# US 12,602,649 B2 · Predicting supply chain policies using machine learning > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,602,649 B2 - Document ID: US12602649B2 - Status: Granted - Kind: B2 - Application number: 17/698,666 - Filed: 2022-03-18 - Published: 2023-09-21 - Granted: 2026-04-14 - Earliest claimed: 2022-03-18 - Assignees: Oracle International Corporation - Applicants: Oracle International Corporation - Inventors: Setareh Borjian; Su-Ming Wu; Shenghao Wang - Web page: https://patentgrep.com/patent/US12602649B2 ## CPC - G06Q 10/087: Inventory or stock management, e.g. order filling, procurement or balancing against orders - G06Q 10/08726: {for replenishment processing, procedures, or recommendations using forecasting or optimisation} - G06Q 30/0605: {Pooling transaction partners, e.g. group buying or group selling} - G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR ## Abstract Embodiments predict supply chain policies using machine learning. A machine learning model trained to predict one or more supply chain metrics for a first product can be stored. The machine learning model can generate a plurality of supply chain metric predictions for the first product using a plurality of candidate replenishment policies for the first product. A candidate replenishment policy with a corresponding supply chain metric prediction that meets a criteria can be selected. The selected replenishment policy can be implemented for the first product within an inventory system, where one or more physical locations are restocked with the first product based on restocking parameters defined by the selected replenishment policy. ## Claims ### Claim 1 A method for predicting supply chain policies using machine learning, the method comprising: generating, using simulation software and one or more processors, simulated training data by performing simulations, for a first set of products and variety of replenishment policies, to generate simulated supply chain metric data labels; generating mixed training data by combining the simulated trained data with historic training data, wherein the historic training data corresponds to historic supply chain metric data labels for a second set of products; training, using the one or more processors via supervised learning, a machine learning model, using the mixed training data via the simulated supply chain metric data labels and the historic supply chain metric data labels, to generate supply chain predictions for one or more products; generating, by the trained machine learning model using the one or more processors, a plurality of supply chain metric predictions for a first product using a plurality of candidate replenishment policies for the first product, wherein, for each of the candidate replenishment policies, the trained machine learning model generates a set of supply chain metric predictions that represents predicted performance of the first product, and wherein the simulation software performs the simulations without generating supply chain predictions via the trained machine learning model; selecting, using the one or more processors, a candidate replenishment policy with a corresponding set of supply chain metric predictions that meets a criteria; and implementing, using the one or more processors, the selected replenishment policy for the first product within an inventory system, wherein one or more physical locations are restocked with the first product based on restocking parameters defined by the selected replenishment policy. ### Claim 2 (depends on 1) The method of claim 1, wherein the sets of supply chain metric predictions are generated by the machine learning model using input for the first product, the input comprising the candidate replenishment policies, product features for the first product, and demand features for the first product. ### Claim 3 (depends on 2) The method of claim 2, wherein the demand features for the first product comprise one or more of mean and standard deviation for product demand, and the product features for the first product comprise one or more of lead time for replenishment, pack size, and minimum order quantity. ### Claim 4 (depends on 1) The method of claim 1, further comprising: ranking the sets of supply chain metric predictions, wherein the selected candidate replenishment policy corresponds to a highest ranked set of supply chain metric predictions. ### Claim 5 (depends on 1) The method of claim 1, wherein the sets of supply chain metric predictions comprise a lost sales metric for the first product, and the sets of supply chain metric predictions are ranked according to the lost sales metric. ### Claim 6 (depends on 5) The method of claim 5, wherein the sets of supply chain metric predictions comprise the lost sales metric and an inventory on hand metric for the first product. ### Claim 7 (depends on 1) The method of claim 1, wherein the candidate replenishment policies comprise a minimum stock level and maximum stock level for the first product. ### Claim 8 (depends on 1) The method of claim 1, wherein the machine learning model comprises one of more of a random forest model and a XG boost model. ### Claim 9 (depends on 1) The method of claim 1, wherein the simulated labeled training data comprises instances of the first set of products, demand features for the product instances, replenishment policies for the product instances, and the simulated supply chain metric data labels. ### Claim 10 A non-transitory computer readable medium having instructions stored thereon that, when executed by a processor, cause the processor to predict supply chain policies using machine learning, wherein, when executed, the instructions cause the processor to: generate, using simulation software, simulated training data by performing simulations, for a first set of products and variety of replenishment policies, to generate simulated supply chain metric data labels; generate mixed training data by combining the simulated trained data with historic training data, wherein the historic training data corresponds to historic supply chain metric data labels for a second set of products; train, via supervised learning, a machine learning model, using the mixed training data via the simulated supply chain metric data labels and the historic supply chain metric data labels, to generate supply chain predictions for one or more products; generate, by the trained machine learning model, a plurality of supply chain metric predictions for a first product using a plurality of candidate replenishment policies for the first product, wherein, for each of the candidate replenishment policies, the trained machine learning model generates a set of supply chain metric predictions that represents predicted performance of the first product, and wherein the simulation software performs the simulations without generating supply chain predictions via the trained machine learning model; select a candidate replenishment policy with a corresponding supply chain metric prediction that meets a criteria; and implement the selected replenishment policy for the first product within an inventory system, wherein one or more physical locations are restocked with the first product based on restocking parameters defined by the selected replenishment policy. ### Claim 11 (depends on 10) The computer readable medium of claim 10, wherein the sets of supply chain metric predictions are generated by the machine learning model using input for the first product, the input comprising the candidate replenishment policies, product features for the first product, and demand features for the first product. ### Claim 12 (depends on 11) The computer readable medium of claim 11, wherein the demand features for the first product comprise one or more of mean and standard deviation for product demand, and the product features for the first product comprise one or more of lead time for replenishment, pack size, and minimum order quantity. ### Claim 13 (depends on 10) The computer readable medium of claim 10, wherein, when executed, the instructions further cause the processor to: ranking the sets of supply chain metric predictions, wherein the selected candidate replenishment policy corresponds to a highest ranked set of supply chain metric predictions. ### Claim 14 (depends on 10) The computer readable medium of claim 10, wherein the sets of supply chain metric predictions comprise a lost sales metric for the first product, and the sets of supply chain metric predictions are ranked according to the lost sales metric. ### Claim 15 (depends on 14) The computer readable medium of claim 14, wherein the sets of supply chain metric predictions comprise the lost sales metric and an inventory on hand metric for the first product. ### Claim 16 (depends on 10) The computer readable medium of claim 10, wherein the candidate replenishment policies comprise a minimum stock level and maximum stock level for the first product. ### Claim 17 A system for predicting supply chain policies using machine learning, the system comprising: a processor; and a memory storing instructions for execution by the processor, the instructions configuring the processor to: generate, using simulation software, simulated training data by performing full-scale simulations, for a first set of products and variety of replenishment policies, to generate simulated supply chain metric data labels; generate mixed training data by combining the simulated trained data with historic training data, wherein the historic training data corresponds to historic supply chain metric data labels for a second set of products; train, via supervised learning, a machine learning model, using the mixed training data via the simulated supply chain metric data labels and the historic supply chain metric data labels, to generate supply chain predictions for one or more products; generate, by the trained machine learning model, a plurality of supply chain metric predictions for a first product using a plurality of candidate replenishment policies for the first product, wherein, for each of the candidate replenishment policies, the trained machine learning model generates a set of supply chain metric predictions that represents predicted performance of the first product, and wherein the simulation software performs the simulations without generating supply chain predictions via the trained machine learning model; select a candidate replenishment policy with a corresponding supply chain metric prediction that meets a criteria; and implement the selected replenishment policy for the first product within an inventory system, wherein one or more physical locations are restocked with the first product based on restocking parameters defined by the selected replenishment policy. ### Claim 18 (depends on 1) The method of claim 1, wherein training instances of the simulated training data comprise demand features for the first set of products, training instances of the historic training data comprise demand features for the second set of products, and the sets of supply chain metric predictions are generated by the trained machine learning model using input for the first product, the input comprising the candidate replenishment policies and the demand features for the first product. ### Claim 19 (depends on 18) The method of claim 18, wherein the demand features comprise one or more of mean and standard deviation for product demand. ### Claim 20 (depends on 1) The method of claim 1, wherein the trained machine learning model generates the supply chain predictions with greater computational efficiency than the simulation software performs the simulations.