An intelligent supply chain demand forecasting and inventory optimization system

Abstract

The present invention relates to the technical field of inventory optimization, and specifically to an intelligent supply chain demand forecasting and inventory optimization system, comprising a supply chain dynamic demand forecasting module for realizing intelligent forecasting of supply chain demand, an inventory intelligent optimization module for intelligently optimizing inventory according to the demand forecast results of the supply chain, and a forecast correction and optimization adjustment module for dynamically correcting demand forecasts and intelligently optimizing and adjusting inventory. The present invention, through the establishment of a three-channel cross-validation prediction and anomaly rejection mechanism, can resist single-channel prediction anomalies, improve the stability and accuracy of supply chain demand forecasts, and reduce inventory risks caused by forecast errors. The present invention dynamically sets safety stock levels according to different stages, so that inventory strategies are more in line with changes in the commodity market, reducing out-of-stock rates and unsalable inventory. The correction amplitude is dynamically adjusted through error fluctuations to avoid over-correction, thereby improving the self-learning and self-evolution capabilities of the long-term supply chain forecasting system.

Description

Technical Field

[0001] The present invention relates to the technical field of inventory optimization, and in particular to an intelligent supply chain demand forecasting and inventory optimization system. Background Art

[0002] Current supply chain management systems generally suffer from single forecasting models and a lack of internal verification mechanisms, making demand forecasts susceptible to local anomalies and resulting in low accuracy. Inventory management is often based on static parameters and cannot dynamically adapt to changes in product lifecycles and market fluctuations, which can easily lead to stockouts or backlogs. Traditional inventory replenishment strategies rely on fixed thresholds and lack a secondary verification mechanism for replenishment results, resulting in delayed replenishment decisions. The lack of effective perception and feedback on supply chain fulfillment status prevents timely adjustments to inventory optimization strategies, further exacerbating supply chain uncertainty risks. Therefore, a comprehensive solution that can implement dynamic forecast correction, intelligent inventory optimization, and closed-loop feedback control of supply chain status is urgently needed to improve the accuracy, responsiveness, and stability of the supply chain system. Summary of the Invention

[0003] In view of the above-mentioned problems, the present invention is proposed.

[0004] To solve the above technical problems, the present invention provides the following technical solutions: an intelligent supply chain demand forecasting and inventory optimization system, comprising:

[0005] Supply chain dynamic demand forecasting module, specifically:

[0006] Collect historical product sales data and construct a time series feature set, which is then divided into three data sets that serve as control groups. A three-channel product forecasting model is then constructed to forecast products in different channels. By comparing the differences in product forecast results between channels, channels with deviating forecasts are eliminated, and the mean forecast value of the remaining channels is used as the supply chain demand forecast result.

[0007] Inventory intelligent optimization module, specifically:

[0008] The inventory safety score is calculated based on supply chain demand forecast results, historical fluctuation standard deviation, and supply cycle. Customer tolerance is adjusted according to the product life cycle, and the inventory safety score is dynamically adjusted. The optimized inventory target is calculated based on the adjusted inventory safety score, and a decision is made whether to trigger a replenishment strategy. A secondary verification of actual inventory levels ensures complete replenishment.

[0009] And the prediction correction and optimization adjustment modules, specifically:

[0010] By introducing an adaptive forecast deviation learning mechanism, the supply chain forecast results for future cycles are dynamically corrected and adjusted. Based on the corrected supply chain demand forecast results and the supply chain fulfillment status, the inventory target optimization amount is positively corrected.

[0011] As a preferred solution of the intelligent supply chain demand forecasting and inventory optimization system described in the present invention, the construction of the three-channel commodity forecasting model is specifically as follows:

[0012] Construct a three-channel commodity forecasting model, including the first forecasting channel, the second forecasting channel, and the third forecasting channel.

[0013] The input data of the first prediction channel is half of the first control group and half of the second control group, which are combined into a new data set;

[0014] The input data of the second prediction channel is half of the first control group and half of the third control group, which are combined into a new data set;

[0015] The input data of the third prediction channel is half of the second control group and half of the third control group, which are combined into a new data set.

[0016] As a preferred solution of the intelligent supply chain demand forecasting and inventory optimization system of the present invention, the commodity forecasting in different channels is specifically as follows:

[0017] For the product prediction of the first prediction channel, the product prediction is performed through the linear equation, then,

[0018]

[0019] in, represents the input data of the first prediction channel, Represent the weight coefficient and bias term of the linear equation respectively, Indicates the commodity forecast data of the first forecast channel at the current moment;

[0020] For the product prediction of the second prediction channel, the convolutional neural network is used to predict the product, then,

[0021]

[0022] in, represents the input data of the second prediction channel, Represent the weight coefficient and bias term of the convolutional neural network respectively, represents the activation function, Indicates the commodity forecast data of the second forecast channel at the current moment;

[0023] For the product prediction of the third layer prediction channel, the product prediction is performed through the multi-layer perceptron, then,

[0024]

[0025] in, represents the input data of the third prediction channel, Represent the weight coefficient and bias term of the multilayer perceptron respectively, represents the activation function, Indicates the product forecast data of the third forecast channel at the current moment.

[0026] As a preferred solution of the intelligent supply chain demand forecasting and inventory optimization system described in the present invention, the supply chain demand forecasting results are as follows:

[0027] Calculate the difference between the commodity prediction results of each channel, then we have,

[0028]

[0029] in, Indicates a time node, Indicates the total number of time nodes, Indicates the channel sequence number, Respectively represent and Product forecast data of the forecast channel, Indicates the difference between the product prediction results of each channel;

[0030] Set the difference threshold between channels , based on the comparison results of the calculated differences between the commodity prediction results of each channel and the set difference threshold, the accuracy of the three-channel commodity prediction is verified, specifically:

[0031] If the comparison result satisfies the formula , indicating the The difference between the channel and the other two channels exceeds the set difference threshold, which means that the The product prediction results of the prediction channel are inaccurate, so the first The commodity prediction result of the prediction channel and the average of the commodity prediction results of the remaining prediction channels are the supply chain demand prediction results. Conversely, it means that the prediction of the currently constructed three-channel commodity prediction model is accurate, and the average of the commodity prediction results of the three channels is the supply chain demand prediction result.

[0032] As a preferred solution of the intelligent supply chain demand forecasting and inventory optimization system described in the present invention, the inventory safety score is specifically as follows:

[0033] According to the supply chain forecast results , and obtain the current supply chain past Standard deviation of historical sales fluctuations at each time point , and the supply cycle of the current supply chain , to evaluate the current inventory safety supply level, we have,

[0034]

[0035] in, Indicates the current supply chain past The standard deviation of historical sales fluctuations at each time point, represents the supply cycle of the current supply chain, Represents the calculated inventory safety score, which is used for the current inventory safety supply level. It represents user tolerance, which is the user's tolerance for supply chain out-of-stock situations and is regulated by the product life cycle of the supply chain.

[0036] As a preferred solution of the intelligent supply chain demand forecasting and inventory optimization system of the present invention, the product life cycle is specifically as follows:

[0037] Set market activity index based on product search popularity ;

[0038] Calculate the sales change rate of a product during its life cycle, then we have:

[0039]

[0040] in, Indicates the sales volume of the product at the current time. Indicates the sales volume of the product at the previous moment. Indicates the sales rate change of the product at the current moment;

[0041] At the same time, set the judgment threshold for each stage of the life cycle, including the sales volume change rate threshold for the growth stage and the market activity index during the growth phase , the sales volume change rate threshold of mature products and the market activity index in the mature stage , the sales volume change rate threshold of the recession period and the market activity index during the recession ;

[0042] According to the set judgment threshold, the supply chain product life cycle is judged, then:

[0043] If the product sales change rate and market activity index satisfy the formula , indicating that the current supply chain products are in the growth stage;

[0044] If the product sales change rate and market activity index satisfy the formula , indicating that the current supply chain products are in the mature stage;

[0045] If the product sales change rate and market activity index satisfy the formula , indicating that the current supply chain commodities are in a recession period;

[0046] Otherwise, the supply chain goods are in the introduction stage.

[0047] As a preferred solution of the intelligent supply chain demand forecasting and inventory optimization system described in the present invention, the calculation of the inventory target optimization amount based on the adjusted inventory safety score is specifically as follows:

[0048]

[0049] in, represents the supply chain demand forecast result, represents the supply chain commodity supply cycle, represents the calculated stock safety score, Indicates the dynamically adjusted inventory target optimization quantity;

[0050] According to the dynamically adjusted inventory target optimization quantity, determine whether to trigger the replenishment strategy, then,

[0051] Collect the actual inventory at the current moment By comparing the actual inventory level with the dynamically adjusted inventory target optimization level, the replenishment strategy is determined based on the comparison results. Specifically:

[0052] Set replenishment policy trigger thresholds ;

[0053] If the actual inventory quantity is compared with the dynamically adjusted inventory target optimization quantity, the formula is satisfied. , it means that the replenishment strategy is triggered, otherwise, the replenishment strategy is not triggered.

[0054] As a preferred solution of the intelligent supply chain demand forecasting and inventory optimization system of the present invention, the dynamic correction and adjustment of the supply chain forecast results of the future cycle is specifically as follows:

[0055] Based on past time periods Historical demand forecast results of the supply chain within , and the corresponding actual sales data , calculate the supply chain forecast error , then there is,

[0056]

[0057] And calculate the mean of historical forecast errors and standard deviation , dynamic offset correction is performed on the future prediction value, then,

[0058]

[0059] in, represents the supply chain demand forecast result, represents the revised supply chain demand forecast result, Represents the introduced learning rate.

[0060] As a preferred solution of the intelligent supply chain demand forecasting and inventory optimization system described in the present invention, the adaptive forecast deviation learning mechanism is specifically as follows:

[0061] According to the standard deviation and mean of the supply chain forecast error in the recent time window, the learning rate that changes dynamically over time is introduced, then,

[0062]

[0063] in, Indicates the maximum and minimum values of the learning rate, Represent the mean and standard deviation of the supply chain forecast in the recent time window, Represents the learning rate introduced that changes dynamically over time, specifically:

[0064] If the mean supply chain forecast error at the current moment exceeds the mean supply chain forecast error in the recent time window, the learning rate introduced is increased to strengthen the correction of future forecast values;

[0065] If the mean supply chain forecast error at the current moment is smaller than the mean supply chain forecast error in the recent time window, the introduced learning rate is reduced to reduce the correction of future forecast values.

[0066] As a preferred solution of the intelligent supply chain demand forecasting and inventory optimization system described in the present invention, the forward correction inventory target optimization amount is specifically as follows:

[0067] Collect the current supply chain planned replenishment time and the actual replenishment arrival time, and use the difference between the two as the supply chain fulfillment status, specifically:

[0068]

[0069] in, represents the supply chain planned replenishment time, Indicates the actual arrival time. Represents the supply chain fulfillment difference, which is used to judge the current supply chain fulfillment status. Specifically:

[0070] For the supply chain performance difference of the current supply chain, a periodic calculation is performed. If the supply chain performance difference satisfies the formula for three consecutive times , it means that the actual arrival time is later than the planned arrival time, and the safety score of the current inventory is dynamically improved. , in order to achieve the adjustment of inventory target optimization quantity;

[0071] According to the corrected forecast value and supply fulfillment status, the target inventory optimization quantity is calculated again, and we have:

[0072]

[0073] in, represents the revised supply chain demand forecast result, represents the inventory safety score dynamically adjusted by the supply chain fulfillment status, It represents the supply cycle of the current supply chain and the target optimization amount of inventory calculated twice.

[0074] Beneficial effects of the present invention:

[0075] By establishing a three-channel cross-validation prediction and anomaly elimination mechanism, the present invention can resist single-channel prediction anomalies, improve the stability and accuracy of supply chain demand forecasts, and reduce inventory risks caused by forecast errors;

[0076] Dynamically set safety stock levels according to different stages to make inventory strategies more in line with commodity market changes, reducing out-of-stock rates and slow-moving inventory;

[0077] Dynamically adjust the correction amplitude through error fluctuations to avoid over-correction or under-correction, and improve the self-learning and self-evolution capabilities of the long-term supply chain forecasting system;

[0078] It can also perceive the actual arrival status of the supply chain in real time, make positive corrections to forecasting and inventory strategies, and improve the system's ability to respond to supply chain anomalies and ensure supply.

[0079] At the same time, based on forecast results, historical fluctuations and supply cycles, inventory targets are dynamically adjusted to match real-time market demand changes and optimize inventory turnover efficiency and capital utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort. Among them:

[0081] Figure 1 This is a schematic diagram of the overall method steps of an intelligent supply chain demand forecasting and inventory optimization system of the present invention. DETAILED DESCRIPTION

[0082] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0083] Example 1, reference Figure 1 , which is the first embodiment of the present invention, provides an intelligent supply chain demand forecasting and inventory optimization system, including a supply chain dynamic demand forecasting module, an inventory intelligent optimization module, and a forecast correction and optimization adjustment module;

[0084] Specifically, the supply chain dynamic demand forecasting module is used to realize intelligent forecasting of supply chain demand, the inventory intelligent optimization module is used to perform intelligent inventory optimization based on the supply chain demand forecast results, and the forecast correction and optimization adjustment module is used to dynamically correct demand forecasts and intelligently optimize and adjust inventory;

[0085] Furthermore, the supply chain dynamic demand forecasting module realizes dynamic forecasting of supply demand based on the historical sales data of commodities, which is specifically implemented as follows:

[0086] Collecting the past The historical sales data of commodities at each time point is used to build a time series feature set based on the historical sales data of commodities. , then there is, ,in, Indicates the total number of time nodes, Indicates the current time Product sales data, Represents the constructed time series feature set;

[0087] Divide the constructed time series feature set into three data sets that serve as control groups, then we have: ,in, Represents the divided control group data set, namely the first control group, the second control group and the third control group, and satisfies the formula , indicating that the number of data in each control group data set is the same.

[0088] Product forecasts are performed on the divided control group data, and the mean of the product forecast results of all control groups is calculated. The mean of the calculated product forecast results is the supply chain demand forecast result. The specific forecast is as follows:

[0089] Construct a three-channel commodity forecasting model, including the first forecasting channel, the second forecasting channel, and the third forecasting channel.

[0090] The input data of the first prediction channel is half of the first control group and half of the second control group, which are combined into a new data set;

[0091] The input data of the second prediction channel is half of the first control group and half of the third control group, which are combined into a new data set;

[0092] The input data of the third prediction channel is half of the second control group and half of the third control group, which are combined into a new data set.

[0093] If we make product forecasts in different forecast channels, we have:

[0094] For the product prediction of the first prediction channel, the product prediction is performed through the linear equation, then,

[0095]

[0096] in, represents the input data of the first prediction channel, They represent the weight coefficient and bias term of the linear equation respectively, which are set by the implementer according to the actual application scenario. Indicates the commodity forecast data of the first forecast channel at the current moment;

[0097] For the product prediction of the second prediction channel, the convolutional neural network is used to predict the product, then,

[0098]

[0099] in, represents the input data of the second prediction channel, They represent the weight coefficient and bias term of the convolutional neural network respectively, which are set by the implementer according to the actual application scenario. represents the activation function, Indicates the commodity forecast data of the second forecast channel at the current moment;

[0100] For the product prediction of the third layer prediction channel, the product prediction is performed through the multi-layer perceptron, then,

[0101]

[0102] in, represents the input data of the third prediction channel, They represent the weight coefficient and bias term of the multi-layer perceptron, which are set by the implementer according to the actual application scenario. represents the activation function, Indicates the product forecast data of the third forecast channel at the current moment.

[0103] Based on the prediction results of each channel, the accuracy of the three-channel commodity prediction model is verified, specifically:

[0104] Calculate the difference between the commodity prediction results of each channel, then we have,

[0105]

[0106] in, Indicates a time node, Indicates the total number of time nodes, Indicates the channel sequence number, Respectively represent and Product forecast data of the forecast channel, Indicates the difference between the product prediction results of each channel, specifically:

[0107] Indicates the difference between the first prediction channel and the second prediction channel;

[0108] Indicates the difference between the first prediction channel and the third prediction channel;

[0109] Indicates the difference between the second prediction channel and the third prediction channel.

[0110] Set the difference threshold between channels , based on the comparison results of the calculated differences between the commodity prediction results of each channel and the set difference threshold, the accuracy of the three-channel commodity prediction is verified, specifically:

[0111] If the comparison result satisfies the formula , indicating the The difference between the channel and the other two channels exceeds the set difference threshold, which means that the The product prediction results of the prediction channel are inaccurate, so the first The commodity prediction result of the prediction channel and the average of the commodity prediction results of the remaining prediction channels are the supply chain demand prediction results. Conversely, it means that the prediction of the currently constructed three-channel commodity prediction model is accurate, and the average of the commodity prediction results of the three channels is the supply chain demand prediction result.

[0112] Furthermore, the intelligent inventory optimization module performs intelligent inventory optimization based on the forecast results of the supply chain dynamic demand forecast module. The specific optimization is as follows:

[0113] Based on the forecast results of the supply chain, an inventory safety scoring mechanism is constructed. At the same time, according to the calculated inventory safety score, the inventory target optimization quantity is dynamically adjusted, and combined with the current actual inventory quantity, the replenishment strategy is dynamically adjusted to achieve intelligent optimization of commodity inventory.

[0114] Furthermore, the inventory safety scoring mechanism evaluates the current inventory safety supply level based on supply chain forecast results, combined with historical fluctuation coefficients and supply response time, as follows:

[0115] According to the supply chain forecast results , and obtain the current supply chain past Standard deviation of historical sales fluctuations at each time point , and the supply cycle of the current supply chain , to evaluate the current inventory safety supply level, we have,

[0116]

[0117] in, Indicates the current supply chain past The standard deviation of historical sales fluctuations at each time point, represents the supply cycle of the current supply chain, Represents the calculated inventory safety score, which is used for the current inventory safety supply level. Indicates user tolerance, which is the user's tolerance for supply chain out-of-stock situations. It is regulated by the product life cycle of the supply chain and is specifically:

[0118] Divide the supply chain products into four stages: introduction, growth, maturity and decline. Regulate user tolerance according to the supply chain product cycle. Then,

[0119] Set market activity index based on product search popularity , which is determined by the number of searches for the product within the current time period. The current time period is set by the implementer based on the actual application scenario;

[0120] Calculate the sales change rate of a product during its life cycle, then we have:

[0121]

[0122] in, Indicates the sales volume of the product at the current time. Indicates the sales volume of the product at the previous moment. Indicates the sales rate change of the product at the current moment;

[0123] At the same time, set the judgment threshold for each stage of the life cycle, including the sales volume change rate threshold for the growth stage and the market activity index during the growth phase , the sales volume change rate threshold of mature products and the market activity index in the mature stage , the sales volume change rate threshold of the recession period and the market activity index during the recession ;

[0124] According to the set judgment threshold, the supply chain product life cycle is judged, then:

[0125] If the product sales change rate and market activity index satisfy the formula , indicating that the current supply chain products are in the growth stage;

[0126] If the product sales change rate and market activity index satisfy the formula , indicating that the current supply chain products are in the mature stage;

[0127] If the product sales change rate and market activity index satisfy the formula , indicating that the current supply chain commodities are in a recession period;

[0128] Otherwise, the supply chain goods are in the introduction stage.

[0129] It should be noted that user tolerance is set based on the judged product life cycle. The specific value is set by the implementer according to the actual application scenario, but the user tolerance is set from high to low in the order of introduction period, growth period, maturity period and decline period.

[0130] Based on the calculated inventory safety score, the inventory target optimization quantity is dynamically adjusted, and then,

[0131]

[0132] in, represents the supply chain demand forecast result, Indicates the supply cycle of supply chain products, which is set by the implementer based on the actual product supply cycle. represents the calculated stock safety score, Indicates the dynamically adjusted inventory target optimization quantity;

[0133] According to the dynamically adjusted inventory target optimization quantity, determine whether to trigger the replenishment strategy, then,

[0134] Collect the actual inventory at the current moment By comparing the actual inventory level with the dynamically adjusted inventory target optimization level, the replenishment strategy is determined based on the comparison results. Specifically:

[0135] Set replenishment policy trigger thresholds , which is set by the implementers according to the actual application scenario;

[0136] If the actual inventory quantity is compared with the dynamically adjusted inventory target optimization quantity, the formula is satisfied. , it means that the replenishment strategy is triggered, otherwise, the replenishment strategy is not triggered.

[0137] It should be noted that for inventory that triggers the replenishment policy, after the goods are replenished, the actual inventory level will be checked again until the actual inventory level exceeds the set replenishment policy triggering threshold.

[0138] Furthermore, the forecast correction and optimization adjustment module compares the errors between actual sales data in the historical period and the dynamic demand forecast of the supply chain. At the same time, through the adaptive forecast deviation learning mechanism, it dynamically corrects and adjusts the supply chain forecast results of the future period. Based on the corrected supply chain demand forecast results, it re-evaluates the inventory target optimization amount and, combined with the actual replenishment fulfillment coefficient, positively corrects the inventory target optimization amount. The specific implementation is as follows:

[0139] Based on past time periods Supply chain demand forecast results within , and the corresponding actual sales data , calculate the supply chain forecast error, then we have,

[0140]

[0141] in, represents the historical demand forecast results of the supply chain, Indicates the actual sales data corresponding to the historical moment, represents the calculated supply chain forecast error;

[0142] And calculate the mean of historical forecast errors and standard deviation , to achieve dynamic offset correction of future prediction values, then we have,

[0143]

[0144] in, represents the supply chain demand forecast result, represents the mean of the supply chain forecast error, represents the revised supply chain demand forecast result, Represents the learning rate introduced, which changes dynamically over time and controls the deviation correction amplitude, as follows:

[0145] According to the standard deviation and mean of the supply chain forecast error in the recent time window (such as the last 10 days), the learning rate that changes dynamically over time is introduced, then,

[0146]

[0147] in, Indicates the maximum and minimum values of the learning rate. The specific values are set by the implementer according to the actual application scenario. represents the mean and standard deviation of the supply chain forecast in the recent time window, Represents the learning rate introduced to dynamically change over time, which is used to achieve dynamic offset correction of future prediction values. Specifically:

[0148] If the mean supply chain forecast error at the current moment exceeds the mean supply chain forecast error in the recent time window, the learning rate introduced is increased to strengthen the correction of future forecast values;

[0149] If the mean supply chain forecast error at the current moment is smaller than the mean supply chain forecast error in the recent time window, the introduced learning rate is reduced to reduce the correction of future forecast values.

[0150] It should be noted that based on the corrected forecast value and the supply chain fulfillment status, the second calculation of the inventory target optimization quantity is:

[0151] Collect the current supply chain planned replenishment time and the actual replenishment arrival time, and use the difference between the two as the supply chain fulfillment status, specifically:

[0152]

[0153] in, represents the supply chain planned replenishment time, Indicates the actual arrival time. Represents the supply chain fulfillment difference, which is used to judge the current supply chain fulfillment status. Specifically:

[0154] For the supply chain performance difference of the current supply chain, a periodic calculation is performed. If the supply chain performance difference satisfies the formula for three consecutive times , it means that the actual arrival time is later than the planned arrival time, and the safety score of the current inventory is dynamically improved. , in order to achieve the adjustment of inventory target optimization quantity;

[0155] According to the corrected forecast value and supply fulfillment status, the target inventory optimization quantity is calculated again, and we have:

[0156]

[0157] in, represents the revised supply chain demand forecast result, represents the inventory safety score dynamically adjusted by the supply chain fulfillment status, It represents the supply cycle of the current supply chain and the target optimization amount of inventory calculated twice.

[0158] It should be noted that the present invention constructs a supply chain demand forecasting system based on three-channel cross-validation and anomaly rejection mechanisms, combines the dynamic regulation of inventory tolerance with the product life cycle, further introduces adaptive forecast correction and dynamic learning rate mechanisms, and adjusts inventory strategies based on supply chain fulfillment status feedback, thus forming a complete intelligent closed-loop control system for supply chain forecasting, inventory optimization, and forecast correction.

[0159] Through the collaboration between various modules, the accuracy and stability of supply chain demand forecasts have been effectively improved, inventory management has been made intelligent, adaptive and efficient, the dual risks of supply chain out-of-stock and inventory backlogs have been reduced, and the overall resilience and operational efficiency of the supply chain system have been enhanced.

[0160] Furthermore, if the functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0161] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0162] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, and then editing, interpreting, or processing in another suitable manner as necessary, and then storing it in a computer memory.

[0163] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An intelligent supply chain demand forecasting and inventory optimization system, characterized by: include, Supply chain dynamic demand forecasting module, specifically: Collect historical product sales data and construct a time series feature set, which is then divided into three data sets that serve as control groups. A three-channel product forecasting model is then constructed to forecast products in different channels. By comparing the differences in product forecast results between channels, channels with deviating forecasts are eliminated, and the mean forecast value of the remaining channels is used as the supply chain demand forecast result. Inventory intelligent optimization module, specifically: The inventory safety score is calculated based on supply chain demand forecast results, historical fluctuation standard deviation, and supply cycle. Customer tolerance is adjusted according to the product life cycle, and the inventory safety score is dynamically adjusted. The optimized inventory target is calculated based on the adjusted inventory safety score, and a decision is made whether to trigger a replenishment strategy. A secondary verification of actual inventory levels ensures complete replenishment. The inventory safety scores are as follows: According to the supply chain forecast results , and obtain the current supply chain past Standard deviation of historical sales fluctuations at each time point , and the supply cycle of the current supply chain , to evaluate the current inventory safety supply level, we have, in, Indicates the current supply chain past The standard deviation of historical sales fluctuations at each time point, represents the supply cycle of the current supply chain, Represents the calculated inventory safety score, which is used for the current inventory safety supply level. Represents user tolerance, which is the user's tolerance for supply chain out-of-stock situations and is regulated by the product life cycle of the supply chain; The calculation of the inventory target optimization amount based on the adjusted inventory safety score is as follows: in, represents the supply chain demand forecast result, represents the supply chain commodity supply cycle, represents the calculated stock safety score, Indicates the dynamically adjusted inventory target optimization quantity; According to the dynamically adjusted inventory target optimization quantity, determine whether to trigger the replenishment strategy, then, Collect the actual inventory at the current moment By comparing the actual inventory level with the dynamically adjusted inventory target optimization level, the replenishment strategy is determined based on the comparison results. Specifically: Set replenishment policy trigger thresholds ; If the actual inventory quantity is compared with the dynamically adjusted inventory target optimization quantity, the formula is satisfied. , it means that the replenishment strategy is triggered, otherwise, the replenishment strategy is not triggered; And the prediction correction and optimization adjustment modules, specifically: By introducing an adaptive forecast deviation learning mechanism, the supply chain forecast results for future cycles are dynamically corrected and adjusted. Based on the corrected supply chain demand forecast results and the supply chain fulfillment status, the inventory target optimization amount is positively corrected.

2. The intelligent supply chain demand forecasting and inventory optimization system according to claim 1 is characterized in that: The construction of the three-channel commodity prediction model is specifically as follows: Construct a three-channel commodity forecasting model, including the first forecasting channel, the second forecasting channel, and the third forecasting channel. The input data of the first prediction channel is half of the first control group and half of the second control group, which are combined into a new data set; The input data of the second prediction channel is half of the first control group and half of the third control group, which are combined into a new data set; The input data of the third prediction channel is half of the second control group and half of the third control group, which are combined into a new data set.

3. The intelligent supply chain demand forecasting and inventory optimization system according to claim 2 is characterized in that: The specific details of performing commodity forecasting in different channels are as follows: For the product prediction of the first prediction channel, the product prediction is performed through the linear equation, then, in, represents the input data of the first prediction channel, Represent the weight coefficient and bias term of the linear equation respectively, Indicates the commodity forecast data of the first forecast channel at the current moment; For the product prediction of the second prediction channel, the convolutional neural network is used to predict the product, then, in, represents the input data of the second prediction channel, Represent the weight coefficient and bias term of the convolutional neural network respectively, represents the activation function, Indicates the commodity forecast data of the second forecast channel at the current moment; For the product prediction of the third layer prediction channel, the product prediction is performed through the multi-layer perceptron, then, in, represents the input data of the third prediction channel, Represent the weight coefficient and bias term of the multilayer perceptron respectively, represents the activation function, Indicates the product forecast data of the third forecast channel at the current moment.

4. The intelligent supply chain demand forecasting and inventory optimization system according to claim 3 is characterized in that: The supply chain demand forecast results are as follows: Calculate the difference between the commodity prediction results of each channel, then we have, in, Indicates a time node, Indicates the total number of time nodes, Indicates the channel sequence number, Respectively represent and Product forecast data of the forecast channel, Indicates the difference between the product prediction results of each channel; Set the difference threshold between channels , based on the comparison results of the calculated differences between the commodity prediction results of each channel and the set difference threshold, the accuracy of the three-channel commodity prediction is verified, specifically: If the comparison result satisfies the formula , indicating the The difference between the channel and the other two channels exceeds the set difference threshold, which means that the The product prediction results of the prediction channel are inaccurate, so the first The commodity prediction result of the prediction channel and the average of the commodity prediction results of the remaining prediction channels are the supply chain demand prediction results. Conversely, it means that the prediction of the currently constructed three-channel commodity prediction model is accurate, and the average of the commodity prediction results of the three channels is the supply chain demand prediction result.

5. The intelligent supply chain demand forecasting and inventory optimization system according to claim 4 is characterized in that: The product life cycle is as follows: Set market activity index based on product search popularity ; Calculate the sales change rate of a product during its life cycle, then we have: in, Indicates the sales volume of the product at the current time. Indicates the sales volume of the product at the previous moment. Indicates the sales rate change of the product at the current moment; At the same time, set the judgment threshold for each stage of the life cycle, including the sales volume change rate threshold for the growth stage and the market activity index during the growth phase , the sales volume change rate threshold of mature products and the market activity index in the mature stage , the sales volume change rate threshold of the recession period and the market activity index during the recession ; According to the set judgment threshold, the supply chain product life cycle is judged, then: If the product sales change rate and market activity index satisfy the formula , indicating that the current supply chain products are in the growth stage; If the product sales change rate and market activity index satisfy the formula , indicating that the current supply chain products are in the mature stage; If the product sales change rate and market activity index satisfy the formula , indicating that the current supply chain commodities are in a recession period; Otherwise, the supply chain goods are in the introduction stage.

6. The intelligent supply chain demand forecasting and inventory optimization system according to claim 5, characterized in that: The dynamic correction adjustment of the supply chain forecast results for the future cycle is specifically as follows: Based on past time periods Historical demand forecast results of the supply chain within , and the corresponding actual sales data , calculate the supply chain forecast error , then there is, And calculate the mean of historical forecast errors and standard deviation , dynamic offset correction is performed on the future prediction value, then, in, represents the supply chain demand forecast result, represents the revised supply chain demand forecast result, Represents the introduced learning rate.

7. The intelligent supply chain demand forecasting and inventory optimization system according to claim 6, characterized in that: The adaptive prediction bias learning mechanism is specifically as follows: According to the standard deviation and mean of the supply chain forecast error in the recent time window, the learning rate that changes dynamically over time is introduced, then, in, Indicates the maximum and minimum values of the learning rate, Represent the mean and standard deviation of the supply chain forecast in the recent time window, Represents the learning rate introduced that changes dynamically over time, specifically: If the mean supply chain forecast error at the current moment exceeds the mean supply chain forecast error in the recent time window, the learning rate introduced is increased to strengthen the correction of future forecast values; If the mean supply chain forecast error at the current moment is smaller than the mean supply chain forecast error in the recent time window, the introduced learning rate is reduced to reduce the correction of future forecast values.

8. The intelligent supply chain demand forecasting and inventory optimization system according to claim 7, characterized in that: The positive correction inventory target optimization amount is specifically as follows: Collect the current supply chain planned replenishment time and the actual replenishment arrival time, and use the difference between the two as the supply chain fulfillment status, specifically: in, represents the supply chain planned replenishment time, Indicates the actual arrival time. Represents the supply chain fulfillment difference, which is used to judge the current supply chain fulfillment status. Specifically: For the supply chain performance difference of the current supply chain, a periodic calculation is performed. If the supply chain performance difference satisfies the formula for three consecutive times , it means that the actual arrival time is later than the planned arrival time, and the safety score of the current inventory is dynamically improved. , in order to achieve the adjustment of inventory target optimization quantity; According to the corrected forecast value and supply fulfillment status, the target inventory optimization quantity is calculated again, and we have: in, represents the revised supply chain demand forecast result, represents the inventory safety score dynamically adjusted by the supply chain fulfillment status, It represents the supply cycle of the current supply chain and the target optimization amount of inventory calculated twice.

Images