A food material inventory dynamic management method and system

Abstract

The application discloses a kind of food material inventory dynamic management method and system, comprising: through group meal food material distribution system, obtain order period and the basic correlation data set of multiple food material inventory life cycle;Order period influence feature set is constructed, and the life cycle influence weight data of each food material category is determined;Using probability distribution fitting method, obtain multiple food material consumption probability data set;Identify distribution mismatched inventory unit, determine the optimized inventory space distribution scheme;Order type identification model is constructed, and order type is identified, and generate inventory level demand interval;Multiple food material inventory adjustment instruction is generated, and the inventory of each food material category is adjusted;The consumption deviation value of different food materials is calculated, and multiple food material life cycle influence weight data is updated.The application can improve the matching degree between inventory configuration and actual group meal demand under the premise of guaranteeing food material life cycle constraint, reduce inventory backlog and shortage risk.

Description

Technical Field

[0001] This invention relates to the fields of supply chain management and intelligent warehousing technology, and in particular to a method and system for dynamic management of food inventory. Background Technology

[0002] With the widespread application of group catering services in schools, enterprises, industrial parks, and public institutions, the scale of group catering operations continues to expand, and the management complexity of food procurement, storage, and distribution is constantly increasing. Group catering businesses typically involve the centralized supply of multiple categories and batches of ingredients, requiring not only to meet the needs of different meal cycles and order sizes but also to be strictly constrained by the requirements for food shelf life, freshness, and safety management. Therefore, how to achieve reasonable allocation of inventory resources while ensuring the constraints of the food's lifecycle has become a key issue in group catering operation management. Existing group catering ingredient inventory management methods mostly adopt experience-based rules or fixed threshold management models, typically using historical average consumption as the main basis for inventory replenishment and adjustment. This method fails to fully consider the cyclical fluctuations of group catering orders over different time periods, and also fails to reflect the impact of changes in order structure on ingredient demand during special scenarios such as holidays, temporary events, and concentrated meetings. Furthermore, existing inventory management systems usually focus on total inventory or single-dimensional management, lacking a refined description of the food inventory lifecycle. Different batches of ingredients exhibit significant differences in their warehousing time, remaining shelf life, and actual consumption patterns. However, traditional management methods struggle to effectively map these differences to specific inventory units, hindering targeted optimization of inventory space layout. This can lead to situations where some inventory units occupy storage space for extended periods with low consumption probability, while high-consumption-probability units fail to receive adequate allocation, increasing the risk of inventory backlog and near-expiration waste. Furthermore, group meal orders typically exhibit distinct type variations in actual operations, such as regular daily orders, ad-hoc orders, and bulk orders. These order types differ in ingredient types, quantities, and consumption patterns. Existing technological solutions generally lack effective differentiation between order types, resulting in simplistic inventory demand range settings that fail to adapt to complex and ever-changing order structures. This can easily lead to localized stockouts or inventory redundancy, and the lack of proactive identification and inventory adjustment for high-expiration-risk ingredients makes it difficult to promptly mitigate inventory waste caused by near-expiration. Therefore, there is an urgent need for a dynamic management method for group catering ingredient inventory that can comprehensively consider order cycle characteristics, order type differences, and the life cycle characteristics of ingredient inventory, so as to achieve refined characterization of inventory status, effective filtering of abnormal demand, and dynamic adaptive optimization of inventory strategy, thereby improving the overall efficiency and reliability of group catering ingredient inventory management. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention provides a method and system for dynamic management of food inventory.

[0004] The first aspect of this invention provides a method for dynamic management of food inventory, mainly including:

[0005] By using the group meal ingredient delivery system, we can obtain group meal order cycle data and inventory batch data, divide the inventory batch data into life cycle stages, and obtain a basic correlation dataset between the order cycle and the inventory life cycle of multiple ingredients.

[0006] By breaking down and statistically analyzing the food consumption behavior at each stage of the life cycle under different order cycle conditions, a set of characteristics affecting the order cycle is constructed, and the life cycle impact weight data of each food category is determined.

[0007] By extracting the demand distribution feature set of multiple ingredients from the historical order record table, and using the probability distribution fitting method, we obtain the consumption probability distribution function of each ingredient under different order cycles. Combined with the life cycle influence weight data, we obtain the consumption probability dataset of multiple ingredients.

[0008] Based on the multi-ingredient consumption probability dataset, the consumption probability of each ingredient is mapped to the corresponding inventory unit, mismatched inventory units are identified, and an optimized inventory spatial distribution scheme is determined.

[0009] By extracting order behavior feature sets from historical order records, an order type identification model is constructed to identify order types and generate inventory level demand ranges for different order types across various food categories.

[0010] Based on historical data on ingredient demand for different order types, predict the ingredient demand for different order types within a preset time period in the future, generate multi-ingredient inventory adjustment instructions, and adjust the inventory of each ingredient category.

[0011] By executing inventory level adjustment instructions, the inventory level adjustment results are obtained, the consumption deviation values ​​of different ingredients are calculated, the nonlinear mapping model of the impact of order cycle on inventory life cycle is corrected, and the weight data of the impact of multi-ingredient life cycle is updated.

[0012] Furthermore, the process involves acquiring group meal order cycle data and inventory batch data through the group meal ingredient delivery system, dividing the inventory batch data into lifecycle stages, and obtaining a basic correlation dataset between the order cycle and the lifecycle of multiple ingredient inventories, including:

[0013] By using historical order records, order scheduling tables, and inventory management tables from the group meal ingredient delivery system, order cycles are labeled using a time window segmentation method to obtain three types of order time series data: daily, weekly, and monthly cycles. Order cycle label sets are obtained by aligning the order placement time, fulfillment time, and actual meal delivery time in the order data with timestamps. Using the inventory management table, for each ingredient's corresponding inventory batch data, the fields of batch entry time, shelf life in days, actual usage time, and remaining inventory are obtained. The inventory batch data is then divided into lifecycle stages, mapping each batch of each ingredient to the initial, middle, and near-expiration stages of its lifecycle, resulting in a set of inventory lifecycle state sequences divided by ingredient dimension. By matching the order cycle label set with the inventory lifecycle state sequences of each ingredient using time windows, a correspondence table between order cycle, ingredient, and inventory lifecycle stages is constructed. Consistency checks and outlier removal are performed on the correspondence table to obtain a basic association dataset between order cycles and multi-ingredient inventory lifecycles.

[0014] Furthermore, by breaking down and statistically analyzing food consumption behavior at each stage of the lifecycle under different order cycle conditions, a set of order cycle impact features is constructed, and the lifecycle impact weight data for each food category is determined, including:

[0015] Based on the basic association dataset of order cycles and multi-ingredient inventory lifecycles, for each ingredient, the ingredient consumption data under different order cycle conditions and in each lifecycle stage is broken down and processed to obtain the order trigger frequency, unit time consumption change rate, and inventory consumption acceleration corresponding to each order cycle in different lifecycle stages, constructing an order cycle impact feature set divided by ingredient category; based on the changes in order trigger frequency and inventory consumption rate of each ingredient category in the same order cycle, high-interference order cycles are identified, and the ingredient combination demand data of the order cycle are removed from the order cycle impact feature set; based on the order cycle impact feature set, a nonlinear mapping model of the impact of order cycles on inventory lifecycle is established for each ingredient category using a combination of nonlinear regression and piecewise fitting, determining the corresponding lifecycle impact weight data; by comparing the output of the nonlinear mapping model with the historical consumption trajectory of the corresponding ingredient, the model prediction error distribution is calculated. If the prediction error exceeds the preset error threshold, the piecewise boundary parameters and feature weight parameters of the nonlinear mapping model are adjusted, and the fitting calculation is re-executed until the model prediction error stabilizes within the preset range.

[0016] This also includes identifying order cycles with high interference based on changes in order trigger frequency and inventory consumption rate for each category of ingredients within the same order cycle, and removing the ingredient combination demand data for those order cycles from the order cycle impact feature set. Specifically, this includes:

[0017] Based on orders for each category of ingredients within the same order cycle, calculate the rate of change in order trigger frequency and the rate of change in inventory consumption acceleration, and construct the corresponding ingredient consumption rhythm function. Where i represents a specific food category involved in inventory consumption calculation, p represents the order cycle index, used to identify the order cycle divided according to a preset time window, and p+1 represents the next adjacent order cycle. This represents the number of orders that trigger an order and cause food ingredient i to consume inventory within the order cycle p. This represents the acceleration of inventory consumption of ingredient i within the order cycle p; based on the ingredient consumption rhythm function, the ingredient consumption interaction index formula is used. Calculate the interaction index of multiple food ingredients consumption within the same order cycle p. Where L represents the set of ingredients. This is used to characterize the degree of mutual interference between different food consumption rhythms; if the interaction index exceeds the preset interference level threshold, the current order cycle is determined to be in a high-interference state order cycle, and the food combination demand data of that order cycle is marked as a non-life cycle driven sample and removed from the order cycle influence feature set.

[0018] Furthermore, by extracting a set of multi-ingredient demand distribution features from the historical order record table, and using a probability distribution fitting method, the consumption probability distribution function of each ingredient under different order cycles is obtained. Combined with lifecycle impact weight data, a multi-ingredient consumption probability dataset is obtained, including:

[0019] By using historical order records, we obtain ingredient usage data and order cycle identifiers by ingredient category, forming a continuous consumption data sequence set for each ingredient under different order cycle conditions. Through windowed statistical processing of the consumption data sequence set, we calculate the mean demand, fluctuation range, extreme value distribution, and stability index of each ingredient within different order cycles, forming a multi-ingredient demand distribution feature set. Using a probability distribution fitting method, we model the multi-ingredient demand distribution feature set to obtain the consumption probability distribution function for each ingredient under different order cycle conditions. By weighting the consumption probability distribution function with the corresponding ingredient's lifecycle impact weight data, we generate a comprehensive consumption probability matrix for each ingredient under different combinations of order cycles and lifecycle stages. We then normalize and perform boundary checks on the matrix data to obtain a standardized multi-ingredient consumption probability dataset.

[0020] Furthermore, the step of mapping the consumption probability of each ingredient to a corresponding inventory unit based on a multi-ingredient consumption probability dataset, identifying mismatched inventory units, and determining an optimized inventory spatial distribution scheme includes:

[0021] Based on the multi-ingredient consumption probability dataset, and combined with the warehouse unit structure, storage area division, and batch distribution of various ingredients within each inventory unit recorded in the inventory management table, the consumption probability of each ingredient is mapped to the corresponding inventory unit. By matching the life cycle stage of different ingredient batches within the inventory unit with the corresponding consumption probability, the deviation between the current inventory distribution and the multi-ingredient consumption probability distribution under the order cycle is calculated, and inventory units with mismatched distributions are identified based on the magnitude of the deviation. By recalculating the batch allocation for mismatched inventory units, and under the premise of satisfying the life cycle constraints of each ingredient, the quantity ratio and life cycle stage combination relationship of multi-ingredient batches within different inventory units are adjusted, and the consistency of the adjustment results is verified to obtain the optimized inventory space distribution scheme.

[0022] Furthermore, the step involves extracting a set of order behavior features from a historical order record table, constructing an order type identification model, identifying order types, and generating inventory level demand ranges for different order types across various food categories, including:

[0023] Historical order data is obtained from the historical order record table. Stability parameters for order occurrence time intervals, volatility parameters for order usage, order fulfillment deviation and anomaly ratio parameters, and order change frequency and timing parameters are calculated to construct an order behavior feature set, which is then stored in the food delivery order monitoring database. Historical order data includes order placement time, fulfillment time, actual meal delivery time, order usage, order change records, and order fulfillment results. The historical order behavior feature set is obtained from the food delivery order monitoring database, and order types are labeled. A decision tree algorithm is used to train a model to construct an order type identification model. Order types include regular orders, temporary orders, and special customized orders. Based on the historical order behavior feature set, the order type identification model is used to identify the order type and generate corresponding inventory level demand ranges for different order types across different food categories. These inventory level demand ranges include the required proportions of available inventory, reserved inventory, and buffer inventory.

[0024] Furthermore, the step of predicting the demand for ingredients for different order types within a preset time period based on historical demand data for different order types, generating multi-ingredient inventory adjustment instructions, and adjusting the inventory levels of each ingredient category includes:

[0025] By utilizing historical order records, historical data on food demand for different order types is obtained. A Long Short-Term Memory (LSTM) network is used to train models, constructing prediction models for food demand for different order types. These models predict the types and quantities of food needed for different order types within a preset future timeframe. Based on the types and quantities of food needed for different order types within the preset future timeframe, and the corresponding inventory level demand ranges, the difference between the target food inventory and the current food inventory is calculated for each batch of food. Multiple food inventory adjustment instructions are generated based on these differences to adjust the inventory level demand for each food type. Food inventory is continuously monitored through an inventory management table. Based on the food inventory and the predicted food consumption, a food spoilage risk assessment formula is used. Determine the failure risk indicators for each category of food at time point t. , This represents the total inventory of ingredients at time point t. This represents the projected food consumption at time point t, predicted using historical demand data. This represents the number of days remaining from the current time t until the food expires. This represents the life cycle stage coefficient of the food ingredient, indicating its current life cycle stage. If it is nearing its expiration date... The value is 1, 0.5 in the intermediate stage, and 0 in the initial stage. The shelf-life adjustment factor for food ingredients was obtained by fitting historical data; food categories with expiration risk indicators exceeding a preset threshold were marked as high-risk ingredients, and an inventory adjustment formula was used. , Determine the required inventory level for high-risk ingredients at time point t. Based on the required inventory levels and corresponding inventory level demand ranges for high-risk ingredients, adjust the required inventory levels for high-risk ingredients.

[0026] Furthermore, the process of executing inventory level adjustment instructions, obtaining inventory level adjustment results, calculating consumption deviation values ​​for different ingredients, correcting the nonlinear mapping model of the impact of order cycle on inventory lifecycle, and updating the weight data of the impact of multi-ingredient lifecycles includes:

[0027] By executing inventory level adjustment instructions and adjusting the inventory level demand for high-risk ingredients, the available inventory, reserved inventory, and buffer inventory parameters for the corresponding batches in the inventory management table are updated. Based on the inventory level adjustment results, the remaining shelf life, expected usage time window, and lifecycle stage identifier for each ingredient batch are recalculated to generate updated inventory lifecycle status data. By aligning the updated inventory lifecycle status data with actual order fulfillment records, the consumption deviation value between the actual consumption location and the nonlinear mapping model predicting the consumption location of the order cycle on the inventory lifecycle is calculated for different ingredients. By writing the consumption deviation value back into the nonlinear mapping model of the order cycle on the inventory lifecycle, the model parameters are corrected and the weight data of the lifecycle impact of multiple ingredients is updated.

[0028] A second aspect of the present invention provides a dynamic food inventory management system, mainly comprising:

[0029] The food ingredient lifecycle mapping module is used to obtain group meal order cycle data and inventory batch data through the group meal ingredient delivery system, divide the inventory batch data into lifecycle stages, and obtain a basic association dataset between the order cycle and the inventory lifecycle of multiple ingredients.

[0030] The food lifecycle analysis module is used to break down and statistically analyze food consumption behavior at each stage of the lifecycle under different order cycle conditions, construct a set of order cycle impact characteristics, and determine the lifecycle impact weight data for each food category.

[0031] The food consumption analysis module is used to extract a set of multi-food demand distribution features from historical order records, use a probability distribution fitting method to obtain the consumption probability distribution function of each food in different order cycles, and combine the life cycle impact weight data to obtain a multi-food consumption probability dataset.

[0032] The inventory space distribution optimization module is used to map the consumption probability of each ingredient to the corresponding inventory unit based on the multi-ingredient consumption probability dataset, identify inventory units with mismatched distribution, and determine the optimized inventory space distribution scheme.

[0033] The order type identification module is used to extract a set of order behavior features from the historical order record table, build an order type identification model, identify the order type, and generate the inventory level demand range for different order types in each food category;

[0034] The food ingredient inventory adjustment module is used to predict the food ingredient demand of different order types within a preset time period based on historical data of food ingredient demand for different order types, generate multiple food ingredient inventory adjustment instructions, and adjust the inventory of each food ingredient category.

[0035] The execution result feedback optimization module is used to obtain the inventory level adjustment result by executing the inventory level adjustment instruction, calculate the consumption deviation value of different ingredients, correct the nonlinear mapping model of the impact of order cycle on inventory life cycle, and update the weight data of the impact of multi-ingredient life cycle.

[0036] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0037] This invention provides a method and system for dynamic management of food ingredient inventory. Based on historical order and inventory batch data, this invention accurately depicts the lifecycle of multiple food ingredient inventories. By combining order cycle characteristics and food ingredient consumption behavior, it can effectively distinguish between normal demand evolution and abnormal demand data under high-interference conditions, avoiding the misleading impact of short-term abnormal orders on inventory strategies, thereby improving the stability and reliability of inventory decisions. This invention combines order behavior characteristics and order type identification results to perform differentiated prediction of food ingredient demand and configure inventory level ranges for different order types. This enables inventory management strategies to adapt to various business scenarios such as regular orders, temporary orders, and special customized orders, enhancing the responsiveness of inventory configuration to changes in order structure. This invention continuously monitors inventory levels and calculates food ingredient expiration risk by combining predicted consumption, remaining validity period, and lifecycle stage. It dynamically adjusts the inventory level demand for high-risk food ingredients, thereby effectively reducing the risk of near-expiration and inventory backlog. By executing inventory level adjustment instructions and introducing an execution result feedback mechanism, the impact weight of the inventory lifecycle is continuously corrected, achieving dynamic adaptive updates of the inventory management strategy. This invention can improve the matching degree between inventory configuration and actual group meal demand while ensuring the constraints of the food life cycle, reduce the risk of inventory backlog and stockout, and enhance the stability, flexibility and long-term operational reliability of group meal food inventory management. Attached Figure Description

[0038] Figure 1 This is a flowchart of a method for dynamic management of food inventory according to the present invention;

[0039] Figure 2 The present invention provides a process for a dynamic food inventory management method.

[0040] Figure 3 This is a schematic diagram of a dynamic food inventory management system according to the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] like Figure 1-2 This embodiment of a method for dynamic management of food inventory may specifically include:

[0043] Step S101: Obtain group meal order cycle data and inventory batch data through the group meal ingredient delivery system, divide the inventory batch data into life cycle stages, and obtain the basic correlation dataset between order cycle and multi-ingredient inventory life cycle.

[0044] By utilizing historical order records, order scheduling tables, and inventory management tables from the group meal ingredient delivery system, order cycles are labeled using a time window segmentation method to obtain three types of order time series data: daily, weekly, and monthly cycles. By aligning the order placement time, fulfillment time, and actual meal delivery time in the order data with timestamps, a set of order cycle labels is obtained. Using the inventory management table, for each type of ingredient's corresponding inventory batch data, the fields of batch entry time, shelf life in days, actual usage time, and remaining inventory are obtained. The inventory batch data is then divided into lifecycle stages, mapping each batch of each ingredient to the initial stage, middle stage, and near-expiration stage of its lifecycle, resulting in a set of inventory lifecycle state sequences segmented by ingredient. By matching the order cycle label set with the inventory lifecycle state sequences of each ingredient using time windows, a correspondence table between order cycle, ingredient, and inventory lifecycle stages is constructed. Consistency checks and outlier removal are performed on this correspondence table to obtain a basic association dataset between order cycles and multi-ingredient inventory lifecycles.

[0045] For example, in the group meal ingredient delivery system, historical order records, order schedules, and inventory management tables from March 1, 2024 to March 31, 2024 are selected as analysis samples. The order cycles are divided into three categories using a time window segmentation method: daily cycles, weekly cycles, and monthly cycles. Daily cycles are based on calendar days, weekly cycles are based on Monday to Sunday, and monthly cycles are based on calendar months. For example, March 4, 2024, a Monday, falls within the daily order cycle of March 4, the weekly order cycle of the 10th week of 2024, and the monthly order cycle of March 2024. In the order data, the order placement time, fulfillment time, and actual meal delivery time for orders within the specified date are timestamped. For example, if an order was placed at 18:30 on March 3, 2024, fulfilled at 05:00 on March 4, 2024, and actually delivered at 11:30 on March 4, 2024, then this order is uniformly tagged in the order cycle tag set corresponding to March 4, 2024. In the inventory management table, taking chicken breast and potatoes as examples, the entry time, shelf life in days, actual usage time, and remaining inventory information for each batch of inventory are recorded. If there are two batches of chicken breast, with the first batch entering the warehouse on February 25, 2024, having a shelf life of 10 days and a planned usage period from March 2 to March 6, and the second batch entering the warehouse on March 3, 2024, also having a shelf life of 10 days and a planned usage period from March 5 to March 10, then... Based on the warehousing time and shelf life, the first batch of chicken breast was divided into three stages: the initial stage (February 25th to February 28th), the mid-stage (March 1st to March 4th), and the near-expiration stage (March 5th to March 6th). Similarly, the potato inventory was divided into the same lifecycle stages, forming an inventory lifecycle status sequence organized by ingredient. Then, the order cycle tag set was matched with the inventory lifecycle status sequence of each ingredient according to time windows. For example, in the order cycle of March 4th, 2024, it was recorded that 120 kg of chicken breast and 200 kg of potatoes were consumed that day. Time matching determined that the chicken breast consumed that day mainly came from the first batch that arrived on February 25th, and this batch was in the mid-lifecycle stage on March 4th. The potatoes consumed came from the inventory batch that arrived on February 20th, and this batch had entered the near-expiration stage on March 4th. This results in corresponding records such as "March 4, 2024 - Chicken breast - Mid-life cycle" and "March 4, 2024 - Potato - Near expiration stage" in the order cycle - ingredient - inventory life cycle stage correspondence table.After constructing the corresponding relationship table, consistency checks are performed to check for conflicts in the life cycle stages of ingredients, abnormal time spans, or negative inventory balances within the same order cycle. Abnormal records are then removed, ultimately yielding the basic association dataset between the order cycle and the life cycle of multiple ingredients' inventory for subsequent modeling and analysis.

[0046] Step S102 involves breaking down and statistically analyzing the food consumption behavior at each stage of the life cycle under different order cycle conditions, constructing a set of features affecting the order cycle, and determining the life cycle impact weight data for each food category.

[0047] Based on the fundamental correlation dataset between order cycles and multi-ingredient inventory lifecycles, this study breaks down the consumption data of each ingredient under different order cycle conditions and within each lifecycle stage for each ingredient. It obtains the order trigger frequency, unit time consumption change rate, and inventory consumption acceleration for each order cycle at different lifecycle stages, constructing a set of order cycle impact features categorized by ingredient type. Based on the changes in order trigger frequency and inventory consumption rate for each ingredient category within the same order cycle, high-interference order cycles are identified, and their ingredient combination demand data are removed from the order cycle impact feature set. Using the order cycle impact feature set, a nonlinear mapping model of the impact of order cycles on inventory lifecycles is established for each ingredient category using a combination of nonlinear regression and piecewise fitting, determining the corresponding lifecycle impact weights. By comparing the output of the nonlinear mapping model with the historical consumption trajectory of the corresponding ingredient, the model prediction error distribution is calculated. If the prediction error exceeds a preset error threshold, the piecewise boundary parameters and feature weight parameters of the nonlinear mapping model are adjusted, and the fitting calculation is re-executed until the model prediction error stabilizes within a preset range.

[0048] For example, based on the established dataset linking order cycles and multi-ingredient inventory lifecycles, we analyze the consumption behavior of chicken breast under different order cycle conditions. Historical data from March 2024 is selected and broken down into daily and weekly order cycles. Under the daily order cycle, the consumption data of chicken breast is divided according to the inventory lifecycle stage: the initial stage corresponds to days 1-3 after warehousing, the mid-stage corresponds to days 4-7 after warehousing, and the near-expiration stage corresponds to days 8-10 after warehousing. Statistical results show that in the initial stage of the product lifecycle, the average daily order trigger frequency for chicken breast is 4 times, and the daily consumption per unit time increases from 80 kg to 95 kg, indicating a gradual increase in the rate of change in consumption. In the middle stage of the product lifecycle, the average daily order trigger frequency increases to 6 times, and the daily consumption increases from 95 kg to 130 kg, with the rate of increase in consumption further expanding, indicating a positive acceleration in inventory consumption. However, in the near-expiration stage, although the order trigger frequency remains around 6 times, the daily consumption drops rapidly from 130 kg to 90 kg, and the rate of consumption slows significantly, indicating a negative acceleration in consumption. Based on the above statistical results, a set of order cycle impact characteristics was constructed, with chicken breast as the dimension, including changes in order trigger frequency, rate of change in consumption per unit time, and acceleration in inventory consumption under different order cycle conditions. Based on this, a joint analysis of the consumption characteristics of multiple ingredients within the same order cycle was conducted to identify order cycles with high interference. For example, in the 12th week of March 2024, the order trigger frequency for chicken breast increased by 40% compared to the previous week, while the order trigger frequencies for potatoes and broccoli decreased by 35% and 30% respectively during the same period. Furthermore, the inventory consumption rates of the three ingredients changed in opposite directions, exhibiting significant abnormal fluctuations. Because the consumption rhythms of different ingredients within this order cycle showed a significant divergence, it was determined that this cycle was strongly interfered with by temporary event orders or sudden group meal demands. Therefore, the ingredient combination demand data for this week was marked as a high-interference state and removed from the order cycle impact feature set to avoid biasing the lifecycle model. After removing high-interference cycles, a nonlinear mapping model of the impact of order cycle on inventory lifecycle was established for chicken breast based on the retained order cycle impact feature set. The model characterizes the degree of impact of order cycle changes on inventory consumption in different lifecycle stages in a segmented manner. For example, in the initial stage of the lifecycle, the impact weight of order cycle changes on inventory consumption is low, while in the middle stage, the impact weight increases significantly, and in the near-expiration stage, it shows a rapid decay trend. Comparing the model output with the historical actual consumption trajectory of chicken breast, it was found that the model predicted an average daily consumption of 125 kg in some cycles, while the actual consumption was 140 kg, with a prediction error of 12%, exceeding the preset error threshold of 10%.To address this issue, the segmentation boundaries of the lifecycle stages in the model were adjusted, and the weighting coefficients of the order cycle characteristics in the mid-term stage were appropriately increased. The fitting calculation was then re-executed. After multiple rounds of adjustments, the model's prediction error stabilized and converged to within 6%, ultimately yielding a nonlinear mapping model that stably reflects the impact of order cycle changes on the chicken breast inventory lifecycle, along with corresponding lifecycle impact weight data.

[0049] Specifically, based on the changes in order trigger frequency and inventory consumption rate of each type of ingredient within the same order cycle, order cycles with high interference are identified, and the ingredient combination demand data of that order cycle is removed from the set of influence features of the order cycle.

[0050] Based on orders for each category of ingredients within the same order cycle, calculate the rate of change in order trigger frequency and the rate of change in inventory consumption acceleration, and construct the corresponding ingredient consumption rhythm function. Where i represents a specific food category involved in inventory consumption calculation, p represents the order cycle index, used to identify the order cycle divided according to a preset time window, and p+1 represents the next adjacent order cycle. This represents the number of orders that trigger an order and cause food ingredient i to consume inventory within the order cycle p. This represents the acceleration of inventory consumption of ingredient i within an order cycle p. Based on the ingredient consumption rhythm function, the ingredient consumption interaction index formula is used. Calculate the interaction index of multiple food ingredients consumption within the same order cycle p. Where L represents the set of ingredients. This is used to characterize the degree of mutual interference between different food consumption rhythms. If the interaction index exceeds a preset interference threshold, the current order cycle is determined to be a high-interference order cycle, and the food combination demand data for that order cycle is marked as a non-lifecycle-driven sample and removed from the order cycle influence feature set.

[0051] For example, in multi-ingredient inventory management, even if the consumption behavior of each ingredient meets stability and predictability constraints within its own order cycle and lifecycle stage, significant mutual interference effects may still occur between the consumption rhythms of multiple ingredients, causing abrupt changes in the overall inventory consumption behavior. These abrupt changes cannot be identified using a single ingredient model or consumption probability distribution. In the group meal delivery system, weeks 10 and 11 of 2024 are selected as two adjacent order cycles, with week 10 corresponding to order cycle index p and week 11 corresponding to order cycle index p+1. For orders of each category of ingredients within the same order cycle, a corresponding ingredient consumption rhythm function is constructed based on the change rate of order trigger frequency and the change in inventory consumption acceleration. Here, i represents a specific food category involved in inventory consumption calculation. During the order cycles of weeks 10 and 11 in 2024, inventory consumption was recorded for chicken breast, potatoes, and broccoli. Therefore, these three foods collectively constitute the set of foods consumed within the same order cycle. In week 10, chicken breast was triggered by 10 orders, while this number increased to 15 in week 11, indicating a significant increase in the order trigger frequency for chicken breast between adjacent cycles. Simultaneously, based on the inventory consumption records, the inventory consumption acceleration for chicken breast was 2.0 in week 10, indicating an accelerating consumption rate, which increased to 3.0 in week 11, indicating a further strengthening of this accelerating trend. Based on the above data, the consumption rhythm function for chicken breast within order cycle p is positive, reflecting that its order trigger rhythm and inventory consumption inertia are mutually reinforcing within adjacent cycles. During the same period, potatoes saw 20 order triggers in week 10, which decreased to 14 in week 11, showing a significant drop in order trigger frequency. Their inventory consumption acceleration was 1.5 in week 10 and decreased to 0.5 in week 11, indicating that potato consumption not only slowed down but the slowdown trend intensified. The calculated food consumption rhythm function for potatoes within order period p is also positive, but this positive value stems from the simultaneous decrease in order trigger frequency and weakening consumption inertia, reflecting a trend of the consumption rhythm converging towards a low-activity state. Meanwhile, broccoli saw 12 order triggers in week 10, which increased to 18 in week 11, showing a significant increase in order trigger frequency. However, its inventory consumption acceleration decreased from 2.5 in week 10 to 1.0 in week 11, indicating that although order triggers were more frequent, the actual consumption intensity of a single order began to weaken. Therefore, the food consumption rhythm function for broccoli within order period p is negative, reflecting a significant divergence between order rhythm and inventory consumption inertia. After obtaining the consumption rhythm function of the above three ingredients within the same order period p, the ingredient consumption interaction index formula is used. Calculate the interaction index of multiple food ingredients consumption within the same order cycle p. Where L represents the set of ingredients. This is used to characterize the degree of mutual interference between the consumption rhythms of different ingredients. Since the consumption rhythm functions of chicken breast and potatoes are positive, while those of broccoli are negative, and the numerical differences are significant, the sum of the absolute values ​​of the rhythms of each ingredient is significantly greater than the absolute value of the algebraic sum of the rhythms. Consequently, the calculated interaction index is significantly higher than the system's preset interference threshold. Based on this result, order cycle p is determined to be in a high-interference state. It is believed that the consumption behavior of multiple ingredients within this cycle is not driven by normal life cycle evolution, but rather by temporary changes in order structure or mutual interference between the consumption rhythms of ingredients. Therefore, the ingredient combination demand data within this order cycle is marked as a non-life cycle driven sample and removed from the dataset subsequently used to construct the order cycle influence feature set to avoid this abnormal cycle interfering with inventory life cycle modeling and parameter fitting results.

[0052] Step S103: Extract the multi-ingredient demand distribution feature set from the historical order record table, use the probability distribution fitting method to obtain the consumption probability distribution function of each ingredient under different order cycles, and combine the life cycle influence weight data to obtain the multi-ingredient consumption probability dataset.

[0053] By analyzing historical order records, we obtain ingredient usage data and order cycle identifiers by ingredient category, forming a continuous consumption data sequence set for each ingredient under different order cycle conditions. Through windowed statistical processing of the consumption data sequence set, we calculate the mean demand, fluctuation range, extreme value distribution, and stability index of each ingredient within different order cycles, forming a multi-ingredient demand distribution feature set. Using a probability distribution fitting method, we model the multi-ingredient demand distribution feature set, obtaining the consumption probability distribution function for each ingredient under different order cycle conditions. By weighting the consumption probability distribution function with the corresponding ingredient's lifecycle impact weight data, we generate a comprehensive consumption probability matrix for each ingredient under different combinations of order cycles and lifecycle stages. The matrix data is then normalized and boundary-checked to obtain a standardized multi-ingredient consumption probability dataset.

[0054] For example, in a group meal delivery system, historical order records from March 2024 are selected. Ingredient usage data is extracted from these records by category, and the corresponding order cycle is marked using order scheduling information. The order cycle is divided into calendar days and weeks. Taking chicken breast as an example, in week 13 of 2024, the daily consumption of chicken breast from Monday to Friday was recorded as 120 kg, 135 kg, 128 kg, 142 kg, and 130 kg, respectively. This forms a continuous consumption data sequence for chicken breast under the order cycle conditions of that week. Similarly, continuous consumption data sequences for other ingredients such as potatoes and broccoli under different order cycles are formed. After obtaining the continuous consumption data sequences, windowed statistical processing is performed on these sequences to characterize the demand distribution characteristics of the ingredients within the corresponding order cycles. Taking chicken breast in week 13 as an example, statistical calculations of daily consumption data during that week reveal an average demand of 131 kg, a maximum consumption of 142 kg, a minimum consumption of 120 kg, and a fluctuation range of 22 kg. Furthermore, based on the deviation of daily consumption from the average, the demand stability index for chicken breast during this period is calculated to be moderately high, indicating that demand is relatively concentrated but still subject to some fluctuation. Using the same method, the average demand, fluctuation range, extreme value distribution, and stability index within the corresponding order periods for other ingredients are statistically analyzed, ultimately forming a multi-ingredient demand distribution feature set. Based on this, a probability distribution fitting method is used to model the multi-ingredient demand distribution feature set. For chicken breast consumption data in week 13, since its daily consumption exhibits a relatively symmetrical distribution around the average, it is fitted to an approximately normal distribution, thus obtaining the probability distribution function of chicken breast consumption under the conditions of this order period. For potatoes, due to the significantly higher consumption on certain days within this cycle, their consumption data was fitted with a skewed distribution to more accurately reflect actual demand characteristics. Using the above method, the consumption probability distribution functions for each ingredient under different order cycle conditions were obtained. Subsequently, the consumption probability distribution functions were weighted with the lifecycle impact weights of each ingredient to comprehensively consider the combined influence of order cycle demand characteristics and inventory lifecycle stage on consumption behavior. For example, when chicken breast is in the middle stage of its inventory lifecycle, its lifecycle impact weight is set to 1.2, indicating that its consumption priority should be appropriately increased in this stage. This weight is applied to the consumption probability distribution function of chicken breast in the 13th week order cycle, thus obtaining the comprehensive consumption probability value of chicken breast under the combined condition of the 13th week order cycle—middle lifecycle stage. By performing the same calculation process on all ingredients under different order cycle and lifecycle stage combinations, a comprehensive consumption probability matrix is ​​generated.Finally, the matrix is ​​normalized to ensure that the consumption probability of different ingredients and under different cycle conditions is within a uniform numerical range, and the probability values ​​that exceed the reasonable range are checked and corrected to obtain a standardized multi-ingredient consumption probability dataset.

[0055] Step S104: Based on the multi-ingredient consumption probability dataset, map the consumption probability of each ingredient to the corresponding inventory unit, identify inventory units with mismatched distributions, and determine the optimized inventory space distribution scheme.

[0056] Based on a multi-ingredient consumption probability dataset, and combined with the warehouse unit structure, storage area division, and batch distribution of various ingredients within each inventory unit recorded in the inventory management table, the consumption probability of each ingredient is mapped to its corresponding inventory unit. By matching the lifecycle stages of different ingredient batches within an inventory unit with their corresponding consumption probabilities, the deviation between the current inventory distribution and the multi-ingredient consumption probability distribution under the order cycle is calculated, and inventory units with mismatched distributions are identified based on the magnitude of the deviation. Batch reallocation calculations are then performed on these mismatched inventory units. Under the premise of satisfying the lifecycle constraints of each ingredient, the quantity ratio and lifecycle stage combination relationship of multi-ingredient batches within different inventory units are adjusted, and the consistency of the adjustment results is verified to obtain an optimized inventory space distribution scheme.

[0057] For example, in a group meal ingredient distribution center, the warehouse is divided into three storage units: refrigerated area A, refrigerated area B, and ambient temperature area C. Refrigerated areas A and B are mainly used to store cold chain ingredients such as chicken breast and beef, while ambient temperature area C is used to store ambient temperature ingredients such as potatoes and onions. Based on the multi-ingredient consumption probability dataset obtained from the aforementioned modeling process, within the current order cycle, the overall consumption probability of chicken breast is 0.45, beef is 0.30, potatoes are 0.20, and onions are 0.05, indicating that chicken breast is most likely to be consumed first within this order cycle, while onions have the lowest consumption probability. The inventory management table shows that at the current moment, cold storage area A contains 300 kg of chicken breast and 100 kg of beef. Of the chicken breast, 200 kg is in the middle of its life cycle, and 100 kg is nearing its expiration date. Cold storage area B contains 80 kg of chicken breast and 220 kg of beef, with most batches still in the early stages of their life cycle. Ambient temperature area C contains 400 kg of potatoes and 120 kg of onions, with half of the potatoes already in the late stages of their life cycle. First, the consumption probability of each ingredient is mapped to its corresponding storage unit. For example, the consumption probability of chicken breast and beef is mapped to cold storage areas A and B, and the consumption probability of potatoes and onions is mapped to ambient temperature area C. Then, the inventory structure within each storage unit is analyzed, matching the life cycle stage of each ingredient batch within the storage unit with its corresponding consumption probability. Analysis revealed that a significant proportion of the high-consumption-probability chicken breasts in refrigerated zone A were still in the middle of their life cycle, while refrigerated zone B contained a large amount of chicken breasts with a high consumption probability but still in the early stages of their life cycle. Meanwhile, in ambient temperature zone C, onions with a low consumption probability occupied a large portion of the storage near the ex-warehouse location, while some potatoes that had entered the late stages of their life cycle and had a high consumption probability were stored deeper in the storage. Based on these matching results, the system calculated a significant deviation between the inventory distribution in refrigerated zone B and ambient temperature zone C and the multi-ingredient consumption probability distribution under the current order cycle, while the deviation in refrigerated zone A was smaller. Therefore, refrigerated zone B and ambient temperature zone C were identified as mismatched inventory units. For these mismatched inventory units, batch reallocation calculations were performed while satisfying the life cycle constraints of each ingredient. For example, batches of chicken breasts in the early stages of their life cycle in refrigerated zone B were moved to refrigerated zone A, and batches of chicken breasts in refrigerated zone A that were nearing expiration were moved to storage areas closer to the ex-warehouse location, thereby improving the out-of-warehouse efficiency of high-consumption-probability ingredients. Meanwhile, within the ambient temperature zone C, the batch distribution of potatoes and onions was adjusted. Potato batches that had entered the middle and late stages of their life cycle and had a high probability of being consumed were moved to a location that was easier to release from storage, while onion batches with a lower probability of being consumed were moved to a later storage location.After the batch reallocation is completed, the consistency of the adjustment results is checked to confirm that each batch of ingredients has not exceeded its life cycle constraints and that the storage unit capacity and storage rules have not been violated. Finally, an optimized inventory space distribution scheme that matches the probability distribution of multi-ingredient consumption under the current order cycle is obtained.

[0058] Step S105: Extract the set of order behavior features from the historical order record table, construct an order type identification model, identify the order type, and generate the inventory level demand range for different order types in each food category.

[0059] Historical order data is obtained from the historical order record table. Stability parameters for order occurrence time intervals, volatility parameters for order usage, order fulfillment deviation and anomaly ratio parameters, and order change frequency and timing parameters are calculated to construct an order behavior feature set, which is then stored in the food delivery order monitoring database. Historical order data includes order placement time, fulfillment time, actual meal delivery time, order usage, order change records, and order fulfillment results. Using the food delivery order monitoring database, the historical order behavior feature set is obtained and order types are labeled. A decision tree algorithm is used to train a model to construct an order type identification model. Order types include recurring orders, temporary orders, and special customized orders. Based on the historical order behavior feature set, the order type identification model is used to identify order types and generate corresponding inventory level demand ranges for different food categories for different order types. These inventory level demand ranges include the required proportions of available inventory, reserved inventory, and buffer inventory.

[0060] For example, historical order records from January to March 2024 were selected from the group meal ingredient delivery system as the analysis object. This historical order data includes order placement time, fulfillment time, actual meal service time, order quantity, order change records, and order fulfillment results. Statistical analysis of the historical order data revealed that a certain customer placed 60 orders during this period. Most orders were placed around 9:00 AM on Mondays, Wednesdays, and Fridays. The time interval between adjacent orders was consistently around 48 hours, with a standard deviation of only 3 hours. Therefore, the stability parameter of the customer's order time interval is high, reflecting a clear periodicity. Simultaneously, the amount of chicken breast used in each order was concentrated between 95 kg and 105 kg, with an average of 100 kg and a usage fluctuation rate of approximately 5%, indicating relatively stable order quantity. Regarding fulfillment, the records show that 58 orders were completed on schedule, with only 2 orders experiencing delays exceeding 30 minutes, resulting in an abnormal fulfillment rate of approximately 3.3%. Furthermore, only one of the 60 orders involved a usage adjustment, and this change occurred within 5 minutes of order placement. There were no multiple changes or changes occurring close to the meal delivery time, indicating a low number of order changes and a relatively early timing of these changes. Based on these analysis results, an order behavior feature set for this customer was constructed and stored in the food delivery order monitoring database. After constructing the order behavior feature set, historical order behavior feature data from multiple customers and time periods was further extracted from the food delivery order monitoring database. Order types were then labeled according to manual annotation or historical rules, including recurring orders, temporary orders, and special customized orders. For example, recurring orders are characterized by highly stable order time intervals, low usage fluctuations, a low proportion of fulfillment anomalies, and few order changes. Temporary orders typically exhibit irregular order placement times, large usage fluctuations, and occasional fulfillment deviations. Special customized orders often involve multiple order changes and high fulfillment complexity. Based on the labeled historical order behavior feature set, a decision tree algorithm was used for model training to construct an order type recognition model, enabling the model to automatically distinguish different order types based on order behavior features. After the model training is completed, the order type recognition model is used to identify other historical order behavior feature sets. Based on the order type recognition results, corresponding inventory level demand ranges are further generated for different order types and different food categories. For example, for chicken breast in regular orders, the system sets the available inventory to cover the demand for the next two order cycles, accounting for 70% to 80% of the total inventory, with a reserved inventory of 15% to 20% to cope with small fluctuations in usage, and a buffer inventory controlled at 5% to 10% to prevent inventory backlog. For temporary orders, the available inventory ratio is appropriately reduced and the buffer inventory ratio is increased to enhance inventory flexibility. For special customized orders, the reserved inventory ratio is increased to cope with frequent order change requirements.Using the above method, based on the identification of order types, inventory level demand ranges that conform to the order behavior characteristics are generated for different food categories.

[0061] Step S106: Based on historical data of ingredient demand for different order types, predict the ingredient demand for different order types within a preset time period in the future, generate multi-ingredient inventory adjustment instructions, and adjust the inventory of each ingredient category.

[0062] By utilizing historical order records, historical data on food demand for different order types is obtained. A Long Short-Term Memory (LSTM) network is used to train models, constructing prediction models for food demand across different order types. These models predict the types and quantities of food needed for different order types within a preset future timeframe. Based on the types and quantities of food needed for different order types within the preset future timeframe, and the corresponding inventory level demand ranges, the difference between the target and current food inventory levels is calculated for each batch of food. Multiple food inventory adjustment instructions are generated based on these differences to adjust the inventory level demand for each food type. Food inventory is continuously monitored through an inventory management table. Based on the current inventory and predicted consumption, a food spoilage risk assessment formula is used. Determine the failure risk indicators for each category of food at time point t. , This represents the total inventory of ingredients at time point t. This represents the projected food consumption at time point t, predicted using historical demand data. This represents the number of days remaining from the current time t until the food expires. This represents the life cycle stage coefficient of the food ingredient, indicating its current life cycle stage. If it is nearing its expiration date... The value is 1, 0.5 in the intermediate stage, and 0 in the initial stage. The shelf-life adjustment factor for food ingredients is obtained by fitting historical data. Food categories with a failure risk index exceeding a preset threshold are marked as high-risk ingredients, and an inventory adjustment formula is used accordingly. , Determine the required inventory level for high-risk ingredients at time point t. Based on the required inventory levels and corresponding inventory level demand ranges for high-risk ingredients, the required inventory levels for high-risk ingredients will be adjusted.

[0063] For example, based on historical order records, orders are categorized into regular orders, temporary orders, and special customized orders. Historical data on the types of ingredients and their demand quantities for each order type are extracted within a specific historical period. By training on this historical data, long short-term memory (LSTM) network prediction models are constructed for different order types to predict the possible ingredient categories and corresponding demand distributions for each order type within a preset timeframe over the next month. The prediction results show that, within a preset timeframe, such as the next week, regular orders primarily involve chicken breast, potatoes, and broccoli, with a predicted demand of 500 kg for chicken breast, 420 kg for potatoes, and 300 kg for broccoli. Temporary orders primarily involve chicken breast and beef, with a predicted demand of 150 kg for chicken breast and 100 kg for beef. Special customized orders primarily involve chicken breast and broccoli, with a predicted demand of 100 kg for chicken breast and 60 kg for broccoli. Therefore, the predicted total demand for chicken breast, potatoes, broccoli, and beef is 750 kg, 420 kg, 360 kg, and 100 kg respectively within the preset time period. Based on these predictions and the established inventory management strategy, along with the corresponding inventory level demand ranges for different food categories, the inventory level demand range for chicken breast is set at 750 to 820 kg, including the combined demand for available inventory, reserved inventory, and buffer inventory. The inventory level demand range for potatoes is 450 to 500 kg, broccoli is 380 to 430 kg, and beef is 120 to 150 kg. Subsequently, the current inventory status data is read from the inventory management table, revealing that the current available inventory in the warehouse is 680 kg for chicken breast, 460 kg for potatoes, 410 kg for broccoli, and 160 kg for beef. The system calculates the difference between the target inventory level and the current inventory level. The current chicken breast inventory is approximately 70 kg below the lower limit of its inventory level's demand range; potato inventory is within a reasonable range; broccoli inventory is near the lower limit of its range, indicating some replenishment demand; and beef inventory is approximately 10 kg above the upper limit of its inventory level's demand range. Based on these difference analysis results, multi-ingredient inventory adjustment instructions are generated, including instructions to replenish chicken breast to increase inventory by approximately 70 kg, to slightly replenish broccoli to maintain inventory stability, and to suspend beef replenishment and prioritize the consumption of existing inventory. In this way, inventory adjustment instructions can directly reflect the difference between the predicted demand for ingredients for different future order types and the current inventory status, ensuring that inventory configuration is consistent with upcoming order demand. Ingredient inventory levels are continuously monitored through the inventory management table. To assess the stability of this inventory structure under future demand, an ingredient spoilage risk assessment formula is used. The calculation is performed, where the total inventory of chicken breasts at time point t is... The estimated consumption is 750 kg, obtained by forecasting historical demand data. The weight is 730 kg. The remaining number of days from the current time t until the food item expires is... Based on the food lifecycle assessment results, this batch of chicken breast is nearing its expiration date, and its lifecycle stage coefficient is [not specified]. The value is 1, while the shelf-life adjustment coefficient is obtained by fitting historical inventory and loss data. The value is 0.9. Substituting this into the food spoilage risk assessment formula, the spoilage risk index for chicken breast at time point t is 0.86. This value is higher than the preset spoilage risk threshold of 0.8. Therefore, chicken breast is marked as a high-risk food ingredient, indicating that under the current inventory configuration and predicted consumption conditions, chicken breast is at risk of entering a near-expiration state in subsequent cycles. For chicken breast marked as a high-risk food ingredient, the inventory adjustment amount at time point t is further calculated according to the inventory adjustment rules. For example, by reducing replenishment and accelerating the outbound pace, the target inventory of chicken breast is reduced from the original 800 kg to 760 kg, thereby reducing the spoilage risk caused by inventory backlog. Based on this, the system dynamically adjusts the inventory level demand range of chicken breast to meet future order demand while reducing the inventory risk in the near-expiration stage.

[0064] Step S107: By executing the inventory level adjustment instruction, obtain the inventory level adjustment result, calculate the consumption deviation value of different ingredients, correct the nonlinear mapping model of the impact of order cycle on inventory life cycle, and update the weight data of the impact of multi-ingredient life cycle.

[0065] By executing inventory level adjustment instructions and adjusting the inventory level demand for high-risk ingredients, the available inventory, reserved inventory, and buffer inventory parameters for the corresponding batches in the inventory management table are updated. Based on the inventory level adjustment results, the remaining shelf life, expected usage time window, and lifecycle stage identifier for each ingredient batch are recalculated, generating updated inventory lifecycle status data. By aligning the updated inventory lifecycle status data with actual order fulfillment records, the consumption deviation value between the actual consumption location and the nonlinear mapping model of the order cycle's impact on the inventory lifecycle is calculated for different ingredients. By writing the consumption deviation value back into the nonlinear mapping model of the order cycle's impact on the inventory lifecycle, the model parameters are corrected, and the weight data of the multi-ingredient lifecycle impact is updated.

[0066] For example, based on the aforementioned inventory adjustment instructions and the adjustment of inventory level requirements for high-risk ingredients, after completing the inventory operation, the parameters in the inventory management table for the chicken breast batch are updated synchronously. Specifically, the available inventory of this batch of chicken breast is adjusted from 680 kg to 720 kg, the reserved inventory from 50 kg to 30 kg, and the buffer inventory from 70 kg to 40 kg, keeping the total inventory at 760 kg to reduce the risk of spoilage. Based on the updated inventory parameters, the inventory lifecycle status of this batch of chicken breast is recalculated. Combining its production date and shelf-life information, the remaining shelf life of this batch is determined to be 5 days. Based on historical order fulfillment patterns, the expected usage window is predicted to be within the next 4 days for major consumption; therefore, its lifecycle stage is updated to the near-expiration stage. After generating updated inventory lifecycle status data, this data was aligned with actual order fulfillment records. The analysis revealed that the actual consumption of this batch of chicken breast reached 600 kg on day 3 of the order cycle. However, the nonlinear mapping model based on the impact of the order cycle on the inventory lifecycle predicted a theoretical consumption of 550 kg at the same time point. This resulted in a 50 kg consumption deviation between the actual and predicted consumption locations, reflecting that the actual order fulfillment pace was faster than the model anticipated. This deviation was then written back into the corresponding nonlinear mapping model to correct the parameters describing the impact of order cycle changes on the inventory lifecycle. This improved the model's sensitivity to the accelerated consumption characteristics of ingredients nearing expiration in subsequent predictions. Through this correction process, new multi-ingredient lifecycle impact weight data was obtained. For example, the lifecycle impact weight of chicken breast in the latter half of the order cycle was increased from 0.7 to 0.82, making subsequent inventory lifecycle predictions more closely reflect actual order fulfillment behavior.

[0067] like Figure 3 This embodiment of a dynamic food inventory management system may specifically include:

[0068] The food ingredient lifecycle mapping module is used to obtain group meal order cycle data and inventory batch data through the group meal ingredient delivery system, divide the inventory batch data into lifecycle stages, and obtain a basic association dataset between the order cycle and the inventory lifecycle of multiple ingredients.

[0069] The food lifecycle analysis module is used to break down and statistically analyze food consumption behavior at each stage of the lifecycle under different order cycle conditions, construct a set of order cycle impact characteristics, and determine the lifecycle impact weight data for each food category.

[0070] The food consumption analysis module is used to extract a set of demand distribution features for multiple food ingredients from historical order records. It then uses a probability distribution fitting method to obtain the consumption probability distribution function of each food ingredient under different order cycles. Combined with the life cycle impact weight data, it obtains a multi-food consumption probability dataset.

[0071] The inventory space distribution optimization module is used to map the consumption probability of each ingredient to the corresponding inventory unit based on the multi-ingredient consumption probability dataset, identify inventory units with mismatched distributions, and determine the optimized inventory space distribution scheme.

[0072] The order type identification module is used to extract a set of order behavior features from the historical order record table, build an order type identification model, identify the order type, and generate the inventory level demand range for different order types in each food category.

[0073] The food ingredient inventory adjustment module is used to predict the food ingredient demand for different order types within a preset time period based on historical data of food ingredient demand for different order types, generate multiple food ingredient inventory adjustment instructions, and adjust the inventory of each food ingredient category.

[0074] The execution result feedback optimization module is used to obtain the inventory level adjustment result by executing the inventory level adjustment instruction, calculate the consumption deviation value of different ingredients, correct the nonlinear mapping model of the impact of order cycle on inventory life cycle, and update the weight data of the impact of multi-ingredient life cycle.

[0075] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the concept of this application. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for dynamic management of food inventory, characterized in that, The method includes: By using the group meal ingredient delivery system, we can obtain group meal order cycle data and inventory batch data, divide the inventory batch data into life cycle stages, and obtain a basic correlation dataset between the order cycle and the inventory life cycle of multiple ingredients. By breaking down and statistically analyzing the food consumption behavior at each stage of the life cycle under different order cycle conditions, a set of characteristics affecting the order cycle is constructed, and the life cycle impact weight data of each food category is determined. By extracting the demand distribution feature set of multiple ingredients from the historical order record table, and using the probability distribution fitting method, we obtain the consumption probability distribution function of each ingredient under different order cycles. Combined with the life cycle influence weight data, we obtain the consumption probability dataset of multiple ingredients. Based on the multi-ingredient consumption probability dataset, the consumption probability of each ingredient is mapped to the corresponding inventory unit, mismatched inventory units are identified, and an optimized inventory spatial distribution scheme is determined. By extracting order behavior feature sets from historical order records, an order type identification model is constructed to identify order types and generate inventory level demand ranges for different order types across various food categories. Based on historical data on ingredient demand for different order types, predict the ingredient demand for different order types within a preset time period in the future, generate multi-ingredient inventory adjustment instructions, and adjust the inventory of each ingredient category. By executing inventory level adjustment instructions, the inventory level adjustment results are obtained, the consumption deviation values ​​of different ingredients are calculated, the nonlinear mapping model of the impact of order cycle on inventory life cycle is corrected, and the weight data of the impact of multi-ingredient life cycle is updated.

2. The method according to claim 1, wherein, The system obtains group meal order cycle data and inventory batch data through the group meal ingredient delivery system, divides the inventory batch data into lifecycle stages, and obtains a basic correlation dataset between the order cycle and the lifecycle of multiple ingredients' inventory, including: By using historical order records, order scheduling tables, and inventory management tables from the group meal ingredient delivery system, order cycles are labeled using a time window segmentation method to obtain three types of order time series data: daily, weekly, and monthly cycles. Order cycle label sets are obtained by aligning the order placement time, fulfillment time, and actual meal delivery time in the order data with timestamps. Using the inventory management table, for each ingredient's corresponding inventory batch data, the fields of batch entry time, shelf life in days, actual usage time, and remaining inventory are obtained. The inventory batch data is then divided into lifecycle stages, mapping each batch of each ingredient to the initial, middle, and near-expiration stages of its lifecycle, resulting in a set of inventory lifecycle state sequences divided by ingredient dimension. By matching the order cycle label set with the inventory lifecycle state sequences of each ingredient using time windows, a correspondence table between order cycle, ingredient, and inventory lifecycle stages is constructed. Consistency checks and outlier removal are performed on the correspondence table to obtain a basic association dataset between order cycles and multi-ingredient inventory lifecycles.

3. The method according to claim 1, wherein, The method involves breaking down and statistically analyzing food consumption behavior at each stage of the food lifecycle under different order cycle conditions, constructing a set of order cycle impact features, and determining the lifecycle impact weight data for each food category, including: Based on the basic association dataset of order cycle and multi-ingredient inventory lifecycle, the consumption data of each ingredient under different order cycle conditions and in each lifecycle stage is broken down for each ingredient. The order trigger frequency, unit time consumption change rate and inventory consumption acceleration corresponding to each type of order cycle in different lifecycle stages are obtained, and a set of order cycle impact features divided by ingredient category is constructed. Based on the changes in order trigger frequency and inventory consumption rate of each type of ingredient in the same order cycle, order cycles with high interference are identified, and the ingredient combination demand data of the order cycle are removed from the set of order cycle impact features. Based on the set of features affecting order cycles, a combination of nonlinear regression and piecewise fitting is used to establish a nonlinear mapping model for the impact of order cycles on inventory lifecycle for each type of food ingredient, and to determine the corresponding lifecycle impact weight data. By comparing the output of the nonlinear mapping model with the historical consumption trajectory of the corresponding food ingredient, the model prediction error distribution is calculated. If the prediction error exceeds the preset error threshold, the piecewise boundary parameters and feature weight parameters of the nonlinear mapping model are adjusted, and the fitting calculation is re-executed until the model prediction error stabilizes within the preset range.

4. The method according to claim 3, wherein, The process involves identifying high-interference order cycles based on changes in order trigger frequency and inventory consumption rate for each category of ingredients within the same order cycle, and removing ingredient combination demand data from the order cycle impact feature set for those order cycles. This includes: Based on orders for each category of ingredients within the same order cycle, calculate the rate of change in order trigger frequency and the rate of change in inventory consumption acceleration, and construct the corresponding ingredient consumption rhythm function. Where i represents a specific food category involved in inventory consumption calculation, p represents the order cycle index, used to identify the order cycle divided according to a preset time window, and p+1 represents the next adjacent order cycle. This represents the number of orders that trigger an order and cause food ingredient i to consume inventory within the order cycle p. This represents the acceleration of inventory consumption of ingredient i within the order cycle p; based on the ingredient consumption rhythm function, the ingredient consumption interaction index formula is used. Calculate the interaction index of multiple food ingredients consumption within the same order cycle p. Where L represents the set of ingredients. This is used to characterize the degree of mutual interference between different food consumption rhythms; if the interaction index exceeds the preset interference level threshold, the current order cycle is determined to be in a high-interference state order cycle, and the food combination demand data of that order cycle is marked as a non-lifecycle driven sample and removed from the order cycle influence feature set.

5. The method according to claim 1, wherein, The process involves extracting a set of demand distribution features for multiple ingredients from a historical order record table, employing a probability distribution fitting method to obtain the consumption probability distribution function of each ingredient under different order cycles, and combining this with lifecycle impact weight data to obtain a multi-ingredient consumption probability dataset, including: By using historical order records, we obtain ingredient usage data and order cycle identifiers by ingredient category, forming a continuous consumption data sequence set for each ingredient under different order cycle conditions. Through windowed statistical processing of the consumption data sequence set, we calculate the mean demand, fluctuation range, extreme value distribution, and stability index of each ingredient within different order cycles, forming a multi-ingredient demand distribution feature set. Using a probability distribution fitting method, we model the multi-ingredient demand distribution feature set to obtain the consumption probability distribution function for each ingredient under different order cycle conditions. By weighting the consumption probability distribution function with the corresponding ingredient's lifecycle impact weight data, we generate a comprehensive consumption probability matrix for each ingredient under different combinations of order cycles and lifecycle stages. We then normalize and perform boundary checks on the matrix data to obtain a standardized multi-ingredient consumption probability dataset.

6. The method according to claim 1, wherein, The step involves mapping the consumption probability of each ingredient to a corresponding inventory unit based on a multi-ingredient consumption probability dataset, identifying mismatched inventory units, and generating an optimized inventory spatial distribution scheme, including: Based on the multi-ingredient consumption probability dataset, and combined with the warehouse unit structure, storage area division, and batch distribution of various ingredients within each inventory unit recorded in the inventory management table, the consumption probability of each ingredient is mapped to the corresponding inventory unit. By matching the life cycle stage of different ingredient batches within the inventory unit with the corresponding consumption probability, the deviation between the current inventory distribution and the multi-ingredient consumption probability distribution under the order cycle is calculated, and inventory units with mismatched distributions are identified based on the magnitude of the deviation. By recalculating the batch allocation for mismatched inventory units, and under the premise of satisfying the life cycle constraints of each ingredient, the quantity ratio and life cycle stage combination relationship of multi-ingredient batches within different inventory units are adjusted, and the consistency of the adjustment results is verified to obtain the optimized inventory space distribution scheme.

7. The method according to claim 1, wherein, The process involves extracting a set of order behavior features from historical order records, constructing an order type identification model, identifying order types, and generating inventory level demand ranges for different order types across various food categories. This includes: Historical order data is obtained from the historical order record table. Stability parameters of order occurrence time interval, volatility parameters of order usage, order fulfillment deviation and abnormality ratio parameters, and order change frequency and change time sequence parameters are calculated respectively. A set of order behavior characteristics is constructed and stored in the food delivery order monitoring database. Historical order data includes order placement time, fulfillment time, actual meal delivery time, order usage, order change records, and order fulfillment results. By using the food delivery order monitoring database, we obtain a set of historical order behavior characteristics and label the order types. We then use a decision tree algorithm to train a model and build an order type identification model. The order types include regular orders, temporary orders, and special customized orders. Based on the set of historical order behavior characteristics, we use the order type identification model to identify the order type and generate corresponding inventory level demand ranges for different order types on different food categories. The inventory level demand range includes the required range of proportions of available inventory, reserved inventory, and buffer inventory.

8. The method according to claim 1, wherein, The process involves predicting the demand for ingredients for different order types within a preset time period based on historical data of ingredient demand for different order types, generating multi-ingredient inventory adjustment instructions, and adjusting the inventory levels for each ingredient category, including: By utilizing historical order records, historical data on food demand for different order types is obtained. A Long Short-Term Memory (LSTM) network is used to train models, constructing prediction models for food demand for different order types. These models predict the types and quantities of food needed for different order types within a preset future timeframe. Based on the types and quantities of food needed for different order types within the preset future timeframe, and the corresponding inventory level demand ranges, the difference between the target food inventory and the current food inventory is calculated for each batch of food. Multiple food inventory adjustment instructions are generated based on these differences to adjust the inventory level demand for each food type. Food inventory is continuously monitored through an inventory management table. Based on the food inventory and the predicted food consumption, a food spoilage risk assessment formula is used. Determine the failure risk indicators for each category of food at time point t. , This represents the total inventory of ingredients at time point t. This represents the projected food consumption at time point t, predicted using historical demand data. This represents the number of days remaining from the current time t until the food expires. This represents the life cycle stage coefficient of the food ingredient, indicating its current life cycle stage. If it is nearing its expiration date... The value is 1, 0.5 in the intermediate stage, and 0 in the initial stage. The shelf-life adjustment factor representing the food ingredients was obtained by fitting historical data. Food categories with failure risk indicators exceeding a preset threshold are marked as high-risk foods, and an inventory adjustment formula is used. , Determine the required inventory level for high-risk ingredients at time point t. Based on the required inventory levels and corresponding inventory level demand ranges for high-risk ingredients, adjust the required inventory levels for high-risk ingredients.

9. The method according to claim 1, wherein, The process involves executing inventory level adjustment instructions, obtaining inventory level adjustment results, calculating consumption deviation values ​​for different ingredients, correcting the nonlinear mapping model of the impact of order cycles on inventory lifecycles, and updating the weight data of the impact of multi-ingredient lifecycles, including: By executing inventory level adjustment instructions and adjusting the inventory level demand for high-risk ingredients, the available inventory, reserved inventory, and buffer inventory parameters for the corresponding batches in the inventory management table are updated. Based on the inventory level adjustment results, the remaining shelf life, expected usage time window, and lifecycle stage identifier for each ingredient batch are recalculated to generate updated inventory lifecycle status data. By aligning the updated inventory lifecycle status data with actual order fulfillment records, the consumption deviation value between the actual consumption location and the nonlinear mapping model predicting the consumption location of the order cycle on the inventory lifecycle is calculated for different ingredients. By writing the consumption deviation value back into the nonlinear mapping model of the order cycle on the inventory lifecycle, the model parameters are corrected and the weight data of the lifecycle impact of multiple ingredients is updated.

10. A dynamic food inventory management system, comprising executing the steps of a dynamic food inventory management method according to any one of claims 1-9, characterized in that, The system includes: The food ingredient lifecycle mapping module is used to obtain group meal order cycle data and inventory batch data through the group meal ingredient delivery system, divide the inventory batch data into lifecycle stages, and obtain a basic association dataset between the order cycle and the inventory lifecycle of multiple ingredients. The food lifecycle analysis module is used to break down and statistically analyze food consumption behavior at each stage of the lifecycle under different order cycle conditions, construct a set of order cycle impact characteristics, and determine the lifecycle impact weight data for each food category. The food consumption analysis module is used to extract a set of multi-food demand distribution features from historical order records, use a probability distribution fitting method to obtain the consumption probability distribution function of each food in different order cycles, and combine the life cycle impact weight data to obtain a multi-food consumption probability dataset. The inventory space distribution optimization module is used to map the consumption probability of each ingredient to the corresponding inventory unit based on the multi-ingredient consumption probability dataset, identify inventory units with mismatched distribution, and determine the optimized inventory space distribution scheme. The order type identification module is used to extract a set of order behavior features from the historical order record table, build an order type identification model, identify the order type, and generate the inventory level demand range for different order types in each food category; The food ingredient inventory adjustment module is used to predict the food ingredient demand of different order types within a preset time period based on historical data of food ingredient demand for different order types, generate multiple food ingredient inventory adjustment instructions, and adjust the inventory of each food ingredient category. The execution result feedback optimization module is used to obtain the inventory level adjustment result by executing the inventory level adjustment instruction, calculate the consumption deviation value of different ingredients, correct the nonlinear mapping model of the impact of order cycle on inventory life cycle, and update the weight data of the impact of multi-ingredient life cycle.

Images