzing the production results and instructions specifying what to analyze. The analysis may also include predictions. The analysis request may be received from a user, or may be automatically received from the system based on the judgment of a program built into the analysis device 1, etc. Users include users who manage or operate the system, as well as users who receive the analysis results.

[0034] [Instruction input section] Based on the accepted analysis request, the instruction input unit 115 inputs instructions specifying the prerequisites for the analysis and what to analyze as prompts to the LLM 123. As instructions to be input to the LLM 123, it is preferable to use, for example, the instructions shown in (a) to (f) below.

[0035] (a) Considering a data set stored in the memory device 12, which defines the relationship between patterns of combined data on livestock farming and events related to livestock production performance, as background knowledge when analyzing production performance.

[0036] (b) Using the patterns of combinations of data on livestock farming stored in the storage device 12 and the actual values ​​of data related to events related to livestock production performance when analyzing production performance.

[0037] (c) When the actual value differs from the value of the data item in the data set associated with the data item of the actual value, analyze the factors that led to the difference.

[0038] (d) Building predictive models to forecast data on production performance.

[0039] (e) Suggesting improvements to improve production performance.

[0040] (f) Suggest data that should be monitored to prevent declines in production performance.

[0041] Here, specific examples of instructions to be input as prompts to the LLM 123 are described below in (1) to (5). Note that the instructions described in (1) to (5) are merely examples, and are not limited to the described contents.

[0042] (1) The instruction for the “purpose of the analysis” was to “identify the factors that cause increases or decreases in an individual’s daily milk yield and make specific improvement proposals to increase milk yield.”

[0043] (2) The following instructions as a "Data Summary": The following dataset contains various production data and environmental information from the farm. Using this data and the prerequisite knowledge described below, please analyze why the event occurred based on the characteristics and reasons for fluctuations in the milking history data for this farm. Please select and apply the most appropriate analysis method from the methods described in the analysis steps described below. - Cow basic data: $[JSON.stringify(cowStatus, null, 2)] - Milking history data: $[JSON.stringify(cowMilkingHistory, null, 2)] - Barn environment data: $[JSON.stringify(envInfo, null, 2)] - Feed data: $[JSON.stringify(feedInfo, null, 2)]"

[0044] (3) The "analysis steps" are divided into five steps: "Phase 0: Confirmation of prerequisite knowledge," "Phase 1: Data exploration and preprocessing," "Phase 2: Identification of patterns of increase or decrease," "Phase 3: Factor analysis," "Phase 4: Pattern classification," and "Phase 5: Construction of a predictive model." The instructions are to list and input the elements belonging to each step as follows:

[0045] Phase 0: Check your prerequisite knowledge: Please refer to the following as background knowledge when analyzing production results. - Relationship between body characteristics and milk production: $[JSON.stringify(BodyCharacteristicsAndMilkProductionRelation, null, 2)] - Genomic breeding value and productivity relationship: $[JSON.stringify(GenomicValueAndMilkProductionRelation, null, 2)] - Relationship between blood test results and disease and productivity: $[JSON.stringify(BloodTestAndDeseaseRelation, null, 2)] - Nutrition and Milk Production Relation: $[JSON.stringify(FeedNutritionAndMilkProductionRelation, null, 2)]"

[0046] Here, the prerequisite knowledge for Phase 0 constitutes the logic that organizes how fluctuations in livestock data affect production performance. This logic is used to construct a dataset that defines the relationship between patterns of combined livestock data and events related to livestock production performance.

[0047] Phase 1: Data exploration and preprocessing - Identifying missing and outlier values ​​in data quality checks and proposing treatment strategies - Visualize data distribution and trends - Calculation of basic statistics, individual and herd milk yield statistics - Descriptive statistics for each factor

[0048] Phase 2: Identifying patterns of increase and decrease: - Visualization of milk yield trends for individual cows through time series analysis - Classification into increasing trend, decreasing trend, and stable period - Extraction of lactation curve characteristics - Identify factors that cause fluctuations, and calculate the rate of change compared to the previous day, week, or month - Detection of sudden fluctuations (e.g., ±20% or more)

[0049] “Phase 3; Factor analysis: - Correlation analysis, calculation of correlation coefficient between each factor and milk yield increase / decrease - Multicollinearity check - Evaluating the importance of factors through regression analysis and multiple regression analysis - Calculating feature values ​​and importance using machine learning methods (random forest, XGBoost, etc.) - Segment analysis, differences in factors by parity (primiparous, multiparous) - Analysis by lactation stage (early, middle, late lactation) - Analysis by season and environmental conditions

[0050] Phase 4: Pattern Classification: Please group the individuals according to the following criteria (types): - High and stable production: Maintains stable high milk yield - Productivity improvement type: Continuous increase in milk yield - Declining production type: Milk yield tends to decrease - Fluctuating type: Milk production fluctuates drastically Please clarify the distinctive factors of each group.

[0051] Phase 5: Building a predictive model - Building a next day milk yield prediction model - Model accuracy assessment (RMSE, MAE, coefficient of determination) - Ranking of factors important for prediction

[0052] Here, the Phase 5 predictive model is constructed by LLM123 using the datasets listed in the "Data Overview" and the results of the analyses in the "Analysis Steps" from Phase 0 to Phase 4.

[0053] (4) The "Output Request" includes four output items: "A. Executive Summary," "B. Detailed Analysis Results," "C. Strategic Improvement Proposals," and "D. Proposals for Monitoring and Management Indicators." The instructions include inputting the elements belonging to each item as follows: When the "Output Request" is input, LLM123 will generate and output the information specified in the "Output Request" using the data set listed in the "Data Summary" and the results of the analysis in the "Analysis Steps."

[0054] A. Executive Summary: - Major findings (3-5 points) - Identifying individuals requiring urgent attention - Overall improvement potential"

[0055] Detailed analysis results: - Quantifying the impact of each factor on milk production by factor impact ranking - Indication of statistical significance - Individual analysis: Comparison of the characteristics of top and bottom productivity individuals - Identifying individuals with significant room for improvement - The influence of environmental factors: the relationship between THI (temperature and humidity index) and milk yield - Present optimal environmental conditions

[0056] C. Strategic Improvement Proposal: - Immediate improvements - Feed adjustment suggestions - Environmental management improvements - Mid- to long-term improvement measures - Optimizing the feeding management system - Strengthening individual management policy

[0057] D. Proposed monitoring and management indicators: - Indicators to check daily - Setting thresholds for issuing alerts

[0058] (5) Instructions for entering the following points as "Notes for analysis and proposals" are provided: "Be sure to display statistical significance (p-value)," "Make proposals that are feasible in actual dairy farming, not impractical theoretical values," "Make flexible proposals that take into account individual differences," and "Make proposals that also take into account economic efficiency (cost-effectiveness)."

[0059] [Analysis result acquisition section] The analysis result acquisition unit 116 acquires the analysis results of the production performance output from the LLM 123 to which instructions specifying the prerequisites for analysis and what to analyze have been input.

[0060] [Analysis result presentation section] The analysis result presentation unit 117 visualizes and presents the analysis results of production performance output from the LLM 123 in a format desired by the user. Users to whom the analysis results are presented may include, for example, livestock producers, livestock farmers, financial institutions, feed companies, logistics companies, veterinarians, and animal pharmaceutical companies.

[0061] The analysis results can be output in the form of a report, for example. In this case, based on values ​​such as production performance data output as the analysis results, causal inferences such as what specifically is happening and what is presumed to be the cause can be presented in the form of a report. Note that the display format is not limited to a report; for example, the analysis results can be displayed on the screen of the user terminal 2, and when a specific numerical value is hovered over, a summarized causal inference can be displayed as text.

[0062] [Alert output section] The alert output unit 118 outputs an alert when alert information indicating that an abnormality has been found in the data being monitored is output from the LLM 123. The alert may be output as a message such as text, or may be output using sound, video, or the like.

[0063] The data to be monitored and the thresholds may be arbitrarily set by the user, or may be set in a system program. For example, a user may input a monitoring instruction specifying the data to be checked and the thresholds as a prompt to the LLM 123. Alternatively, the monitoring instruction may be input to the LLM 123 by executing a program incorporating the monitoring instruction in which predetermined data and thresholds are set as the data to be monitored and the thresholds.

[0064] [Analyzer operation] An example of the operation of the analysis device 1 according to the embodiment will be described with reference to FIG.

[0065] First, the analysis request receiving unit 114 of the analysis device 1 receives a request for analysis of livestock production results from the user terminal 2 operated by the user (step S101).

[0066] Next, the instruction input unit 115 of the analysis device 1 inputs instructions specifying the prerequisites for analyzing the production results and what to analyze as a prompt to the LLM 123 based on the request for analyzing the production results received in step S101 above (step S102).

[0067] Next, the analysis result acquisition unit 116 of the analysis device 1 acquires the analysis results of the production performance output from the LLM 123 to which the instruction was input in step S102 (step S103).

[0068] Next, the analysis result presentation unit 117 of the analysis device 1 visualizes and presents the analysis results of the production performance acquired in step S103 in a format desired by the user (step S104), and then ends this operation.

[0069] As described above, according to the analysis device 1 of the embodiment, when instructions specifying the prerequisites for analysis and what to analyze are input to the LLM123 based on a request to analyze production results, the instructions can include at least the following: to consider a data set that defines the relationship between a pattern combining data related to livestock farming and events related to production results as prerequisite knowledge when analyzing production results; to use the actual values ​​of the data related to the patterns and events when analyzing production results; and, if the actual values ​​differ from the values ​​of the data items included in the events associated with the data items of the actual values, to analyze the factors that caused the difference, and the analysis results of production results can be obtained from the LLM123 to which these instructions have been input.

[0070] This makes it possible to obtain analytical results for production performance in which the factors that caused the difference are analyzed, taking into account a data set that defines the relationship between the patterns of combined livestock data and events related to production performance, when the actual values ​​differ from the values ​​included in events related to production performance.

[0071] Therefore, with the analysis device 1 according to the embodiment, when productivity in livestock farming fluctuates, it becomes possible to analyze the factors that cause the fluctuation.

[0072] The present invention is not limited to the above-described embodiment, and can be embodied in various other forms without departing from the spirit of the present invention. Therefore, the above-described embodiment is merely an example in all respects and should not be interpreted as being limiting. For example, the order of the above-described processing steps can be arbitrarily changed or executed in parallel as long as no contradiction occurs in the processing content. [Explanation of symbols]

[0073] 1...analysis device, 2...user terminal, 11...processor, 12...storage device, 13...communication interface, 111...data collection unit, 112...data set setting unit, 113...performance data acquisition unit, 114...analysis request reception unit, 115...instruction input unit, 116...analysis result acquisition unit, 117...analysis result presentation unit, 118...alert output unit, 121...program, 122...data, 123...large-scale language model (LLM), N...network

Claims

1. an analysis request receiving unit that receives a request for analysis of livestock production results; an instruction input unit that inputs into a language model the prerequisites for analysis and instructions specifying what to analyze, which are included in the accepted analysis request; an analysis result acquisition unit that acquires the analysis result of the production performance output from the language model; Equipped with The instruction input unit inputs at least the following as the instruction: A data set that defines the relationship between a pattern of combined data on livestock farming and an event related to the production performance, which is stored in a memory unit, is considered as background knowledge when analyzing the production performance; using the actual values ​​of the data related to the patterns and the events stored in the storage unit when analyzing the production performance; and When the performance value differs from the value of a data item included in the event associated with the data item of the performance value, analyzing the cause of the difference; input into the language model; Analyzer.

2. the instruction input unit further inputs, as the instruction, to the language model, an instruction to construct a prediction model for predicting data related to the production performance; The analytical device according to claim 1 .

3. the instruction input unit further inputs, as the instruction, to the language model, a suggestion of an improvement measure for improving the production performance. The analytical device according to claim 1 .

4. the instruction input unit further inputs, as the instruction, to the language model, a suggestion of a data item that should be monitored in order to prevent a decline in the production performance. The analytical device according to claim 1 .

5. A data collection unit that collects data related to livestock farming; A data set setting unit that stores the data set set based on the collected data related to livestock farming in the storage unit; The analytical device of claim 1 further comprising:

6. Further provided is a performance data acquisition unit that acquires performance values ​​of the livestock-related data and stores them in the storage unit. The analytical device according to claim 1 .

7. further comprising an analysis result presentation unit that visualizes and presents the analysis result of the production performance output from the language model in a format desired by a user. The analytical device according to claim 1 .

8. an alert output unit that outputs an alert when alert information indicating that an abnormality has been found in data of a predetermined monitoring target is output from the language model; The analytical device according to claim 1 .

9. The data related to livestock farming includes any one of livestock individual data, livestock product data, environmental data, feed data, breeding data, and financial data. The analytical device according to claim 1 .

10. The production performance includes any one of livestock product quality, livestock product production, livestock product grading, livestock product weight, and reproductive performance; The analytical device according to claim 1 .

11. 1. A processor-implemented method comprising: The first step is to accept a request for analysis of livestock production performance; a second step of inputting into a language model the preconditions for analysis and instructions specifying what to analyze, which are included in the accepted analysis request; a third step of acquiring an analysis result of the production performance output from the language model; Including, The second step includes, as the instruction, at least: A data set that defines the relationship between a pattern of combined data on livestock farming and an event related to the production performance, which is stored in a memory unit, is considered as background knowledge when analyzing the production performance; using the actual values ​​of the data related to the patterns and the events stored in the storage unit when analyzing the production performance; and When the performance value differs from the value of a data item included in the event associated with the data item of the performance value, analyzing the cause of the difference; input into the language model; Analysis method.

12. an analysis request reception unit that receives requests for analysis of livestock production results; an instruction input unit that inputs, into a language model, prerequisites for analysis and instructions specifying what to analyze, which are included in the accepted analysis request; an analysis result acquisition unit that acquires the analysis result of the production performance output from the language model; and make the computer function as The instruction input unit inputs at least the following as the instruction: A data set that defines the relationship between a pattern of combined data on livestock farming and an event related to the production performance, which is stored in a memory unit, is considered as background knowledge when analyzing the production performance; using the actual values ​​of the data related to the patterns and the events stored in the storage unit when analyzing the production performance; and When the performance value differs from the value of a data item included in the event associated with the data item of the performance value, analyzing the cause of the difference; input into the language model; Analysis program.

Images