Causal explanation support device, causal explanation support system, causal explanation support method and program

Abstract

This enables explanations of the causal relationship between leading indicators used by users to forecast product demand and the products being forecasted. [Solution] The causal explanation support device 1 includes an information acquisition unit 11 that acquires leading indicator name information indicating the name of a leading indicator in product demand forecasting, target product name information indicating the name of the product for which demand forecasting is to be made, and time lag information indicating the time lag from the occurrence of an event of a leading indicator in demand forecasting until demand for the product occurs; a prompt creation unit 12 that creates a causal explanation prompt that instructs the artificial intelligence unit to answer the causal relationship between the leading indicator of product demand forecasting and the target product in terms of a collection of sub-causes that explain the logic between cause and effect, based on the leading indicator name information, target product name information, and time lag information, and inputs it to the artificial intelligence unit; and an information output unit 14 that outputs causal information indicating the causal relationship between the leading indicator of product demand forecasting and the target product based on the answer from the artificial intelligence unit.

Description

Technical Field

[0001] The present disclosure relates to a causal explanation support device, a causal explanation support system, a causal explanation support method, and a program.

Background Art

[0002] In the manufacturing industry, when demand forecasting using external statistical values as leading indicators is incorporated into operations, an explanation of the relevance between the leading indicators and the product to be forecast may be required. However, humans cannot necessarily understand and explain the causal relationship between the leading indicators and the product to be forecast. Patent Document 1 discloses a monitoring support device that estimates and outputs causal relationships between pieces of time-series data of equipment states.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The device described in Patent Document 1 can analyze causal relationships between a limited number of pieces of time-series data, but since this method requires brute-force analysis of the data, it is not possible to estimate causal relationships with the product to be forecast for a huge number of events in the world that are used as leading indicators for product demand forecasting.

[0005] The present disclosure has been made to solve the above problems, and an object thereof is to enable a user to explain the causal relationship between a leading indicator for product demand forecasting and the product to be forecast.

Means for Solving the Problems

[0006] To achieve the above objective, the causal explanation support device relating to this disclosure comprises an information acquisition unit, a prompt creation unit, and an information output unit. The information acquisition unit acquires leading indicator name information indicating the name of a leading indicator in product demand forecasting, target product name information indicating the name of the product for which demand forecasting is to be performed, and time lag information indicating the time lag from the occurrence of the leading indicator event in demand forecasting until the demand for the product occurs. Based on the leading indicator name information, target product name information, and time lag information, the prompt creation unit creates a causal explanation prompt that instructs the artificial intelligence unit to respond with a set of sub-causes that explain the intermediate logic between cause and effect regarding the causal relationship between the leading indicator of product demand forecasting and the target product, and inputs it to the artificial intelligence unit. Based on the response from the artificial intelligence unit, the information output unit outputs causal information indicating the causal relationship between the leading indicator of product demand forecasting and the target product. [Effects of the Invention]

[0007] According to this disclosure, by outputting causal information that shows the causal relationship between leading indicators for product demand forecasting and the products being forecasted, it becomes possible for users to explain the causal relationship between leading indicators for product demand forecasting and the products being forecasted. [Brief explanation of the drawing]

[0008] [Figure 1] A diagram showing an example configuration of the causal explanation support system according to Embodiment 1. [Figure 2] A diagram showing an example of a causal explanation prompt according to Embodiment 1. [Figure 3] Flowchart showing the causal explanation support process according to Embodiment 1 [Figure 4] A diagram showing an example configuration of the causal explanation support system according to Embodiment 2. [Figure 5] Flowchart showing the causal explanation support process according to Embodiment 2 [Figure 6] This figure shows an example configuration of the causal explanation support system according to Embodiment 3. [Figure 7] Flowchart showing the causal explanation support process according to Embodiment 3 [Figure 8]A diagram showing an example configuration of the causal explanation support system according to Embodiment 4. [Figure 9] Flowchart showing the causal explanation support process according to Embodiment 4 [Figure 10] A diagram showing an example of the hardware configuration of the causal explanation support device according to Embodiments 1 to 4. [Modes for carrying out the invention]

[0009] The causal explanation support device, causal explanation support system, causal explanation support method, and program according to this embodiment will be described in detail below with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals. The causal explanation support device outputs causal information showing the causal relationship between leading indicators for product demand forecasting and the product being forecasted. In the following description, the causal relationship between leading indicators for product demand forecasting and the product being forecasted consists of a collection of sub-causes that explain the intermediate logic between cause and effect, such as event A causing event B, and event B causing event C.

[0010] (Embodiment 1) The configuration of the causal explanation support system 100 and the functional configuration of the causal explanation support device 1 according to Embodiment 1 will be explained with reference to Figure 1. As shown in Figure 1, the causal explanation support system 100 consists of a demand forecasting device 2 that forecasts product demand and a causal explanation support device 1 that outputs causal information showing the causal relationship between leading indicators of product demand forecasts and the products to be forecasted.

[0011] The causal explanation support device 1 includes an information acquisition unit 11 that acquires leading indicator name information indicating the name of a leading indicator in demand forecasting, target product name information indicating the name of the product for which demand forecasting is to be made, and time lag information indicating the time lag from the occurrence of an event of a leading indicator in demand forecasting until demand for the product occurs; a prompt creation unit 12 that creates a causal explanation prompt instructing the artificial intelligence unit 13 to answer the causal relationship between the leading indicator of product demand forecasting and the target product as a collection of subcauses based on the leading indicator name information, target product name information, time lag information and example information indicating an example of causal explanation, and inputs it to the artificial intelligence unit 13; an artificial intelligence unit 13 that executes the causal explanation prompt input from the prompt creation unit 12 and answers the causal relationship between the leading indicator of product demand forecasting and the target product; and an information output unit 14 that outputs causal information indicating the causal relationship between the leading indicator of product demand forecasting and the target product based on the answer from the artificial intelligence unit 13. The artificial intelligence unit 13 is, for example, a text generation AI using GPT (Generative Pre-trained Transformer).

[0012] The information acquisition unit 11 acquires leading indicator name information, target product name information, and time lag information from the demand forecasting device 2. Note that the leading indicator name information, target product name information, and time lag information are not limited to the demand forecasting device 2; they may also be acquired from external devices or systems, or input by the user into the causal explanation support device 1. The prompt creation unit 12 creates a causal explanation prompt based on the leading indicator name information, target product name information, time lag information, and case information, instructing the system to respond with a set of sub-causes explaining the causal relationship between the leading indicators for product demand forecasting and the target product. Case information may, for example, be stored in advance, acquired from external devices or systems, or input by the user into the causal explanation support device 1. Case information is, for example, information explaining the causal relationship of the phenomenon "When the wind blows, the barrel maker profits."

[0013] Here, the causal explanation prompt will be described using Figure 2. In the example shown in Figure 2, the causal explanation prompt describes "remote work penetration rate" as the leading indicator name, "insulation material" as the target product name, "8 months later" as the time lag, and "the causality of the phenomenon that if the wind blows, the cooper barrel maker profits" as the causal explanation example, and gives an instruction to explain the causality between the leading indicator for product demand forecasting and the forecast target product using a set of sub-causalities. In the example of Figure 2, the causal explanation prompt is created as a one-shot prompt.

[0014] Returning to Figure 1, the prompt creation unit 12 inputs the created causal explanation prompt to the artificial intelligence unit 13. The artificial intelligence unit 13 executes the causal explanation prompt input from the prompt creation unit 12, and answers the causality between the leading indicator for product demand forecasting and the forecast target product. The causality between the leading indicator for product demand forecasting and the forecast target product is answered as a set of sub-causalities arranged in order from the occurrence of the event serving as the leading indicator to the occurrence of demand for the forecast target product. The information output unit 14 generates and outputs causal information indicating the causality between the leading indicator for product demand forecasting and the forecast target product, based on the answer from the artificial intelligence unit 13. The output method of the causal information may be, for example, screen display, or transmission to a terminal used by the user.

[0015] Next, the flow of causal explanation support processing executed by the causal explanation support device 1 will be described using Figure 3. The causal explanation support processing shown in Figure 3 starts, for example, when the causal explanation support device 1 is activated. When the information acquisition unit 11 of the causal explanation support device 1 does not acquire the leading indicator name information, target product name information and time lag information (step S11; NO), the processing proceeds to step S17. When the information acquisition unit 11 has acquired the leading indicator name information, target product name information and time lag information (step S11; YES), the prompt creation unit 12 creates a causal explanation prompt that instructs to answer the causality between the leading indicator for product demand forecasting and the forecast target product as a set of sub-causalities, based on the leading indicator name information, target product name information, time lag information and example information (step S12), and inputs it to the artificial intelligence unit 13 (step S13).

[0016] The artificial intelligence unit 13 executes the causal explanation prompt input from the prompt creation unit 12, and returns an answer regarding the causality between the leading indicator for product demand forecasting and the target product to be forecasted (step S14). The information output unit 14 generates causal information indicating the causality between the leading indicator for product demand forecasting and the target product to be forecasted based on the answer from the artificial intelligence unit 13 (step S15), and outputs the causal information (step S16). If the power is not turned off (step S17; NO), the process returns to step S11, and repeats steps S11 to S17. If the power is turned off (step S17; YES), the process ends.

[0017] According to the causal explanation support device 1 according to Embodiment 1, by outputting the causal information indicating the causality between the leading indicator for product demand forecasting and the target product to be forecasted, it becomes possible for the user to explain the causal relationship between the leading indicator for product demand forecasting and the target product to be forecasted. In addition, the user can confirm the certainty of the product demand forecast by viewing the output causal information and understanding the causal relationship between the leading indicator and the target product to be forecasted.

[0018] (Embodiment 2) In Embodiment 1, the artificial intelligence unit 13 is a general text generation AI, but in Embodiment 2, the text generation AI of the artificial intelligence unit 13 is fine-tuned using information related to the leading indicator for product demand forecasting and the target product to be forecasted. In the following description, parts different from Embodiment 1 will be mainly described.

[0019] The functional configuration of the causal explanation support device 1 according to Embodiment 2 will be described with reference to Figure 4. As shown in Figure 4, the causal explanation support device 1 includes, in addition to an information acquisition unit 11, a prompt creation unit 12, an artificial intelligence unit 13, and an information output unit 14, a related information acquisition unit 15 that acquires related information related to the leading indicator for product demand forecasting and the target product to be forecasted, and a fine-tuning unit 16 that fine-tunes the text generation AI of the artificial intelligence unit 13 using the related information.

[0020] The fine-tuning unit 16 fine-tunes the text generation AI of the artificial intelligence unit 13 using the related information acquired by the related information acquisition unit 15. Related information includes, for example, industry reports, company reports, and social situation reports. Company reports include, for example, information such as supply chains. Related information may be acquired from an external device or system, or the user may input it into the causal explanation support device 1. Alternatively, it may be acquired from the demand forecasting device 2. The information output unit 14 generates and outputs causal information showing the causal relationship between leading indicators for product demand forecasting and the product being forecasted, based on the response from the artificial intelligence unit 13 that has been fine-tuned with the related information.

[0021] Next, the flow of the causal explanation support process executed by the causal explanation support device 1 will be explained using Figure 5. The causal explanation support process shown in Figure 5 starts, for example, when the causal explanation support device 1 is activated. The related information acquisition unit 15 of the causal explanation support device 1 acquires leading indicators for product demand forecasting and related information related to the product to be forecasted (step S21). The fine-tuning unit 16 fine-tunes the text generation AI of the artificial intelligence unit 13 using the related information (step S22). Steps S23 to S29 are the same as steps S11 to S17 shown in Figure 3, so the explanation will be omitted.

[0022] According to the causal explanation support device 1 of Embodiment 2, by fine-tuning the text generation AI of the artificial intelligence unit 13 with relevant information, the artificial intelligence unit 13 becomes capable of answering the causal relationship between leading indicators for product demand forecasting and the target product based on more specialized information, thereby improving the accuracy of the answer.

[0023] (Embodiment 3) In Embodiment 2, causal information was generated and output based on the causal relationship between the leading indicators of product demand forecasts and the target product, as provided by the finely tuned artificial intelligence unit 13. In Embodiment 3, however, fact-checking is performed on the responses from the finely tuned artificial intelligence unit 13, and causal information is generated and output based on the causal relationship between the leading indicators of product demand forecasts and the target product, which are determined to be correct. The following description will mainly explain the differences from Embodiment 2.

[0024] The functional configuration of the causal explanation support device 1 according to Embodiment 3 will be explained with reference to Figure 6. As shown in Figure 6, the causal explanation support device 1 includes an information acquisition unit 11, a prompt creation unit 12, an artificial intelligence unit 13, an information output unit 14, a related information acquisition unit 15, a fine-tuning unit 16, and a fact-checking unit 17 that receives a response from the artificial intelligence unit 13 and performs fact-checking.

[0025] The causal relationship between the leading indicators and the target product for product demand forecasting, as provided by the artificial intelligence unit 13, is a collection of sub-causal relationships arranged in order from the occurrence of the leading indicator event to the occurrence of demand for the target product.

[0026] The fact-checking unit 17 chunks industry reports, company reports, social affairs reports, and other related information acquired by the related information acquisition unit 15, and stores the divided chunks in the chunk database 171. The fact-checking unit 17 creates a searchable vector store 172 by embedding the divided chunks. Note that the industry reports, company reports, social affairs reports, etc. that the fact-checking unit 17 chunks and embeds are not limited to related information acquired by the related information acquisition unit 15, but may also be acquired from other devices or systems, the internet, etc.

[0027] The fact-checking unit 17 creates sub-causal search queries for each of the sub-causal relationships that constitute the causal relationship between the leading indicators of the product demand forecast and the product being forecasted, as provided by the artificial intelligence unit 13. A sub-causal search query is a sentence that shows the cause and effect of a sub-causal relationship, such as "A caused B to occur." The fact-checking unit 17 searches the vector store 172 using the sub-causal search queries to identify sentences that contain the events of the cause and effect of the sub-causal relationship. The fact-checking unit 17 uses the search results from the vector store 172 and the chunks stored in the chunk DB 171 to create a sub-causal determination prompt that determines whether the sub-causal relationship is correct or not.

[0028] The fact-checking unit 17 inputs a sub-causal determination prompt to the artificial intelligence unit 13. The sub-causal determination prompt, for example, first describes the sub-causal relationship and includes a statement such as, "Determine whether the above causal relationship is true or false based on the following context. However, only the context should be used, and general knowledge should not be used," and includes a sentence listing the events of the cause and effect of the sub-causal relationship. The artificial intelligence unit 13 executes the sub-causal determination prompt and responds whether the sub-causal relationship is true or false.

[0029] If there are subcausal relationships that are determined to be incorrect, the fact-checking unit 17 determines that there is no causal relationship between the leading indicators for product demand forecasting and the product being forecasted, if the number of subcausal relationships determined to be incorrect exceeds a threshold, and notifies the information output unit 14. When the information output unit 14 is notified that there is no causal relationship between the leading indicators for product demand forecasting and the product being forecasted, it generates and outputs error information indicating that there is no causal relationship between the leading indicators for product demand forecasting and the product being forecasted.

[0030] If the number of subcausal relationships determined to be incorrect does not exceed a threshold, the fact-checking unit 17 sends the incorrect subcausal relationships to the prompt generation unit 12. Upon receiving the incorrect subcausal relationships, the prompt generation unit 12 creates a causal explanation correction prompt that includes information indicating the incorrect subcausal relationships and instructs the prompt generation unit to explain the causal relationship between the leading indicators for product demand forecasting and the target product. The prompt generation unit 12 inputs the created causal explanation correction prompt to the artificial intelligence unit 13. The artificial intelligence unit 13 executes the causal explanation correction prompt input from the prompt generation unit 12 and responds with the causal relationship between the leading indicators for product demand forecasting and the target product, excluding the incorrect subcausal relationships. The above process is repeated until all subcausal relationships are determined to be correct.

[0031] The prompt generation unit 12 sends the causal relationship between the leading indicators of the demand forecast for the product for which all subcausal relationships have been determined to be correct and the product being forecasted to the information output unit 14. The information output unit 14 generates and outputs causal information showing the causal relationship between the leading indicators of the demand forecast for the product for which all subcausal relationships have been determined to be correct and the product being forecasted.

[0032] Next, the flow of the causal explanation support process executed by the causal explanation support device 1 will be explained using Figure 7. The causal explanation support process shown in Figure 7 starts, for example, when the causal explanation support device 1 is activated. Steps S31 to S34 are the same as steps S21 to S24 shown in Figure 5, so their explanation will be omitted.

[0033] The prompt creation unit 12 of the causal explanation support device 1 inputs the created causal explanation prompt to the artificial intelligence unit 13 (step S35). The artificial intelligence unit 13 executes the causal explanation prompt and responds with the causal relationship between the leading indicators for product demand forecasting and the product to be forecasted (step S36). The fact-checking unit 17 repeats the following process for the sub-causal relationships that constitute the causal relationship between the leading indicators for product demand forecasting and the product to be forecasted, as responded by the artificial intelligence unit 13 (step S37). The fact-checking unit 17 creates a sub-causal search query (step S38). The fact-checking unit 17 searches the vector store 172 using the sub-causal search query (step S39) and identifies the text containing the cause and effect events of the sub-causal relationship.

[0034] The fact-checking unit 17 uses the search results from the vector store 172 and the chunks stored in the chunk DB 171 to create a sub-causal determination prompt to determine whether the sub-causal relationship is correct or not (step S40). The fact-checking unit 17 inputs the sub-causal determination prompt to the artificial intelligence unit 13 (step S41). The artificial intelligence unit 13 executes the sub-causal determination prompt and responds whether the sub-causal relationship is correct or not (step S42).

[0035] The fact-checking unit 17 determines whether all subcausal relationships are true or false, and when it finishes repeating the process (step S43), if there are any subcausal relationships that are determined to be false (step S44; NO), it determines whether the number of false subcausal relationships exceeds a threshold (step S45). If the number of false subcausal relationships does not exceed a threshold (step S45; NO), the fact-checking unit 17 sends the false subcausal relationships to the prompt creation unit 12. When the prompt creation unit 12 receives the false subcausal relationships, it creates a causal explanation correction prompt that includes content indicating the false subcausal relationships and instructs the prompt to explain the causal relationship between the leading indicators for product demand forecasting and the product being forecasted (step S46). The process returns to step S35, and steps S35 to S46 are repeated.

[0036] If the number of incorrect subcausal relationships exceeds the threshold (step S45; YES), the fact-checking unit 17 notifies the information output unit 14 that there is no causal relationship between the leading indicators for product demand forecasting and the product being forecasted. The information output unit 14 generates error information indicating that there is no causal relationship between the leading indicators for product demand forecasting and the product being forecasted (step S47), and outputs the error information (step S48). The process then proceeds to step S51.

[0037] If all subcausal relationships are determined to be correct (step S44; YES), the fact-checking unit 17 sends the causal relationship between the leading indicators for the demand forecast of the product for which all subcausal relationships were determined to be correct and the product being forecasted to the information output unit 14. The information output unit 14 generates causal information showing the causal relationship between the leading indicators for the demand forecast of the product for which all subcausal relationships were determined to be correct and the product being forecasted (step S49), and outputs the causal information (step S50). If the power is not turned off (step S51; NO), the process returns to step S33 and steps S33 to S51 are repeated. If the power is turned off (step S51; YES), the process ends.

[0038] According to the causal explanation support device 1 of Embodiment 3, by fact-checking the response of the artificial intelligence unit 13, the reliability of the causal relationship between the leading indicators of product demand forecasting indicated by the causal information and the product being forecasted is improved.

[0039] (Embodiment 4) In Embodiment 3, the response from the finely tuned artificial intelligence unit 13 was fact-checked using the artificial intelligence unit 13, whereas in Embodiment 4, fact-checking is performed using statistical data. The following description will mainly explain the differences from Embodiment 3.

[0040] The functional configuration of the causal explanation support device 1 according to Embodiment 4 will be explained with reference to Figure 8. As shown in Figure 8, the causal explanation support device 1 comprises an information acquisition unit 11, a prompt creation unit 12, an artificial intelligence unit 13, an information output unit 14, a related information acquisition unit 15, a fine-tuning unit 16, and a fact-checking unit 17.

[0041] The fact-checking unit 17 includes a statistical data storage unit 173 that pre-stores statistical data for various events. The statistical data is time-series data linked to statistical data names, such as the daily weather in XX prefecture in 2024, or the daily number of automobiles sold by △△ company in 2024. The fact-checking unit 17 calculates the relationship between the causal event and the resulting event of a subcausal relationship by analyzing the time-series data. First, since the causal event and the resulting event of a subcausal relationship are expressed in words and are not time-series data, the fact-checking unit 17 uses natural language processing to identify statistical data names that are close to the names of the causal and resulting events of the subcausal relationship. This allows the time-series data linked to the statistical data names to be extracted from the statistical data storage unit 173. The process of identifying statistical data names for the names of the causal and resulting events of the subcausal relationship may, for example, be done by word search, or an AI model such as BERT (Bidirectional Encoder Representations from Transformers) may be used to calculate the similarity between the names of the causal and resulting events of the subcausal relationship and the names of the statistical data, and those with high similarity may be extracted. Furthermore, if there are multiple statistical data points where the names and words of the cause and effect of each subcausal link match or are highly similar, you may select multiple options.

[0042] The fact-checking unit 17 performs statistical analysis using statistical data corresponding to the cause and the result of a subcausal relationship. Here, an example of correlation analysis is described. Because there is a time lag between the occurrence of the cause of a subcausal relationship and the occurrence of the result, the fact-checking unit 17 performs a time shift of k months so that when the statistical data corresponding to the cause is Xt, the statistical data corresponding to the corresponding result is Y(tk), and then performs correlation analysis. k is k≧0, and if, for example, the time lag between the occurrence of the cause of a subcausal relationship and the occurrence of the result is known, that time lag may be set as the upper limit. If the correlation coefficient exceeds the threshold α, the fact-checking unit 17 determines that the subcausal relationship is correct, and if the correlation coefficient does not exceed the threshold α, the fact-checking unit 17 determines that the subcausal relationship is incorrect. Furthermore, the fact-checking unit 17 may determine whether the subcausal relationship is correct or not using other logic, such as the Euclid distance between time series data or statistical causal inference, rather than just correlation analysis. Also, the fact-checking unit 17 may obtain statistical data from another device or system, the internet, etc.

[0043] Next, the flow of the causal explanation support process executed by the causal explanation support device 1 will be explained using Figure 9. The causal explanation support process shown in Figure 9 starts, for example, when the causal explanation support device 1 is activated. In the example shown in Figure 9, correlation analysis is used as the statistical analysis between the statistical data corresponding to the cause and the statistical data corresponding to the effect of the subcauses. Steps S61 to S67 are the same as steps 31 to S37 shown in Figure 7, so the explanation will be omitted.

[0044] The fact-checking unit 17 of the causal explanation support device 1 uses natural language processing to identify statistical data names that are close to the names of the cause and effect of each subcausal link (step S68). The fact-checking unit 17 extracts statistical data corresponding to the cause and the effect of the subcausal link from the statistical data storage unit 173 (step S69). The fact-checking unit 17 performs a correlation analysis on the statistical data corresponding to the cause and the effect of the subcausal link (step S70). The fact-checking unit 17 determines whether the subcausal link is correct or not based on whether the correlation coefficient exceeds the threshold α (step S71). Steps S72 to S80 are the same as steps 43 to S51 shown in Figure 7, so their explanation is omitted.

[0045] According to the causal explanation support device 1 of Embodiment 4, by fact-checking the response of the artificial intelligence unit 13, the certainty of the causal relationship between the leading indicators of product demand forecasting indicated by the causal information and the product being forecasted is improved.

[0046] The hardware configuration of the causal explanation support device 1 will be explained using Figure 10. As shown in Figure 10, the causal explanation support device 1 includes a temporary storage unit 301, a storage unit 302, a calculation unit 303, an input unit 304, a transmitting / receiving unit 305, and a display unit 306. The temporary storage unit 301, storage unit 302, input unit 304, transmitting / receiving unit 305, and display unit 306 are all connected to the calculation unit 303 via a BUS.

[0047] The calculation unit 303 is, for example, a CPU (Central Processing Unit). The calculation unit 303 executes the processes of the prompt creation unit 12, the artificial intelligence unit 13, the information output unit 14, the fine tuning unit 16, and the fact checking unit 17 according to the control program stored in the memory unit 302.

[0048] The temporary storage unit 301 is, for example, RAM (Random-Access Memory). The temporary storage unit 301 loads the control program stored in the storage unit 302 and uses it as a work area for the calculation unit 303.

[0049] The storage unit 302 is a non-volatile memory such as flash memory, hard disk, DVD-RAM (Digital Versatile Disc - Random Access Memory), or DVD-RW (Digital Versatile Disc - ReWritable). The storage unit 302 pre-stores a program for causing the calculation unit 303 to perform processing of the causal explanation support device 1, and also supplies the information stored by this program to the calculation unit 303 according to the instructions of the calculation unit 303, and stores the information supplied from the calculation unit 303. The chunk DB 171, vector store 172, and statistical data storage unit 173 are configured in the storage unit 302.

[0050] The input unit 304 is an interface device that connects input devices such as keyboards, pointing devices, and voice input devices to a BUS. Information entered by the user is supplied to the calculation unit 303 via the input unit 304. In a configuration where the user inputs leading indicator name information, target product name information, and time lag information to the information acquisition unit 11, the input unit 304 functions as the information acquisition unit 11. In a configuration where the user inputs related information to the related information acquisition unit 15, the input unit 304 functions as the related information acquisition unit 15.

[0051] The transmitting / receiving unit 305 is a network termination device or wireless communication device connected to a network, and a serial interface or LAN (Local Area Network) interface connected to them. The transmitting / receiving unit 305 functions as an information acquisition unit 11, an information output unit 14, and a related information acquisition unit 15.

[0052] The display unit 306 is a display device such as an LCD (Liquid Crystal Display) or an organic EL (electroluminescence) display. In a configuration where the information output unit 14 displays causal information on the screen, the display unit 306 functions as the information output unit 14.

[0053] The information acquisition unit 11, prompt creation unit 12, artificial intelligence unit 13, information output unit 14, related information acquisition unit 15, fine tuning unit 16, and fact checking unit 17 of the causal explanation support device 1 shown in Figures 1, 4, and 6 are executed by a control program that uses resources such as the temporary storage unit 301, calculation unit 303, storage unit 302, input unit 304, transmission / reception unit 305, and display unit 306 to perform the processing.

[0054] Furthermore, the aforementioned hardware configuration and flowchart are examples only and can be changed and modified as needed.

[0055] The core components of the causal explanation support device 1, such as the calculation unit 303, temporary storage unit 301, storage unit 302, input unit 304, transmission / reception unit 305, and display unit 306, can be implemented using a standard computer system, rather than a dedicated system. For example, the causal explanation support device 1 can be configured by distributing a computer-readable recording medium such as a flexible disk, CD-ROM (Compact Disc - Read Only Memory), or DVD-ROM (Digital Versatile Disc - Read Only Memory) containing a computer program for performing the aforementioned operations, and then installing the computer program on a computer. Alternatively, the computer program can be stored in a storage device on a server device on a communication network such as the Internet, and the causal explanation support device 1 can be configured by downloading it from a standard computer system.

[0056] Furthermore, if the functions of the causal explanation support device 1 are realized through a division of labor between the OS (Operating System) and the application program, or through collaboration between the OS and the application program, then only the application program portion may be stored on the recording medium or storage device.

[0057] Furthermore, it is possible to superimpose a computer program onto the carrier wave and provide it via a communication network. For example, the computer program may be posted on a bulletin board system (BBS) on the communication network and provided via the communication network. The system may then be configured to execute the aforementioned processing by starting this computer program and running it under the control of the OS, just like other application programs.

[0058] In the embodiments 1 to 4 described above, the causal explanation support device 1 and the demand forecasting device 2 were provided separately, but the invention is not limited to this, and the causal explanation support device 1 and the demand forecasting device 2 may be implemented in a single device.

[0059] In embodiments 1 to 4 described above, the prompt creation unit 12 of the causal explanation support device 1 created a causal explanation prompt based on leading indicator name information, target product name information, time lag information, and example information showing an example of causal explanation. However, it is not limited to this, and for example, a causal explanation prompt may be created based on leading indicator name information, target product name information, and time lag information. In this case, the output format should be specified in the causal explanation prompt so that the causal relationship between the leading indicator for product demand forecasting and the target product is answered as a collection of sub-causes.

[0060] In the above embodiments 1 to 4, the causal explanation support device 1 is equipped with an artificial intelligence unit 13, but the artificial intelligence unit 13 may be, for example, a text generation AI provided on an external server.

[0061] In embodiments 2 to 4 described above, the causal explanation support device 1 is equipped with a related information acquisition unit 15. However, it is not limited to this, and for example, related information relating to events that can serve as leading indicators for predictions and products that can become the target products for prediction may be stored in advance.

[0062] In embodiments 1 to 4 described above, the causal explanation support device 1 outputs causal information showing the causal relationship between leading indicators for product demand forecasting and the product being forecasted. However, it may also output causal information showing the causal relationship between leading indicators for product demand forecasting and the product being forecasted, not limited to products. In this case, the information acquisition unit 11 acquires leading indicator name information indicating the name of the leading indicator in the demand forecast, target product name information indicating the name of the product for which demand forecasting is to be performed, and time lag information indicating the time lag from the occurrence of the leading indicator event in the demand forecasting until the demand for the product occurs. The prompt creation unit 12 creates a causal explanation prompt that instructs the system to respond with a set of sub-causes showing the causal relationship between the leading indicator for product demand forecasting and the product being forecasted, based on the leading indicator name information, target product name information, time lag information, and example information showing an example of causal explanation, and inputs it to the artificial intelligence unit 13.

[0063] Although Embodiments 1 to 4 described above were explained separately, Embodiment 1 and Embodiment 3 may be combined, or Embodiment 1 and Embodiment 4 may be combined.

[0064] Although preferred embodiments have been described in detail above, the invention is not limited to the embodiments described above, and various modifications and substitutions can be made to Embodiment 1 described above without departing from the scope of the claims.

[0065] The various aspects of this disclosure are summarized below as an appendix.

[0066] (Note 1) An information acquisition unit that acquires leading indicator name information indicating the name of a leading indicator in product demand forecasting, target product name information indicating the name of the product for which demand forecasting is to be performed, and time lag information indicating the time lag from the occurrence of an event of a leading indicator in demand forecasting until demand for the product occurs. A prompt creation unit creates a causal explanation prompt that instructs the artificial intelligence unit to respond with a set of sub-causes explaining the intermediate logic between cause and effect, based on the aforementioned leading indicator name information, the aforementioned target product name information, and the aforementioned time lag information, regarding the causal relationship between the leading indicator for product demand forecasting and the target product, and inputs this prompt into the artificial intelligence unit. Based on the response from the artificial intelligence unit, an information output unit outputs causal information showing the causal relationship between the leading indicators for the demand forecast of the product and the product being forecasted. Equipped with, Causal explanation support device. (Note 2) The prompt generation unit, Based on the preceding indicator name information, the target product name information, the time lag information, and the example information showing an example of causal explanation, the causal explanation prompt is created and input to the artificial intelligence unit. The causal explanation support device described in Appendix 1. (Note 3) The artificial intelligence unit further executes the causal explanation prompt and responds with a set of sub-causes indicating the causal relationship between the leading indicators for the demand forecast of the product and the product being forecasted. A causal explanation support device as described in Appendix 1 or 2. (Note 4) The system further includes a fine-tuning unit that fine-tunes the artificial intelligence unit using leading indicators for demand forecasting of the aforementioned product and related information concerning the product being forecasted. A causal explanation support device as described in any of the appendices 1 to 3. (Note 5) The system further includes a fact-checking unit that performs fact-checking on the causal relationship between the leading indicators of the demand forecast for the product provided by the artificial intelligence unit and the product being forecasted. The fact-checking unit creates a sub-causal determination prompt for each sub-causal event, which determines whether the sub-causal event is true or false, and inputs it to the artificial intelligence unit. The information output unit outputs causal information showing the causal relationship between the leading indicators of the demand forecast for the product, for which the artificial intelligence unit has determined all of the sub-causal relationships are correct, and the product being forecasted. A causal explanation support device as described in any of the appendices 1 to 4. (Note 6) The system further includes a fact-checking unit that performs fact-checking on the causal relationship between the leading indicators of the demand forecast for the product provided by the artificial intelligence unit and the product being forecasted. The fact-checking unit performs a statistical analysis on each sub-causal relationship using statistical data corresponding to the cause and statistical data corresponding to the effect of that sub-causal relationship to determine whether it is correct or not. The information output unit outputs causal information showing the causal relationship between the leading indicators of the demand forecast for the product and the product being forecasted, for which the fact-checking unit has determined all of the sub-causal relationships to be correct. A causal explanation support device as described in any of the appendices 1 to 4. (Note 7) A causal explanation support device described in any of Appendix 1 to 6, A demand forecasting device that forecasts product demand, A causal explanation support system equipped with the following features. (Note 8) The causal explanation support device performs the following: The steps include creating a causal explanation prompt that instructs the system to respond with a set of sub-causes explaining the intermediate logic between cause and effect, based on leading indicator name information indicating the name of the leading indicator in the product demand forecast, target product name information indicating the name of the product for which demand forecasting is to be made, and time lag information indicating the time lag between the occurrence of the leading indicator event in the demand forecast and the occurrence of demand for the product; and The steps include inputting the aforementioned causal explanation prompt to the artificial intelligence unit, Based on the response from the artificial intelligence unit, the step of outputting causal information showing the causal relationship between the leading indicators for the demand forecast of the product and the product to be forecasted, Methods for supporting causal explanations. (Note 9) Computers, A prompt creation unit creates a causal explanation prompt that instructs the artificial intelligence unit to respond with a set of sub-causes explaining the intermediate logic between cause and effect, based on leading indicator name information indicating the name of the leading indicator in the product demand forecast, target product name information indicating the name of the product for which demand forecasting is to be made, and time lag information indicating the time lag from the occurrence of the leading indicator event in the demand forecast to the occurrence of demand for the product. This prompt is then input into the artificial intelligence unit. An information output unit outputs causal information showing the causal relationship between the leading indicators for the demand forecast of the product and the product being forecasted, based on the response from the artificial intelligence unit. A program that makes it function as such. [Explanation of Symbols]

[0067] 1 Causal explanation support device, 2 Demand forecasting device, 11 Information acquisition unit, 12 Prompt creation unit, 13 Artificial intelligence unit, 14 Information output unit, 15 Related information acquisition unit, 16 Fine tuning unit, 17 Fact checking unit, 100 Causal explanation support system, 171 Chunk DB, 172 Vector store, 173 Statistical data storage unit, 301 Temporary storage unit, 302 Storage unit, 303 Calculation unit, 304 Input unit, 305 Transmission / reception unit, 306 Display unit.

Claims

1. An information acquisition unit that acquires leading indicator name information indicating the name of a leading indicator in product demand forecasting, target product name information indicating the name of the product for which demand forecasting is to be performed, and time lag information indicating the time lag from the occurrence of an event of a leading indicator in demand forecasting until demand for the product occurs. A prompt creation unit creates a causal explanation prompt that instructs the artificial intelligence unit to respond with a set of sub-causes explaining the intermediate logic between cause and effect, based on the aforementioned leading indicator name information, the aforementioned target product name information, and the aforementioned time lag information, regarding the causal relationship between the leading indicator for product demand forecasting and the target product, and inputs this prompt into the artificial intelligence unit. Based on the response from the artificial intelligence unit, an information output unit outputs causal information showing the causal relationship between the leading indicators for the demand forecast of the product and the product being forecasted. Equipped with, Causal explanation support device.

2. The prompt generation unit, Based on the preceding indicator name information, the target product name information, the time lag information, and the example information showing an example of causal explanation, the causal explanation prompt is created and input to the artificial intelligence unit. The causal explanation support device according to claim 1.

3. The artificial intelligence unit further executes the causal explanation prompt and responds with a set of sub-causes indicating the causal relationship between the leading indicators for the demand forecast of the product and the product being forecasted. The causal explanation support device according to claim 1 or 2.

4. The system further includes a fine-tuning unit that fine-tunes the artificial intelligence unit using leading indicators for demand forecasting of the aforementioned product and related information concerning the product being forecasted. The causal explanation support device according to claim 1 or 2.

5. The system further includes a fact-checking unit that performs fact-checking on the causal relationship between the leading indicators of the demand forecast for the product provided by the artificial intelligence unit and the product being forecasted. The fact-checking unit creates a sub-causal determination prompt for each sub-causal event, which determines whether the sub-causal event is true or false, and inputs it to the artificial intelligence unit. The information output unit outputs causal information showing the causal relationship between the leading indicators of the demand forecast for the product, for which the artificial intelligence unit has determined all of the sub-causal relationships are correct, and the product being forecasted. The causal explanation support device according to claim 1 or 2.

6. The system further includes a fact-checking unit that performs fact-checking on the causal relationship between the leading indicators of the demand forecast for the product provided by the artificial intelligence unit and the product being forecasted. The fact-checking unit performs a statistical analysis on each sub-causal relationship using statistical data corresponding to the cause and statistical data corresponding to the effect of that sub-causal relationship to determine whether it is correct or not. The information output unit outputs causal information showing the causal relationship between the leading indicators of the demand forecast for the product and the product being forecasted, for which the fact-checking unit has determined all of the sub-causal relationships to be correct. The causal explanation support device according to claim 1 or 2.

7. A causal explanation support device according to claim 1 or 2, A demand forecasting device that forecasts product demand, A causal explanation support system equipped with the following features.

8. The causal explanation support device performs the following: The steps include creating a causal explanation prompt that instructs the system to respond with a set of sub-causes explaining the intermediate logic between cause and effect, based on leading indicator name information indicating the name of the leading indicator in the product demand forecast, target product name information indicating the name of the product for which demand forecasting is to be made, and time lag information indicating the time lag between the occurrence of the leading indicator event in the demand forecast and the occurrence of demand for the product; and The steps include inputting the aforementioned causal explanation prompt to the artificial intelligence unit, Based on the response from the artificial intelligence unit, the step of outputting causal information showing the causal relationship between the leading indicators for the demand forecast of the product and the product to be forecasted, Methods for supporting causal explanations.

9. Computers, A prompt creation unit creates a causal explanation prompt that instructs the artificial intelligence unit to respond with a set of sub-causes explaining the intermediate logic between cause and effect, based on leading indicator name information indicating the name of the leading indicator in the product demand forecast, target product name information indicating the name of the product for which demand forecasting is to be made, and time lag information indicating the time lag from the occurrence of the leading indicator event in the demand forecast to the occurrence of demand for the product. This prompt is then input into the artificial intelligence unit. An information output unit outputs causal information showing the causal relationship between the leading indicators for the demand forecast of the product and the product being forecasted, based on the response from the artificial intelligence unit. A program that makes it function as such.

Images