Inventory management system, learning device, inference device, inventory management method and program

Abstract

An inventory management system (100) comprises: a learning device (1) including a first data acquisition unit (11) that acquires inventory placement amount change information indicating a change in the inventory placement amount of an article in each base location resulting from an inventory placement amount change action, PSI plan history information indicating the plan and history of the PSI of an article, and demand amount information indicating the demand amount of an article not listed in the PSI, and a model generation unit (12) that generates a trained model through learning of an inventory placement amount change action by using the inventory placement amount change information, the PSI plan history information, and the demand amount information as learning data; and an inference device (4) including an inference unit (42) that obtains an inventory placement amount change action by inputting the PSI plan history information and the demand amount information acquired by a second data acquisition unit (41) into the trained model, and an output unit (43) that outputs inventory placement amount change action information indicating the inventory placement amount change action obtained by the inference unit (42).

Description

[Technical Field]

[0001] The present disclosure relates to an inventory management system, a learning device, an inference device, an inventory management method, and a program. [Background technology]

[0002] In business activities, in order to control inventory allocation at each location and overall, demand is predicted using past shipping records, and inventory is replenished and transferred between locations to ensure appropriate inventory allocation.

[0003] Patent Document 1 discloses an appropriate inventory management method for pharmaceuticals that can automate the replenishment of pharmaceuticals based on the time-series consumption characteristics of pharmaceuticals consumed at medical institutions or dispensing pharmacies. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2004-284689 Summary of the Invention [Problem to be solved by the invention]

[0005] In the technology described in Patent Document 1, when creating inventory plans, future demand is predicted using only past performance data, which means that changes in demand are followed after the fact, which could result in lost sales opportunities and excess inventory.

[0006] The present disclosure has been made to solve the above-mentioned problems, and aims to enable more appropriate inventory allocation that suppresses lost sales opportunities, excess inventory, etc. [Means for solving the problem]

[0007] To achieve the above object, an inventory management system according to the present disclosure includes a learning device and an inference device. The learning device includes a first data acquisition unit and a model generation unit. The first data acquisition unit acquires inventory allocation amount change information indicating changes in inventory allocation amounts of items at each location due to inventory allocation amount change actions, PSI plan performance information indicating planned and actual Production Sales Inventory (PSI) for items, and demand amount information indicating demand amounts of items not included in the PSI. The model generation unit uses the inventory allocation amount change information, PSI plan performance information, and demand amount information acquired by the first data acquisition unit as learning data to learn inventory allocation amount change actions and generate a trained model. The demand information includes one or more of the requested quantity of the item, the request source, the requested destination, the requested date, the requested delivery date, or the lead time from the requested date to the requested delivery date. The inference device includes a second data acquisition unit, an inference unit, and an output unit. The second data acquisition unit acquires PSI plan performance information and demand quantity information. The inference unit inputs the PSI plan performance information and demand quantity information acquired by the second data acquisition unit into a trained model to obtain an inventory placement quantity change action. The output unit outputs inventory placement quantity change action information indicating the inventory placement quantity change action acquired by the inference unit. [Effects of the Invention]

[0008] According to the present disclosure, by using not only inventory allocation quantity change information and PSI planned performance information, but also demand quantity information indicating the demand quantity of items not included in the PSI as training data, a trained model that learns actions to change the inventory allocation quantity of items at each location can be generated, enabling more appropriate inventory allocation that reduces sales opportunity losses, excess inventory, etc. [Brief explanation of the drawings]

[0009] [Figure 1] A block diagram showing a configuration example of an inventory management system according to a first embodiment. [Figure 2] FIG. 10 is a schematic diagram showing an inventory allocation amount change action according to the first embodiment; [Figure 3] FIG. 10 is a diagram showing an example of inventory allocation amount change information according to the first embodiment. [Figure 4] FIG. 1 is a diagram showing an example of a PSI plan included in PSI plan performance information according to the first embodiment. [Figure 5] FIG. 1 is a diagram showing an example of PSI results included in PSI plan result information according to the first embodiment. [Figure 6] FIG. 1 is a diagram showing an example of demand information according to the first embodiment. [Figure 7] FIG. 1 is a block diagram showing an example of a functional configuration of a learning device according to a first embodiment. [Figure 8] Flowchart showing learning process 1 according to embodiment 1 [Figure 9] 1 is a flowchart showing an inference process according to the first embodiment. [Figure 10] FIG. 10 is a diagram showing an example of actual transportation cost information according to the second embodiment. [Figure 11] FIG. 10 is a diagram showing an example of inventory storage resource occupancy information according to the second embodiment. [Figure 12] Flowchart showing learning process 2 according to embodiment 2 [Figure 13] FIG. 1 is a diagram showing an example of the hardware configuration of a learning device and an inference device according to first and second embodiments. DETAILED DESCRIPTION OF THE INVENTION

[0010] An inventory management system, a learning device, an inference device, an inventory management method, and a program according to the present embodiment will be described in detail below with reference to the drawings. The same or corresponding parts in the drawings are designated by the same reference numerals. In this embodiment, an example of managing product inventory placement amounts at each location of a company with multiple locations will be described.

[0011] (Embodiment 1) The configuration of an inventory management system 100 according to a first embodiment will be described with reference to FIG. 1. The inventory management system 100 includes a production management system 2 that stores inventory allocation amount change information indicating changes in inventory allocation amounts due to changes in product inventory amounts at each base, PSI plan performance information indicating planned and actual product PSI, and demand amount information indicating product demand; a learning device 1 that uses the inventory allocation amount change information, PSI plan performance information, and demand amount information as learning data to learn changes in product inventory amounts at each base and generate a trained model; a trained model storage unit 3 that stores the trained model generated by the learning device 1; and an inference device 4 that uses the trained model to infer and output changes in product inventory amounts at each base to ensure appropriate inventory allocation. Hereinafter, changes in product inventory amounts at each base will be abbreviated as inventory allocation amount change action.

[0012] The inventory allocation amount change action will be explained using Figure 2. The inventory allocation amount change action refers to the supply of inventory from the supplier factory (for example, the product production factory) shown in Figure 2 to destination bases Ba, Bb, and Bc (for example, inventory management bases of a sales company, national inventory management bases managed by the production factory, etc.), as well as the transfer of inventory between destination bases. When inventory is replenished from the factory to the bases, the factory's inventory decreases, and when products are sold from each base to agents, end users, etc., the inventory at each base decreases. Furthermore, when products are produced or procured at the factory, the factory's inventory increases, and when inventory is replenished from the factory to the bases, the inventory at the bases increases.

[0013] The inventory allocation quantity change information will be described with reference to Figure 3. As shown in Figure 3, the inventory allocation quantity change information includes the total inventory quantity of the supplier and the supply destination, and the inventory allocation quantity ratios of the supplier and the supply destination. For example, the total inventory quantity of product Pa at the supplier and the supply destination is 100 units, and the inventory allocation quantity ratios of the supplier and the supply destination, i.e., the factory, base Ba, base Bb, and base Bc, are 50%, 20%, 20%, and 10%, respectively. The learning device 1 may acquire inventory allocation quantity change information each time an inventory allocation quantity change action is performed, or may acquire inventory allocation quantity change information for a certain period all at once.

[0014] The PSI plan performance information will be explained using FIGS. 4 and 5. The PSI plan performance information includes the PSI plan and PSI performance for a product. FIG. 4 shows the monthly product P (production and procurement), S (shipment), and I (inventory) plans included in the PSI plan performance information. P (production and procurement) is an increase in I (inventory), and S (shipment) is a decrease in I (inventory). For example, as of April, the PSI plan has P (production and procurement) of 17 units, S (shipment) of 19 units, and I (inventory) of 20 units. One month ahead, P (production and procurement) is 11 units, and S (shipment) is 14 units, so I (inventory) is 20 + 11 - 14 = 17 units. FIG. 5 shows the monthly product P (production and procurement), S (shipment), and I (inventory) performance included in the PSI plan performance information. For example, as of April, the actual P (production and procurement) figures were 13 units, S (shipment) was 17 units, and I (inventory) was 15 units. In Figures 4 and 5, the units of the product PSI plan and actual figures were number of units (units), but this is not limiting and may be monetary value (yen), for example.

[0015] Demand information will be explained using Figure 6. Demand information is information that indicates the demand for products that have not yet been recorded in the PSI, and corresponds to requests from supply destinations to suppliers for inventory replenishment, orders from agents, end users (customers) to supply destinations, etc.

[0016] In the example of Figure 6, the demand information includes the requested quantity of each product, the request source, the request destination, the requested date, the requested delivery date, and the lead time from the requested date to the requested delivery date. For example, on January 2, 2022, customer Ca requests five units of product Pa from base Ba by the delivery date of January 10, 2022. The lead time from the requested date to the requested delivery date is eight days.

[0017] 1, the learning device 1 includes a data acquisition unit 11 that acquires inventory allocation amount change information, PSI plan performance information, and demand amount information from the production management system 2, and a model generation unit 12 that learns inventory allocation amount change actions using the inventory allocation amount change information, PSI plan performance information, and demand amount information as learning data, and generates a learned model. The data acquisition unit 11 is an example of a first data acquisition unit.

[0018] The model generation unit 12 learns, for example, an inventory allocation change action that improves inventory turnover or minimizes a deterioration in inventory turnover. In this case, when PSI plan performance information and demand information are input to the trained model generated by the model generation unit 12, an inventory allocation change action that improves inventory turnover or minimizes a deterioration in inventory turnover is output. By executing the inventory allocation change action output from the trained model to increase or decrease the inventory allocation amount overall and at each location, it is possible to reduce sales opportunity losses, excess inventory, and other issues. This trained model is effective, for example, when a customer requests a short lead time and the supplier (factory) cannot meet the demand. Specifically, if there is a factory in Japan and various locations in Europe, even if the supply from the Japanese factory cannot meet the demand, it may be possible to respond by transferring inventory between the European locations.

[0019] Furthermore, if there is a bias (excess inventory, shortage of inventory, etc.) when comparing the inventory allocation amount at each location relative to the demand amount for that location between locations, even if there is sufficient inventory overall to meet the demands of all customers and the market, this will result in uneven distribution of inventory between locations, leading to lost sales opportunities, excess inventory, etc. In contrast, by using the trained model described above, it is possible to eliminate uneven distribution of inventory between locations and prevent lost sales opportunities, excess inventory, etc.

[0020] The learning algorithm used by the model generation unit 12 can be any known algorithm such as supervised learning, unsupervised learning, or reinforcement learning. As an example, a case where reinforcement learning is applied will be described. In reinforcement learning, an agent (subject of action) in a certain environment observes the current state (environmental parameters) and decides on an action to take. The environment changes dynamically depending on the agent's actions, and the agent is given a reward according to the environmental changes. The agent repeats this process and learns the course of action that will obtain the most reward through a series of actions. Typical reinforcement learning methods include Q-learning and TD-learning. For example, in the case of Q-learning, a general update formula for the action value function Q(s, a) is expressed by Equation 1.

[0021]

number

[0022] In equation 1, s t represents the state of the environment at time t, and a t represents the action at time t. Action a t Therefore, the state is s t+1 Changes to r t+1 represents the reward obtained depending on the change in state, γ represents the discount rate, and α represents the learning coefficient. Note that γ is in the range of 0<γ≦1, and α is in the range of 0<α≦1. When the model generation unit 12 uses reinforcement learning, the action a t is an inventory placement change action, and the planned and actual PSI and the product demand are in state st The model generation unit 12 generates the state s t Best Practices in a t Learn.

[0023] The update formula expressed by Equation 1 increases the action value Q if the action value Q of the action a with the highest Q value at time t+1 is greater than the action value Q of the action a executed at time t, and decreases the action value Q if the opposite is true. In other words, it updates the action value Q of action a at time t to an action value function Q(s, a) that brings it closer to the best action value at time t+1. In this way, the best action value in a certain environment is propagated sequentially to the action value in the previous environment.

[0024] When generating a trained model by reinforcement learning, the model generation unit 12 includes a reward calculation unit 121 and a function update unit 122, as shown in FIG.

[0025] The remuneration calculation unit 121 calculates the remuneration based on the inventory allocation amount change action indicated by the inventory allocation amount information, the PSI plan and PSI results indicated by the PSI plan result information, and the product demand amount indicated by the demand amount information. The remuneration calculation unit 121 calculates the remuneration r based on changes in inventory turnover. For example, if the inventory turnover improves, the remuneration r is increased (e.g., a remuneration of "1"); if the inventory turnover worsens, the remuneration r is decreased (e.g., a remuneration of "-1"). For example, since the annual inventory turnover is calculated by "annual shipment volume / inventory volume," there are four ways to increase the remuneration: "increase shipment volume and maintain or reduce inventory volume," "maintain shipment volume and reduce inventory volume," "increase shipment volume and increase inventory volume by a smaller amount," and "reduce shipment volume and reduce inventory volume by a larger amount."

[0026] The function update unit 122 updates the function for determining the inventory placement amount change action that maximizes the reward for the input state, according to the reward calculated by the reward calculation unit 121. For example, in the case of Q-learning, the action value function Q(s t ,a t) is used as a function to calculate the inventory placement amount change action that maximizes the reward for the input state.

[0027] The model generation unit 12 repeatedly executes the above learning until a learning end condition is met. The learning end condition may be, for example, when the state s t When the learning is completed, the learned model storage unit 3 stores the action-value function Q(s) updated by the function update unit 122. t ,a t ), i.e., stores the trained model.

[0028] In the above description, the learning device 1 stores the generated trained model in an externally provided trained model storage unit 3, but the trained model storage unit 3 may be provided inside the learning device 1. Alternatively, the trained model storage unit 3 may be configured in a system or device external to the inventory management system 100.

[0029] 1, the inference device 4 includes a data acquisition unit 41 that acquires PSI plan performance information and demand quantity information from the production management system 2, an inference unit 42 that inputs the PSI plan performance information and demand quantity information into a trained model stored in the trained model storage unit 3 and infers an inventory placement quantity change action, and an output unit 43 that outputs inventory placement quantity change action information indicating the inventory placement quantity change action inferred by the inference unit 42. The data acquisition unit 41 is an example of a second data acquisition unit.

[0030] The inference unit 42 inputs the PSI plan performance information and demand quantity information acquired by the data acquisition unit 41 into the trained model stored in the trained model storage unit 3 to obtain an inventory placement quantity change action. The output unit 43 outputs inventory placement quantity change action information indicating the inventory placement quantity change action obtained by the inference unit 42 from the trained model. The inventory placement quantity change action information may be output by, for example, a screen display, an audio output, or transmission to a terminal used by a user.

[0031] In the above description, the inference device 4 uses a trained model stored in an externally provided trained model storage unit 3, but the trained model storage unit 3 may also be provided inside the inference device 4.

[0032] As described above, the processing performed by the inventory management system 100 includes a learning phase in which inventory allocation quantity change actions are learned, and an application phase in which inventory allocation quantity change actions are inferred using the learned model.

[0033] Here, the flow of learning process 1 executed by the learning device 1 in the learning phase will be explained using Figure 8. Learning process 1 shown in Figure 8 is an example of learning process when reinforcement learning is applied as the learning algorithm. Learning process 1 starts when the learning device 1 is powered on. If the data acquisition unit 11 of the learning device 1 has not acquired inventory allocation amount change information, PSI plan performance information, and demand amount information required for learning from the production management system 2 (step S11; NO), the process proceeds to step S18.

[0034] When the data acquisition unit 11 acquires inventory allocation quantity change information, PSI plan performance information, and demand quantity information required for learning from the production management system 2 (step S11; YES), the reward calculation unit 121 of the model generation unit 12 determines whether the inventory allocation quantity change action has improved the inventory turnover rate based on the inventory allocation quantity change action indicated by the inventory allocation quantity information, the PSI plan and PSI performance indicated by the PSI plan performance information, and the product demand quantity indicated by the demand quantity information (step S12). Step S11 is an example of a first data acquisition step.

[0035] If the inventory turnover rate improves as a result of the inventory placement quantity change action (step S12; YES), the remuneration calculation unit 121 increases the remuneration r (step S13). In step S13, the remuneration calculation unit 121 gives a remuneration of, for example, "1". If the inventory turnover rate worsens as a result of the inventory placement quantity change action (step S12; NO), the remuneration calculation unit 121 reduces the remuneration r (step S14). In step S14, the remuneration calculation unit 121 gives a remuneration of, for example, "-1".

[0036] The function update unit 122 of the model generation unit 12 updates the function for determining the inventory placement amount change action for the input state according to the reward calculated by the reward calculation unit 121 (step S15). For example, in the case of Q-learning, the action value function Q(s t ,a t ) is used as a function to calculate the inventory placement amount change action that maximizes the reward for the input state.

[0037] The model generation unit 12 determines whether or not the learning termination condition is satisfied (step S16). If the learning termination condition is not satisfied (step S16; NO), the process returns to step S12, and steps S12 to S16 are repeated. If the learning termination condition is satisfied (step S16; YES), the learned model storage unit 3 stores the action-value function Q(s t ,a t ), that is, the learned model is stored (step S17). The learning end condition is, for example, t reaches a terminal state, the number of repetitions exceeds a predetermined number, etc. Steps S12 to S16 are an example of a model generation step.

[0038] If the power of the learning device 1 is not turned off (step S18; NO), the process returns to step S11 and steps S11 to S18 are repeated. If the power of the learning device 1 is turned off (step S18; YES), the process ends.

[0039] 8, learning of inventory placement quantity change actions begins when the data acquisition unit 11 acquires inventory placement quantity change information, PSI plan performance information, and demand quantity information required for learning from the production management system 2, but this is not limiting. For example, a configuration may be adopted in which learning of inventory placement quantity change actions begins at a timing specified by the user to generate a trained model.

[0040] Next, the flow of the inference process executed by the inference device 4 in the utilization phase will be described with reference to FIG. 9. The inference process shown in FIG. 9 starts when the inference device 4 is powered on. If the data acquisition unit 41 of the inference device 4 does not acquire PSI plan performance information and demand quantity information from the production control system 2 (step S21; NO), the process proceeds to step S25. If the data acquisition unit 41 acquires PSI plan performance information and demand quantity information from the production control system 2 (step S21; YES), the inference unit 42 inputs the PSI plan performance information and demand quantity information into the trained model stored in the trained model storage unit 3 (step S22) and obtains an inventory placement quantity change action (step S23). Step S21 is an example of a second data acquisition step. Steps S22 and S23 are examples of inference steps. The output unit 43 outputs inventory placement quantity change action information indicating the inventory placement quantity change action obtained by the inference unit 42 (step S24). The inventory allocation amount change action information may be output, for example, by a screen display, an audio output, or by transmission to a terminal used by the user.

[0041] If the power supply of the inference device 4 is not turned off (step S25; NO), the process returns to step S21 and steps S21 to S25 are repeated. If the power supply of the inference device 4 is turned off (step S25; YES), the process ends.

[0042] According to the inventory management system 100 of embodiment 1, not only inventory allocation quantity change information and PSI planned performance information, but also demand quantity information indicating the demand quantity of products not included in the PSI is used as learning data to generate a trained model that has learned the inventory allocation quantity change actions for products at each location, thereby enabling more appropriate inventory allocation that reduces sales opportunity losses, excess inventory, etc.

[0043] (Embodiment 2) The configuration of the inventory management system 100 and the functional configuration of the inference device 4 in the second embodiment are the same as those in the first embodiment. In the second embodiment, the learning device 1 uses, as learning data, inventory placement quantity change information, PSI plan actual result information, and demand quantity information, as well as actual transportation cost information indicating the actual transportation costs resulting from inventory placement quantity change actions, and inventory storage resource occupancy information indicating the storage resource occupancy rate of the product inventory at each base, to learn the inventory placement quantity change actions for the products at each base and generate a trained model.

[0044] The actual transportation cost information will be explained using Fig. 10. As shown in Fig. 10, the actual transportation cost information includes the transportation date, transportation origin, transportation destination, and transportation cost of each product. For example, when product Pa is transported from base Ba to customer Ca on January 2, 2022, the transportation cost is 10,000 yen.

[0045] The inventory storage resource occupancy information will be explained using FIG. 11. As shown in FIG. 11, the inventory storage resource occupancy information includes a date and a storage resource occupancy, which is the rate at which the product inventory at each base occupies the storage resource on that date. For example, the storage resource occupancy at base Ba on January 2, 2022 is 90%. The storage resource occupancy is calculated periodically, for example.

[0046] The data acquisition unit 11 of the learning device 1 acquires inventory allocation quantity change information, PSI plan actual information and demand quantity information, transportation cost actual information, and inventory storage resource occupancy rate information from the production management system 2. The model generation unit 12 learns inventory allocation quantity change actions using the inventory allocation quantity change information, PSI plan actual information, demand quantity information, transportation cost actual information, and inventory storage resource occupancy rate information as learning data, and generates a trained model. The model generation unit 12 stores the generated trained model in the trained model storage unit 3.

[0047] The model generation unit 12 learns an inventory allocation quantity change action that improves inventory turnover or prevents a deterioration in inventory turnover, reduces transportation costs, and ensures that the storage resource occupancy rate at each location meets an acceptable value (e.g., 90%). In this case, when PSI plan performance information, demand quantity information, transportation cost performance information, and inventory storage resource occupancy rate information are input to the trained model generated by the model generation unit 12, an inventory allocation quantity change action that improves inventory turnover or prevents a deterioration in inventory turnover, reduces transportation costs, and ensures that the storage resource occupancy rate at each location meets an acceptable value is output. By executing the inventory allocation quantity change action output from the trained model to increase or decrease the overall inventory allocation amount and at each location, the storage resource occupancy rate at each location does not exceed an acceptable value, thereby preventing sales opportunity losses, excess inventory, and the like, and reducing transportation costs. This trained model is effective, for example, when a customer requests a short lead time and the supplier (factory) cannot meet the required delivery time. Specifically, if a company has a factory in Japan and various bases in Europe, even if the supply from the Japanese factory cannot keep up with demand, it may be possible to handle the situation by transferring inventory between the European bases.

[0048] Furthermore, if there is a bias (excess inventory, shortage of inventory, etc.) when comparing the inventory allocation amount at each location relative to the demand amount for that location between locations, even if there is sufficient inventory overall to meet all customer and market demand, inventory unevenness will occur between locations, leading to lost sales opportunities, excess inventory, etc. In contrast, by using the trained model described above, it is possible to eliminate inventory unevenness between locations and prevent lost sales opportunities, excess inventory, etc., while also reducing transportation costs and preventing storage resource occupancy rates from exceeding allowable values.

[0049] The learning algorithm used by the model generation unit 12 may be a known algorithm such as supervised learning, unsupervised learning, or reinforcement learning, as in the first embodiment. In these learning algorithms, a constraint is that the storage resource occupancy rate at each base does not exceed an allowable value.

[0050] For example, when the model generation unit 12 uses Q-learning, the action a t is an inventory placement change action, and the planned and actual PSI, the product demand, and the transportation cost are in the state s t The constraint is that the storage resource occupancy rate does not exceed the allowable value. The model generation unit 12 generates a state s at time t that satisfies the constraint. t Best Practices in a t Learn.

[0051] In the second embodiment, the reward r is calculated from two rewards r1 and r2. For example, if the inventory turnover rate improves as a result of the inventory placement quantity change action, the reward r1 is increased. If the inventory turnover rate worsens as a result of the inventory placement quantity change action, the reward r1 is decreased. If the total transportation costs due to the inventory placement quantity change action are equal to or less than a threshold, the reward r2 is increased. If the total transportation costs due to the inventory placement quantity change action are higher than a threshold, the reward r2 is decreased. The threshold for determining whether to increase the reward r2 may be different from the threshold for determining whether to decrease the reward r2. The reward r = r1 + r2. Alternatively, the reward r1 and the reward r2 may be weighted. In this case, the reward r = w1r1 + w2r2. The other functions of the learning device 1 are the same as those of the first embodiment.

[0052] Here, the flow of learning process 2 executed by the learning device 1 in the learning phase will be described with reference to Fig. 12. Learning process 2 shown in Fig. 12 is an example of learning process when reinforcement learning is applied as the learning algorithm. Learning process 1 starts when the learning device 1 is powered on. If the data acquisition unit 11 of the learning device 1 has not acquired the inventory allocation amount change information, PSI plan performance information, demand amount information, transportation cost performance information, and inventory storage resource occupancy rate information required for learning from the production management system 2 (step S31; NO), the process proceeds to step S43.

[0053] When the data acquisition unit 11 acquires the inventory allocation quantity change information, PSI plan performance information, demand quantity information, transportation cost performance information, and inventory storage resource occupancy rate information required for learning from the production management system 2 (step S31; YES), the remuneration calculation unit 121 of the model generation unit 12 determines whether the inventory allocation quantity change action satisfies the constraint (i.e., the storage resource occupancy rate does not exceed the allowable value) based on the storage resource occupancy rate indicated by the inventory storage resource occupancy rate information (step S32). If the inventory allocation quantity change action does not satisfy the constraint (step S32; NO), the process proceeds to step S41. If the inventory allocation quantity change action satisfies the constraint (step S32; YES), the remuneration calculation unit 121 of the model generation unit 12 determines whether the inventory turnover rate has improved based on the inventory allocation quantity change action indicated by the inventory allocation quantity information, the PSI plan and PSI performance indicated by the PSI plan performance information, and the product demand amount indicated by the demand quantity information (step S33).

[0054] If the inventory turnover rate improves as a result of the inventory placement amount change action (step S33; YES), the remuneration calculation unit 121 increases the remuneration r1 (step S34).If the inventory turnover rate worsens as a result of the inventory placement amount change action (step S33; NO), the remuneration calculation unit 121 decreases the remuneration r1 (step S35).

[0055] The remuneration calculation unit 121 of the model generation unit 12 determines whether the total transportation costs due to the inventory allocation quantity change action is below a threshold value based on the inventory allocation quantity change action indicated by the inventory allocation quantity information, the PSI plan and PSI results indicated by the PSI plan result information, the product demand quantity indicated by the demand quantity information, and the transportation cost result information (step S36).

[0056] If the total transportation costs due to the inventory placement quantity change action is equal to or less than the threshold (step S36; YES), the remuneration calculation unit 121 increases the remuneration r2 (step S37). If the total transportation costs due to the inventory placement quantity change action is greater than the threshold (step S36; NO), the remuneration calculation unit 121 decreases the remuneration r2 (step S38). The remuneration calculation unit 121 calculates the remuneration r based on the remuneration r1 and the remuneration r2 (step S39). The remuneration r = r1 + r2. When weighting the remuneration r1 and the remuneration r2, the remuneration r = w1r1 + w2r2 is calculated.

[0057] The function update unit 122 of the model generation unit 12 updates the function for determining the inventory placement amount change action for the input state according to the reward calculated by the reward calculation unit 121 (step S40). For example, in the case of Q-learning, the action value function Q(s t ,a t ) is used as a function to calculate the inventory placement amount change action that maximizes the reward for the input state.

[0058] The model generation unit 12 determines whether or not the learning termination condition is satisfied (step S41). If the learning termination condition is not satisfied (step S41; NO), the process returns to step S32, and steps S32 to S41 are repeated. If the learning termination condition is satisfied (step S41; YES), the learned model storage unit 3 stores the action-value function Q(s t ,a t ), that is, the learned model is stored (step S42). The learning end condition is, for example, t reaches a terminal state, the number of repetitions exceeds a predetermined number, etc. Steps S32 to S36 are an example of a model generation step.

[0059] If the power of the learning device 1 is not turned off (step S43; NO), the process returns to step S31 and steps S31 to S43 are repeated. If the power of the learning device 1 is turned off (step S43; YES), the process ends.

[0060] 12, learning of inventory allocation amount change actions begins when the data acquisition unit 11 acquires inventory allocation amount change information, PSI plan performance information, demand amount information, transportation cost performance information, and inventory storage resource occupancy rate information required for learning from the production management system 2. However, this is not limiting. For example, learning of inventory allocation amount change actions may be started at a timing specified by the user to generate a trained model.

[0061] According to the inventory management system 100 of embodiment 2, not only inventory allocation quantity change information and PSI planned performance information, but also demand quantity information indicating the demand quantity of products not included in the PSI is used as learning data to generate a trained model that has learned the inventory allocation quantity change actions for products at each base. This makes it possible to eliminate inventory uneven distribution between bases, reduce sales opportunity losses, excess inventory, etc., reduce transportation costs, prevent storage resource occupancy rates from exceeding allowable values, and enable more appropriate inventory allocation.

[0062] In the first and second embodiments described above, the data acquisition unit 11 of the learning device 1 and the data acquisition unit 41 of the inference device 4 both acquire PSI plan / actual information including PSI plans and actual results, but this is not limited thereto. For example, the data acquisition unit 11 of the learning device 1 may acquire PSI actual information indicating PSI actual results, and the data acquisition unit 41 of the inference device 4 may acquire PSI plan information indicating PSI plans. In this case, the model generation unit 12 of the learning device 1 uses the PSI actual information and demand quantity information as learning data to learn an inventory placement quantity change action and generate a trained model, and the inference unit 42 of the inference device 4 inputs the PSI plan information and demand quantity information into the trained model stored in the trained model storage unit 3 to infer an inventory placement quantity change action.

[0063] In the above-described first and second embodiments, the inventory management system 100 includes the learning device 1 and the inference device 4, and the inference device 4 infers an inventory placement quantity change action using a trained model generated by the learning device 1. However, this is not limited to this. The inference device 4 may also acquire a trained model from another system or device and infer an inventory placement quantity change action using the acquired trained model.

[0064] In the above-described first and second embodiments, the case where reinforcement learning is applied to the learning algorithm used in the model generation unit 12 has been described, but the present invention is not limited to this. As for the learning algorithm, supervised learning, unsupervised learning, semi-supervised learning, or the like can also be applied in addition to reinforcement learning. Furthermore, as the learning algorithm used in the model generation unit 12, deep learning that learns to extract feature quantities themselves can also be used, and the model generation unit 12 may also perform machine learning according to other known methods, such as neural networks and functional logic programming.

[0065] In the above-described first and second embodiments, the learning device 1 and the inference device 4 are included in the inventory management system 100, but the learning device 1 and the inference device 4 may be separate devices or may exist on a cloud server.

[0066] In the first embodiment described above, the data acquisition unit 11 of the learning device 1 acquires inventory allocation quantity change information, PSI plan performance information, and demand quantity information, and the data acquisition unit 41 of the inference device 4 acquires PSI plan performance information and demand quantity information from the production management system 2, but this is not limited to this. For example, the information may be acquired from another system or device, or may be directly input by the user. Similarly, in the second embodiment, the data acquisition unit 11 of the learning device 1 may acquire inventory allocation quantity change information, PSI plan performance information and demand quantity information, transportation cost performance information, and inventory storage resource occupancy information from another system or device, or may be directly input by the user.

[0067] In the first and second embodiments described above, the model generation unit 12 learns inventory placement quantity change actions using learning data acquired from the production control system 2. However, learning data may be acquired from multiple related production control systems 2 used in the same area, or from multiple production control systems 2 operating independently in different areas. A configuration may also be adopted in which a production control system 2 that collects learning data can be added to or removed from the target system at any time. Furthermore, a learning device 1 that has learned inventory placement quantity change actions for one inventory control system 100 may be applied to another inventory control system, and the learning device 1 may be configured to relearn and update inventory placement quantity change actions for the other inventory control system.

[0068] In the above first and second embodiments, an example of managing product inventory allocation at each location of a company with multiple locations has been described, but this is not limited to products; it can also be any item whose inventory is managed based on a PSI plan.

[0069] In the second embodiment described above, the reward r is calculated from two rewards r1 and r2. For example, if the inventory turnover rate improves, reward r1 is increased. If the inventory turnover rate worsens, reward r1 is decreased. If the total transportation cost is equal to or less than a threshold, reward r2 is increased. If the total transportation cost is higher than a threshold, reward r2 is decreased. Alternatively, the data acquisition unit 11 of the learning device 1 may acquire management target information indicating management target values ​​for inventory turnover rate and transportation cost in addition to inventory allocation amount change information, PSI plan actual information and demand amount information, transportation cost actual information, and inventory storage resource occupancy information, and determine the reward based on whether the management target achievement rate improves.

[0070] In this case, for example, if the business target achievement rate for inventory turnover improves, reward r1 is increased. If the business target achievement rate for inventory turnover deteriorates, reward r1 is decreased. If the business target achievement rate for transportation costs improves, reward r2 is increased. If the business target achievement rate for transportation costs deteriorates, reward r2 is decreased. As in embodiment 2, reward r = r1 + r2. If reward r1 and reward r2 are weighted, reward r = w1r1 + w2r2 is calculated. Alternatively, reward r may be increased if the combined changes in the business target achievement rates for inventory turnover and transportation costs improve, and may be decreased if the combined changes in the business target achievement rates for inventory turnover and transportation costs deteriorate. In this variant, when inventory allocation quantity change information, PSI plan actual information and demand quantity information, transportation cost actual information, inventory storage resource occupancy information, and management target information are input into the trained model generated by the model generation unit 12, an inventory allocation quantity change action is output that improves the management target achievement rate for inventory turnover and transportation costs, or prevents a deterioration in the management target achievement rate for inventory turnover and transportation costs.

[0071] The hardware configuration of the learning device 1 and the inference device 4 will be described with reference to Fig. 13. As shown in Fig. 13, the learning device 1 and the inference device 4 include a temporary storage unit 101, a storage unit 102, a calculation unit 103, an input unit 104, a transmission / reception unit 105, and a display unit 106. The temporary storage unit 101, the storage unit 102, the input unit 104, the transmission / reception unit 105, and the display unit 106 are all connected to the calculation unit 103 via a BUS.

[0072] The calculation unit 103 is, for example, a CPU (Central Processing Unit). The calculation unit 103 executes the processes of the model generation unit 12 of the learning device 1 and the inference unit 42 of the inference device 4 in accordance with a control program stored in the storage unit 102.

[0073] The temporary storage unit 101 is, for example, a RAM (Random-Access Memory). The temporary storage unit 101 loads the control program stored in the storage unit 102 and is used as a work area for the calculation unit 103.

[0074] The storage unit 102 is a non-volatile memory such as a flash memory, a hard disk, a DVD-RAM (Digital Versatile Disc - Random Access Memory), or a DVD-RW (Digital Versatile Disc - Rewritable). The storage unit 102 pre-stores a program for causing the calculation unit 103 to perform the processing of the learning device 1 and the inference device 4, and also supplies the data stored by this program to the calculation unit 103 in accordance with instructions from the calculation unit 103, and stores the data supplied from the calculation unit 103. In a configuration in which the learning device 1 and the inference device 4 are internally provided with the trained model storage unit 3, the trained model storage unit 3 is configured in the storage unit 102.

[0075] The input unit 104 is an interface device that connects input devices such as a keyboard, pointing device, and voice input device to the BUS. Information input by the user is supplied to the calculation unit 103 via the input unit 104. In a configuration in which the user directly inputs information to the data acquisition unit 11 of the learning device 1, the input unit 104 functions as the data acquisition unit 11. In a configuration in which the user directly inputs information to the data acquisition unit 41 of the inference device 4, the input unit 104 functions as the data acquisition unit 41.

[0076] The transmitting / receiving unit 105 is a network termination device or a wireless communication device that connects to the network, and a serial interface or a LAN (Local Area Network) interface that connects to them. The transmitting / receiving unit 105 functions as the model generation unit 12 of the learning device 1 and the inference unit 42 of the inference device 4. In a configuration in which the output unit 43 of the inference device 4 transmits inventory placement quantity change action information to a terminal used by a user, the transmitting / receiving unit 105 functions as the output unit 43.

[0077] The display unit 106 is a display device such as an LCD (Liquid Crystal Display), an organic EL (Electroluminescence) display, etc. In a configuration in which the output unit 43 of the inference device 4 displays the inventory placement quantity change action information on a screen, the display unit 106 functions as the output unit 43.

[0078] The processing of the data acquisition unit 11 and model generation unit 12 of the learning device 1, and the data acquisition unit 41, inference unit 42, and output unit 43 of the inference device 4 shown in Figure 1 is executed by a control program using resources such as the temporary storage unit 101, calculation unit 103, storage unit 102, input unit 104, transmission / reception unit 105, and display unit 106.

[0079] Furthermore, the above hardware configuration and flowchart are merely examples and can be changed and modified as desired.

[0080] The core processing components of learning device 1 and inference device 4, such as calculation unit 103, temporary storage unit 101, storage unit 102, input unit 104, transmission / reception unit 105, and display unit 106, can be realized using an ordinary computer system rather than a dedicated system. For example, a computer program for executing the above operations may be stored and distributed on 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), and the learning device 1 and inference device 4 that execute the above processing may be configured by installing the computer program on a computer. Alternatively, the computer program may be stored in a storage device of a server device on a communication network, such as the Internet, and downloaded by an ordinary computer system to configure learning device 1 and inference device 4.

[0081] In addition, when the functions of the learning device 1 and the inference device 4 are realized by sharing the functions between an OS (Operating System) and an application program, or by collaboration between the OS and an application program, only the application program portion may be stored on a recording medium or storage device.

[0082] It is also possible to superimpose a computer program on a 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 computer program may then be started and executed under the control of an OS in the same way as other application programs, thereby enabling the above-mentioned processing to be performed.

[0083] Although the preferred embodiments have been described in detail above, the present invention is not limited to the above-described embodiments, and various modifications and substitutions can be made to the above-described embodiments without departing from the scope of the claims.

[0084] Various aspects of the present disclosure are summarized below as appendices.

[0085] (Appendix 1) An inventory management system including a learning device and an inference device, The learning device a first data acquisition unit that acquires inventory allocation amount change information indicating a change in inventory allocation amount of an item at each base due to an inventory allocation amount change action, PSI plan and actual result information indicating a PSI (Production Sales Inventory) plan and actual result of the item, and demand amount information indicating a demand amount of the item not included in the PSI; a model generation unit that learns the inventory placement quantity change action and generates a trained model using the inventory placement quantity change information, the PSI plan performance information, and the demand quantity information acquired by the first data acquisition unit as training data; and and The inference device a second data acquisition unit that acquires the PSI plan performance information and the demand amount information; an inference unit that inputs the PSI plan performance information and the demand quantity information acquired by the second data acquisition unit into the trained model to obtain the inventory placement quantity change action; an output unit that outputs inventory quantity change action information indicating the inventory quantity change action obtained by the inference unit; An inventory control system having: (Appendix 2) the model generation unit uses the inventory placement quantity change information, the PSI plan performance information, and the demand quantity information acquired by the first data acquisition unit as learning data to learn the inventory placement quantity change action that improves inventory turnover or minimizes deterioration of inventory turnover; 1. An inventory control system as described in Appendix 1. (Appendix 3) the first data acquisition unit further acquires actual transportation cost information indicating actual transportation costs due to the inventory placement amount change action, and inventory storage resource occupancy information indicating a storage resource occupancy rate of the product inventory at each base; the model generation unit uses the inventory placement quantity change information, the PSI plan actual result information, the demand quantity information, the transportation cost actual result information, and the inventory storage resource occupancy rate information acquired by the first data acquisition unit as learning data to learn the inventory placement quantity change action that improves inventory turnover or prevents a deterioration in inventory turnover, reduces transportation costs, and satisfies an allowable value for the storage resource occupancy rate at each base; 1. An inventory control system as described in Appendix 1. (Appendix 4) the first data acquisition unit further acquires management target information indicating management target values ​​for inventory turnover and transportation costs; the model generation unit uses the inventory placement quantity change information, the PSI plan actual result information, the demand quantity information, the transportation cost actual result information, the inventory storage resource occupancy rate information, and the management target information acquired by the first data acquisition unit as learning data to learn the inventory placement quantity change action that improves the management target achievement rate of inventory turnover and transportation cost or suppresses a deterioration in the management target achievement rate of inventory turnover and transportation cost, and in which the storage resource occupancy rate at each base satisfies an allowable value; The inventory control system described in Appendix 3. (Appendix 5) a first data acquisition unit that acquires inventory allocation amount change information indicating a change in the inventory allocation amount of the item at each base due to the inventory allocation amount change action, PSI plan and actual result information indicating the planned and actual PSI of the item, and demand amount information indicating the demand amount of the item that is not included in the PSI; a model generation unit that learns the inventory placement quantity change action and generates a trained model using the inventory placement quantity change information, the PSI plan performance information, and the demand quantity information acquired by the first data acquisition unit as training data; and A learning device comprising: (Appendix 6) an inference device that infers an inventory placement quantity change action using a trained model generated by training the inventory placement quantity change action using, as training data, inventory placement quantity change information indicating a change in inventory placement quantity of an item at each base due to an inventory placement quantity change action, PSI plan and performance information indicating a plan and performance of PSI of the item, and demand quantity information indicating a demand quantity of the item not included in the PSI, a second data acquisition unit that acquires the PSI plan performance information and the demand amount information; an inference unit that inputs the PSI plan performance information and the demand quantity information acquired by the second data acquisition unit into the trained model to obtain the inventory placement quantity change action; an output unit that outputs inventory quantity change action information indicating the inventory quantity change action obtained by the inference unit; An inference device comprising: (Appendix 7) The learning device executes a first data acquisition step of acquiring inventory allocation amount change information indicating a change in the inventory allocation amount of the item at each base due to an inventory allocation amount change action, PSI plan and actual result information indicating the planned and actual PSI of the item, and demand amount information indicating the demand amount of the item not included in the PSI; a model generation step of learning the inventory placement amount change action and generating a trained model using the inventory placement amount change information, the PSI plan performance information, and the demand amount information acquired in the first data acquisition step as training data; The inference device executes a second data acquisition step of acquiring the PSI plan performance information and the demand amount information; an inference step of inputting the PSI plan result information and the demand quantity information acquired in the second data acquisition step into the trained model to obtain the inventory placement quantity change action; An inventory control method having the following. (Appendix 8) Computer, a first data acquisition unit that acquires inventory allocation amount change information indicating a change in the inventory allocation amount of the item at each base due to an inventory allocation amount change action, PSI plan and actual result information indicating the planned and actual PSI of the item, and demand amount information indicating the demand amount of the item that is not included in the PSI; a model generation unit that uses the inventory placement amount change information, the PSI plan performance information, and the demand amount information acquired by the first data acquisition unit as learning data to learn an inventory placement amount change action for the item at each base and generates a trained model; A program that functions as a (Appendix 9) a computer that uses inventory placement quantity change information indicating a change in the inventory placement quantity of an item at each base due to an inventory placement quantity change action, PSI plan and performance information indicating the plan and performance of the PSI of the item, and demand quantity information indicating the demand quantity of the item not included in the PSI as learning data, and uses a trained model generated by learning the inventory placement quantity change action; a second data acquisition unit that acquires the PSI plan performance information and the demand amount information; an inference unit that inputs the PSI plan performance information and the demand quantity information acquired by the second data acquisition unit into the trained model and obtains the inventory placement quantity change action; A program that functions as a

[0086] It should be noted that the present disclosure allows various embodiments and modifications without departing from the broad spirit and scope of the present disclosure. Furthermore, the above-described embodiments are intended to explain the disclosure and do not limit the scope of the present disclosure. That is, the scope of the present disclosure is defined by the claims, not the embodiments. Various modifications made within the scope of the claims and within the meaning of the disclosure equivalent thereto are considered to be within the scope of the present disclosure.

[0087] This application is based on Japanese Patent Application No. 2022-97333, filed on June 16, 2022. The entire specification, claims, and drawings of Japanese Patent Application No. 2022-97333 are incorporated herein by reference. [Explanation of symbols]

[0088] 1 Learning device, 2 Production management system, 3 Learned model memory unit, 4 Inference device, 11 Data acquisition unit, 12 Model generation unit, 41 Data acquisition unit, 42 Inference unit, 43 Output unit, 100 Inventory management system, 101 Temporary memory unit, 102 Memory unit, 103 Calculation unit, 104 Input unit, 105 Transmission / reception unit, 106 Display unit, 121 Reward calculation unit, 122 Function update unit.

Claims

1. An inventory management system including a learning device and an inference device, The learning device a first data acquisition unit that acquires inventory allocation amount change information indicating a change in the inventory allocation amount of the item at each base due to an inventory allocation amount change action, PSI plan and actual result information indicating a PSI (Production Sales Inventory) plan and actual result of the item, and demand amount information indicating a demand amount of the item not included in the PSI; a model generation unit that generates a trained model by learning the inventory placement amount change action using the inventory placement amount change information, the PSI plan performance information, and the demand amount information acquired by the first data acquisition unit as training data; and and the demand information includes one or more of a requested quantity of the item, a request source, a requested destination, a requested date, a requested delivery date, or a lead time from the requested date to the requested delivery date; The inference device a second data acquisition unit that acquires the PSI planned performance information and the demand amount information; an inference unit that inputs the PSI plan performance information and the demand quantity information acquired by the second data acquisition unit into the trained model to obtain the inventory placement quantity change action; an output unit that outputs inventory quantity change action information indicating the inventory quantity change action obtained by the inference unit; An inventory control system having:

2. the model generation unit uses the inventory placement quantity change information, the PSI plan performance information, and the demand quantity information acquired by the first data acquisition unit as learning data to learn the inventory placement quantity change action that improves inventory turnover or minimizes deterioration of inventory turnover; The inventory management system of claim 1 .

3. the first data acquisition unit further acquires actual transportation cost information indicating actual transportation costs due to the inventory placement amount change action, and inventory storage resource occupancy information indicating a storage resource occupancy rate of the product inventory at each base; the model generation unit uses the inventory placement quantity change information, the PSI plan actual result information, the demand quantity information, the transportation cost actual result information, and the inventory storage resource occupancy rate information acquired by the first data acquisition unit as learning data to learn the inventory placement quantity change action that improves inventory turnover or suppresses deterioration of inventory turnover, suppresses transportation costs, and ensures that the storage resource occupancy rate at each base satisfies an allowable value; The inventory management system of claim 1 .

4. The first data acquisition unit further acquires management target information indicating management target values ​​of inventory turnover and transportation costs; the model generation unit uses the inventory placement quantity change information, the PSI plan actual result information, the demand quantity information, the transportation cost actual result information, the inventory storage resource occupancy rate information, and the management target information acquired by the first data acquisition unit as learning data to learn the inventory placement quantity change action that improves the management target achievement rate of inventory turnover and transportation cost or suppresses a deterioration in the management target achievement rate of inventory turnover and transportation cost, and in which the storage resource occupancy rate at each base satisfies an allowable value; The inventory management system according to claim 3 .

5. a first data acquisition unit that acquires inventory allocation amount change information indicating a change in the inventory allocation amount of the item at each base due to an inventory allocation amount change action, PSI plan and actual result information indicating the planned and actual PSI of the item, and demand amount information indicating the demand amount of the item not included in the PSI; a model generation unit that generates a trained model by learning the inventory placement amount change action using the inventory placement amount change information, the PSI plan performance information, and the demand amount information acquired by the first data acquisition unit as training data; and Equipped with The demand information includes one or more of the requested quantity of the item, the request source, the requested destination, the requested date, the requested delivery date, or the lead time from the requested date to the requested delivery date.

6. an inference device that infers an inventory placement quantity change action using a trained model generated by training the inventory placement quantity change action, using as training data inventory placement quantity change information indicating a change in inventory placement quantity of an item at each base due to an inventory placement quantity change action, PSI plan and performance information indicating a PSI plan and performance of the item, and demand quantity information indicating a demand quantity of the item not included in the PSI, a second data acquisition unit that acquires the PSI planned performance information and the demand amount information; an inference unit that inputs the PSI plan performance information and the demand quantity information acquired by the second data acquisition unit into the trained model to obtain the inventory placement quantity change action; an output unit that outputs inventory quantity change action information indicating the inventory quantity change action obtained by the inference unit; Equipped with The demand information includes one or more of the requested quantity of the item, the request source, the requested destination, the requested date, the requested delivery date, or the lead time from the requested date to the requested delivery date.

7. The learning device executes a first data acquisition step of acquiring inventory allocation amount change information indicating a change in the inventory allocation amount of the item at each base due to an inventory allocation amount change action, PSI plan performance information indicating the planned and actual PSI of the item, and demand amount information indicating the demand amount of the item not included in the PSI; a model generation step of learning the inventory placement amount change action and generating a trained model using the inventory placement amount change information, the PSI plan performance information, and the demand amount information acquired in the first data acquisition step as training data; The inference device executes a second data acquisition step of acquiring the PSI planned performance information and the demand quantity information; an inference step of inputting the PSI plan performance information and the demand quantity information acquired in the second data acquisition step into the trained model to obtain the inventory placement quantity change action; and An inventory management method in which the demand information includes one or more of the requested quantity of the item, the request source, the requested destination, the requested date, the requested delivery date, or the lead time from the requested date to the requested delivery date.

8. Computer, a first data acquisition unit that acquires inventory allocation amount change information indicating a change in the inventory allocation amount of the item at each base due to an inventory allocation amount change action, PSI plan and actual result information indicating the planned and actual PSI of the item, and demand amount information indicating the demand amount of the item not included in the PSI; a model generation unit that uses the inventory placement amount change information, the PSI plan performance information, and the demand amount information acquired by the first data acquisition unit as learning data to learn an inventory placement amount change action for the item at each base and generates a trained model; It functions as The demand information includes one or more of the requested quantity of the item, the request source, the requested destination, the requested date, the requested delivery date, or the lead time from the requested date to the requested delivery date.

9. a computer that uses inventory placement quantity change information indicating a change in the inventory placement quantity of an item at each base due to an inventory placement quantity change action, PSI plan and actual result information indicating a plan and actual result of the PSI of the item, and demand quantity information indicating a demand quantity of the item not included in the PSI as learning data, and uses a trained model that has been generated by learning the inventory placement quantity change action; a second data acquisition unit that acquires the PSI planned performance information and the demand amount information; an inference unit that inputs the PSI plan performance information and the demand quantity information acquired by the second data acquisition unit into the trained model to obtain the inventory placement quantity change action; It functions as The demand information includes one or more of the requested quantity of the item, the request source, the requested destination, the requested date, the requested delivery date, or the lead time from the requested date to the requested delivery date.

Images