Inventory optimization system and method based on multi-source data

Abstract

The invention discloses an inventory optimization system and method based on multi-source data, and belongs to the technical field of inventory optimization. The production data and the sales data are acquired and fused to obtain the multi-source data, the LSTM neural network is trained by using the multi-source data, and the future market demand is predicted; on the basis of future market demands, in the aspect of space resource consumption and the aspect of time resource consumption, target functions and constraint conditions are set respectively, the corresponding target functions are solved, and the minimum warehouse resource consumption is obtained through calculation; processing the historical sales data by using a moving average method, and predicting the future sales volume of the warehouse products; combining the future sales volume with the existing stock volume and the in-transit volume, and calculating the replenishment volume; obtaining logistics data, determining the replenishment time according to the logistics transportation time, and adjusting the replenishment amount according to the transportation capacity; and based on the replenishment time and the replenishment amount, using a dynamic planning algorithm to obtain an optimal distribution frequency, and matching the optimal distribution frequency with the inventory to obtain an inventory use plan.

Description

Technical Field

[0001] The present invention relates to the technical field of inventory optimization, and particularly to an inventory optimization system and method based on multi-source data. Background Art

[0002] Inventory, as a key link connecting production and sales, its management efficiency directly affects the operating costs and market competitiveness of enterprises; the rapid changes in market demand, the acceleration of product updates, and the increasing requirements of customers for service quality all prompt enterprises to seek more scientific and efficient inventory management methods.

[0003] Traditional forecasting methods, such as simple time series analysis or empirical forecasting, cannot effectively handle complex market data and dynamically changing market demands. In terms of warehouse resource consumption, the existing technologies lack comprehensive and systematic optimization methods. In terms of space resources, warehouse layouts often lack scientific planning, resulting in low space utilization; in terms of time resources, products stay in the warehouse for too long, increasing inventory costs and the risk of goods damage. In determining the replenishment quantity and replenishment time of existing inventory management, there is often a lack of scientific basis and it is easily affected by subjective factors. Summary of the Invention

[0004] The purpose of the present invention is to provide an inventory optimization system and method based on multi-source data to solve the problems raised in the existing technologies.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] In the first aspect, the present invention provides an inventory optimization method based on multi-source data, including the following steps:

[0007] Obtain production data and sales data, fuse them to obtain multi-source data, and use the multi-source data to train an LSTM neural network to predict future market demands;

[0008] Based on the future market demands, set objective functions and constraint conditions respectively in terms of space resource consumption and time resource consumption, solve the corresponding objective functions, and calculate the minimum warehouse resource consumption;

[0009] Apply the moving average method to process historical sales data, predict the future sales volume of warehouse products; combine the future sales volume with the existing inventory and in-transit quantity to calculate the replenishment quantity;

[0010] Obtain logistics data, determine the replenishment time according to the logistics transportation time, and adjust the replenishment quantity according to the transportation capacity; based on the replenishment time and replenishment quantity, use the dynamic programming algorithm to obtain the optimal delivery frequency, and match the optimal delivery frequency with the inventory to obtain the inventory usage plan.

[0011] In combination with the first aspect, in the first implementation manner of the first aspect of this application, the steps of obtaining production data and sales data, fusing them to obtain multi-source data, and using the multi-source data to train an LSTM neural network to predict future market demand include:

[0012] Obtain production data and sales data, use the product ID as the associated field to ensure the temporal alignment of production data and sales data, create a data table in a relational database, and fuse the production data and sales data to obtain multi-source data;

[0013] Set the input layer, hidden layer, and output layer of the LSTM neural network, determine the parameters of the LSTM neural network, arrange the multi-source data in chronological order, divide it into a training set, a validation set, and a test set according to a ratio, and construct a data sequence; perform forward propagation and backward propagation, and conduct the evaluation and adjustment of the LSTM neural network; the user inputs the production data and sales data of the product into the trained LSTM neural network to obtain the predicted future market demand, specifically the production capacity limit, supply chain constraints, and order fulfillment rate.

[0014] In combination with the first aspect, in the second implementation manner of the first aspect of this application, based on the future market demand, set the objective function and constraint conditions respectively in terms of spatial resource consumption and temporal resource consumption, solve the corresponding objective function, and calculate the minimum warehouse resource consumption, including:

[0015] The warehouse resource consumption includes spatial resource consumption and temporal resource consumption; in terms of spatial resource consumption, on the premise of meeting the future market demand, reduce the occupancy of warehouse space by products, and use an optimization layout algorithm to optimize the warehouse layout according to the occupied space and sales speed of products, reducing spatial resource consumption; in terms of temporal resource consumption, according to the future market demand and sales data, arrange the procurement time through a time optimization algorithm to determine the optimal procurement time point and reduce the in-warehouse storage time of products;

[0016] Based on the optimized warehouse layout, obtain the spatial resource consumption index by calculating the utilization rate of the total warehouse volume; convert the in-warehouse storage time of products corresponding to the optimal procurement time point into a time length index to obtain the temporal resource consumption index; based on the spatial resource consumption index and the temporal resource consumption index, obtain the minimum warehouse resource consumption through weighted summation.

[0017] In combination with the first aspect, in the third implementation manner of the first aspect of this application, the steps of using an optimization layout algorithm to optimize the warehouse layout according to the occupied space and sales speed of products and reducing spatial resource consumption include:

[0018] Obtain the occupied space and sales speed of the product. The optimization layout algorithm is used to select the genetic algorithm. The objective function of the genetic algorithm is to reduce the consumption of space resources. The constraint conditions of the genetic algorithm include product correlation constraints, storage condition constraints, and channel and handling equipment constraints. Among them, the product correlation constraint means that products with correlation are placed in adjacent positions to reduce the picking path and time. The storage condition constraint represents the storage condition requirements of the product. For products with the same storage condition requirements, they are arranged in the same storage area. The channel and handling equipment constraint represents the setting of the reserved channels in the warehouse and the operating space of the handling equipment;

[0019] In the genetic algorithm, determine the size of the initial population, set the maximum number of iterations, crossover probability, and mutation probability; divide the warehouse into several storage areas, each area corresponding to a gene position of the chromosome, and the value of the gene position represents the product number stored in that area; during the iteration process, generate a new layout plan through crossover and mutation operations, and decode the new plan to obtain the actual warehouse layout; according to the objective function and constraint conditions of the genetic algorithm, calculate the fitness value for each layout plan, and the fitness value reflects the pros and cons of the layout plan in terms of reducing the consumption of space resources; continuously generate a new population through selection, crossover, and mutation operations; evaluate the optimal layout plan generated by the genetic algorithm, make adjustments and optimizations, and implement the optimal layout plan.

[0020] Combined with the first aspect, in the fourth implementation manner of the first aspect of the present application, the procurement time is arranged through a time optimization algorithm according to future market demands and sales data to determine the best procurement time point and reduce the in - warehouse storage time of the product, including:

[0021] The time optimization algorithm selects the dynamic programming algorithm. According to the procurement time range and the time distribution of future market demands, the procurement time is divided into several stages; define the state of each stage, including the current inventory quantity, remaining production capacity, and remaining supply capacity of the supplier; determine the decision variable of each stage, specifically the procurement quantity; establish a state transition equation according to production, supply, and sales situations; with the goal of minimizing the in - warehouse storage time of the product, construct an objective function, and calculate the sum of the products of the inventory quantity at each stage and the in - warehouse storage time of the product; the constraint conditions of the dynamic programming algorithm include market demand constraints, production capacity constraints, supply capacity constraints, and non - negative inventory constraints;

[0022] Determine the initial state, including the initial inventory quantity, the initial production capacity of the enterprise, and the initial supply capacity of the supplier, and set the objective function value in the initial stage to 0; starting from the first stage, according to the state transition equation, the objective function, and the constraint conditions, calculate the next-stage state and the objective function value corresponding to different procurement quantities in the current state; select the procurement quantity that minimizes the objective function value as the optimal decision for the current stage, take the state of the next stage as the new current state, and enter the calculation of the next stage; when the calculations for all stages are completed, obtain the optimal procurement time points from the initial stage to the final stage through the backtracking process; starting from the optimal procurement time point of the final stage, gradually backtrack to the initial stage according to the state transition equation and the decision record to determine the best procurement time points and procurement quantities for each stage.

[0023] Combined with the first aspect, in the fifth implementation manner of the first aspect of the present application, the use of the moving average method to process historical sales data to predict the future sales volume of warehouse products includes:

[0024] Determine the moving average period and match it with the seasonal period of the product; the moving average period is n, and for the historical sales data from 1 to n months, calculate the moving average values in sequence; extrapolate the moving average values to predict the sales volume in the (n + 1)-th month; as time goes by, compare the predicted future sales volume with the actual sales data; calculate the prediction error using the mean absolute error to measure the deviation degree between the predicted value and the actual value, and when the deviation degree exceeds the set threshold, adjust the moving average period.

[0025] Combined with the first aspect, in the sixth implementation manner of the first aspect of the present application, the calculation of the replenishment quantity by combining the future sales volume with the existing inventory quantity and the quantity in transit includes:

[0026] The first calculation formula for the replenishment quantity is RQ = FS - CI - ITQ, where RQ is the replenishment quantity, FS is the future sales volume, CI is the existing inventory quantity, and ITQ is the quantity in transit; for products with historical sales data fluctuations exceeding the set threshold, introduce a safety inventory TQ and use the second calculation formula for the replenishment quantity, specifically: RQ = FS - CI - ITQ - TQ.

[0027] Combined with the first aspect, in the seventh implementation manner of the first aspect of the present application, the obtaining of logistics data, determining the replenishment time according to the logistics transportation time, and adjusting the replenishment quantity according to the transportation capacity includes:

[0028] Establish a data sharing interface with the logistics supplier to extract the average transportation time of various products from the supplier's shipping place to the warehouse; obtain the transportation capacity of different transportation tools from the logistics supplier, including the maximum single shipment volume and the transportation frequency per unit time;

[0029] Determine the replenishment time according to the logistics transportation time, establish a replenishment time model, and the formula is as follows: Among them, RT is the replenishment time, in days, ADFS is the future average daily sales volume, obtained by dividing the future sales volume FS by the number of days in the current month, LTT is the logistics transportation time, and SLT is the safety lead time, representing the additional time reserved by users; when the logistics transportation is delayed, the replenishment time is advanced accordingly according to the actual delay situation; when the sales speed increases or decreases, recalculate the time to sell out the existing inventory and the in-transit quantity, and then adjust the replenishment time;

[0030] Compare the calculated replenishment quantity with the transportation capacity of the logistics supplier. When the replenishment quantity exceeds the single maximum transportation volume of the transportation vehicle, split the replenishment quantity.

[0031] Combined with the first aspect, in the eighth implementation manner of the first aspect of the present application, based on the replenishment time and the replenishment quantity, use the dynamic programming algorithm to obtain the optimal delivery frequency, match the optimal delivery frequency with the inventory, and obtain an inventory usage plan, including:

[0032] Based on the replenishment time, divide the inventory management cycle into several stages, define the state variables of each stage, including the current inventory level, the remaining replenishment quantity, and the number of completed deliveries; determine the decision variable of each stage as the delivery frequency, specifically, decide whether to make a delivery and the number of deliveries in this stage; establish a state transition equation based on the change of inventory, replenishment operations, and the delivery process; construct an objective function with the goal of minimizing inventory backlog and out-of-stock risks; determine the initial state, specifically, the inventory, the remaining replenishment quantity, and the number of deliveries in the first stage; set the boundary conditions of each stage, specifically, the upper and lower limits of the delivery frequency, the upper limit is determined according to the transportation capacity, and the lower limit is determined according to the need to ensure the normal turnover of the inventory;

[0033] Starting from the first stage, according to the state transition equation, the objective function, and the boundary conditions, calculate the next-stage state and the objective function value corresponding to different delivery frequency decisions in the current state; select the delivery frequency that minimizes the objective function value as the optimal decision for the current stage, take the state of the next stage as the new current state, and enter the calculation of the next stage; when the calculations of all stages are completed, obtain the optimal delivery frequency from the initial stage to the final stage through the backtracking process;

[0034] Establish an association model between the optimal delivery frequency and the inventory, coordinate the inventory and the delivery time nodes, so as to obtain an inventory usage plan.

[0035] In the second aspect, the present invention provides an inventory optimization system based on multi-source data, including:

[0036] Data Acquisition and Prediction Module: It includes a data acquisition unit and a demand prediction unit. Among them, the data acquisition unit acquires production data and sales data, and fuses them to obtain multi-source data. The demand prediction unit uses the multi-source data to train an LSTM neural network to predict future market demand.

[0037] Objective Solving Module: It includes an initial setting unit and an optimization algorithm solving unit. Among them, the initial setting unit, based on future market demand, sets the objective function and constraints respectively in terms of spatial resource consumption and time resource consumption. The optimization algorithm solving unit solves the corresponding objective function to calculate the minimum warehouse resource consumption.

[0038] Replenishment Quantity Calculation Module: It includes a moving average processing unit and a replenishment quantity calculation unit. Among them, the moving average processing unit uses the moving average method to process historical sales data to predict the future sales volume of warehouse products. The replenishment quantity calculation unit combines the future sales volume with the existing inventory and in-transit quantity to calculate the replenishment quantity.

[0039] Inventory Planning Module: It includes a replenishment quantity adjustment unit, a dynamic programming solving unit, and an inventory planning unit. Among them, the replenishment quantity adjustment unit acquires logistics data, determines the replenishment time according to the logistics transportation time, and adjusts the replenishment quantity according to the transportation capacity. The dynamic programming solving unit, based on the replenishment time and replenishment quantity, uses the dynamic programming algorithm to obtain the optimal delivery frequency. The inventory planning unit matches the optimal delivery frequency with the inventory to obtain an inventory usage plan.

[0040] Compared with the prior art, the beneficial effects of the present invention are:

[0041] 1. Based on future market demand, the present invention sets the objective function and constraints respectively from two aspects of spatial resource consumption and time resource consumption, and uses an optimization algorithm to solve. It can realize the rational utilization of warehouse space resources and effectively shorten the in-warehouse storage time of products on the premise of meeting market demand.

[0042] 2. The present invention uses the moving average method to process historical sales data, accurately predicts future sales volume, combines the existing inventory and in-transit quantity, and scientifically calculates the replenishment quantity. It determines the replenishment time according to the logistics transportation time, adjusts the replenishment quantity according to the transportation capacity, and uses the dynamic programming algorithm to obtain the optimal delivery frequency. It can closely coordinate the replenishment and distribution links, and improve the efficiency and flexibility of inventory management.

[0043] 3. The present invention matches the optimal delivery frequency with the inventory to formulate a reasonable inventory usage plan. By optimizing the inventory layout, formulating a scientific inventory usage strategy, and real-time monitoring the inventory level, it can effectively improve the inventory turnover rate. Description of the Drawings

[0044] Figure 1It is a schematic diagram of the steps of an inventory optimization method based on multi-source data according to the present invention;

[0045] Figure 2 It is a system structure diagram of an inventory optimization system based on multi-source data according to the present invention. Detailed implementation manners

[0046] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0047] Embodiment: As Figure 1 - Figure 2 shown, the present invention provides a technical solution,

[0048] As Figure 1 shown in the schematic diagram of the steps of an inventory optimization method based on multi-source data according to the present invention, the present invention provides an inventory optimization method based on multi-source data, including the following steps:

[0049] Step S100: Obtain production data and sales data, fuse them to obtain multi-source data, and use the multi-source data to train an LSTM neural network to predict future market demand;

[0050] Specifically, obtain production data and sales data, use the product ID as the associated field to ensure that the production data and sales data are aligned in time, create a data table in a relational database, and fuse the production data and sales data to obtain multi-source data;

[0051] Set the input layer, hidden layer, and output layer of the LSTM neural network, determine the parameters of the LSTM neural network, arrange the multi-source data in chronological order, divide it into a training set, a validation set, and a test set according to a ratio, and construct a data sequence; perform forward propagation and backward propagation, and perform evaluation and adjustment of the LSTM neural network; the user inputs the production data and sales data of the product into the trained LSTM neural network to obtain the predicted future market demand, specifically the production capacity limit, supply chain constraints, and order fulfillment rate.

[0052] In a specific embodiment, in the LSTM neural network, the number of input layer nodes is determined according to the dimensions of production data and sales data. Here, the production data includes three dimensions: production quantity, production batch, and production time. The sales data includes three dimensions: sales quantity, sales channels, and customer types. Therefore, the number of input layer nodes is set to 6. Two hidden layers are set, and the number of neurons in each hidden layer is set to 50. The ReLU function is selected as the activation function for the hidden layers. The number of output layer nodes is set to 3, corresponding to production capacity limit, supply chain constraint, and order fulfillment rate respectively. The linear function is selected as the activation function for the output layer. The learning rate is set to 0.001, the number of iterations is set to 1000 times, and the mean squared error (MSE) is selected as the loss function. Arrange the 12 multi-source data in chronological order and divide them into a training set, a validation set, and a test set according to the ratio of 70%, 15%, and 15%. That is, the training set contains 8 data, and the validation set and the test set each contain 2 data. Combine the production data and sales data in each data into a data sequence as the input of the LSTM neural network. Through forward propagation, input the data in the training set into the LSTM neural network, and after the calculation of the hidden layer, output the prediction result. Compare the prediction result with the actual production capacity limit, supply chain constraint, and order fulfillment rate, and calculate the mean squared error as the loss value. Then, through backpropagation, adjust the parameters of the neural network according to the loss value to continuously optimize the model. During the training process, use the validation set to evaluate the model and observe the change of the loss value of the model on the validation set. When the loss value on the validation set no longer decreases, it is considered that the model has converged and the training is stopped.

[0053] Input the 2 data in the test set into the trained LSTM neural network. For example, the production data corresponding to one of the test data is: production quantity of 12,000 units, 3 production batches, and production times from March 1st to March 10th, from March 15th to March 20th, and from March 25th to March 31st respectively; the sales data is: sales quantity of 10,000 units, 6,000 units sold through the online e-commerce platform, 4,500 units purchased by individual consumers, 1,500 units purchased by enterprise customers, 4,000 units sold through offline stores, 3,500 units purchased by individual consumers, and 500 units purchased by enterprise customers. After the calculation of the LSTM neural network, the predicted future market demand result is obtained. The predicted production capacity limit is 13,000 units (indicating the maximum quantity that the enterprise can theoretically produce in the next stage), the supply chain constraint is that the raw material supply may be delayed by 1 week (affecting the production progress), and the order fulfillment rate is 85% (indicating the proportion of orders that are expected to be fulfilled under the current production and sales situation). By comparing with the actual situation, it is found that the prediction result is relatively close to the actual situation, verifying the effectiveness of the LSTM neural network in predicting the future market demand.

[0054] Step S200: Based on future market demands, set the objective function and constraints in terms of space resource consumption and time resource consumption respectively, solve the corresponding objective function, and calculate the minimum warehouse resource consumption.

[0055] Specifically, the warehouse resource consumption includes space resource consumption and time resource consumption. In terms of space resource consumption, on the premise of meeting future market demands, reduce the occupancy of the warehouse space by products, and use the optimized layout algorithm to optimize the warehouse layout according to the occupied space and sales speed of products, thereby reducing the space resource consumption. In terms of time resource consumption, according to future market demands and sales data, arrange the procurement time through the time optimization algorithm, determine the optimal procurement time point, and reduce the storage time of products in the warehouse.

[0056] Based on the optimized warehouse layout, obtain the space resource consumption index by calculating the utilization rate of the total volume of the warehouse; convert the storage time of products corresponding to the optimal procurement time point into a time length index to obtain the time resource consumption index; based on the space resource consumption index and the time resource consumption index, obtain the minimum warehouse resource consumption through weighted summation.

[0057] Furthermore, obtain the occupied space and sales speed of products. The optimized layout algorithm selects the genetic algorithm. The objective function of the genetic algorithm is to reduce the space resource consumption. The constraints of the genetic algorithm include product correlation constraints, storage condition constraints, and channel and handling equipment constraints. Among them, the product correlation constraint means that products with correlation are placed in adjacent positions to reduce the picking path and time. The storage condition constraint means the storage condition requirements of products. Products with the same storage condition requirements are arranged in the same storage area. The channel and handling equipment constraint means the setting of reserved channels in the warehouse and the operating space of handling equipment.

[0058] In the genetic algorithm, determine the size of the initial population, set the maximum number of iterations, crossover probability, and mutation probability; divide the warehouse into several storage areas, each area corresponding to a gene position of the chromosome, and the value of the gene position represents the product number stored in that area; during the iteration process, generate a new layout plan through crossover and mutation operations, and decode the new plan to obtain the actual warehouse layout; according to the objective function and constraints of the genetic algorithm, calculate the fitness value for each layout plan, and the fitness value reflects the pros and cons of the layout plan in terms of reducing the space resource consumption; continuously generate a new population through selection, crossover, and mutation operations; evaluate the optimal layout plan generated by the genetic algorithm, make adjustments and optimizations, and implement the optimal layout plan.

[0059] Further, the time optimization algorithm selects the dynamic programming algorithm. According to the procurement time range and the time distribution of future market demand, the procurement time is divided into several stages; the states of each stage are defined, including the current inventory quantity, the remaining production capacity, and the remaining supply capacity of the supplier; the decision variables of each stage are determined, specifically the procurement quantity; according to the production, supply, and sales situations, a state transition equation is established; with the goal of minimizing the in - stock storage time of the product, an objective function is constructed to calculate the sum of the products of the inventory quantity at each stage and the in - stock storage time of the product; the constraint conditions of the dynamic programming algorithm include market demand constraints, production capacity constraints, supply capacity constraints, and non - negative inventory constraints;

[0060] Determine the initial state, including the initial inventory quantity, the enterprise's initial production capacity, and the supplier's initial supply capacity, and set the objective function value of the initial stage to 0; starting from the first stage, according to the state transition equation, the objective function, and the constraint conditions, calculate the next - stage states and objective function values corresponding to different procurement quantities in the current state; select the procurement quantity that minimizes the objective function value as the optimal decision for the current stage, take the next - stage state as the new current state, and enter the calculation of the next stage; when the calculations for all stages are completed, obtain the optimal procurement time points from the initial stage to the final stage through the backtracking process; starting from the optimal procurement time point of the final stage, according to the state transition equation and the decision record, gradually backtrack to the initial stage to determine the best procurement time points and procurement quantities for each stage.

[0061] In a specific embodiment, genetic algorithm settings are made. Initial population size: Set to 50 layout plans, that is, 50 chromosomes. Maximum number of iterations: Set to 200 times. Crossover probability: Set to 0.8. Mutation probability: Set to 0.05. Product correlation constraint: It is known that model A and its matching mobile phone case are correlated and need to be placed adjacent to each other. Storage condition constraint: Models A, B, and C are all ordinary electronic products with the same storage conditions and can be arranged in the same storage area. Channel and handling equipment constraint: A 2 - meter width needs to be reserved for the warehouse channels, and at least 10 square meters of operating space for handling equipment is required for each storage area.

[0062] Each storage area corresponds to a gene locus on the chromosome, and the value of the gene locus represents the product number stored in that area. For example, the chromosome [1, 2, 0, 1, 2, 3, 0, 3, 2, 1] indicates that products of model A are stored in areas 1, 4, and 10, products of model B are stored in areas 2, 5, and 9, areas 3 and 7 are empty, and products of model C are stored in areas 6 and 8. According to each layout plan, calculate the utilization rate of the total volume of the warehouse. For example, in a certain layout plan, model A occupies 3 areas, with a total occupied space of 3×250×0.05 = 37.5 cubic meters; model B occupies 3 areas, with an occupied space of 3×250×0.06 = 45 cubic meters; model C occupies 2 areas, with an occupied space of 2×250×0.08 = 40 cubic meters, and the total occupied space is 37.5 + 45 + 40 = 122.5 cubic meters. Then the utilization rate of the total volume of the warehouse in this plan is 122.5÷2500 = 4.9%. The fitness value is related to the utilization rate of the total volume of the warehouse. The higher the utilization rate, the greater the fitness value (for ease of calculation, here the fitness value = the utilization rate of the total volume of the warehouse × 100), and the fitness value of this plan is 4.9.

[0063] Through the roulette wheel selection method, select the chromosomes with high fitness to enter the next generation. For example, if the total fitness value is 200 and the fitness value of a certain chromosome is 5, then the probability of its being selected is 5÷200 = 2.5%. The selected chromosomes are subjected to crossover and mutation operations to generate new layout plans. After 200 iterations, the optimal layout plan is obtained. The optimal layout plan is [1, 1, 2, 2, 3, 3, 1, 2, 3, 0]. At this time, the utilization rate of the total volume of the warehouse is 6%, which is significantly improved compared with the initial plan. After implementing this plan, the consumption of space resources is reduced, and the picking path is also optimized due to the product correlation constraints.

[0064] Set up the dynamic programming algorithm, and divide the next 6 months into 6 stages. The state of each stage includes the current inventory quantity (I), the remaining production capacity (P), and the remaining supply capacity of the supplier (S). For example, the initial state of the first stage is (500, 1000, 800). The decision variable is the procurement quantity (Q) for each stage.

[0065] Market demand constraint: The inventory quantity plus the production quantity and the procurement quantity in each stage should meet the market demand, that is, I + P + Q ≥ market demand. Production capacity constraint: The production quantity cannot exceed the remaining production capacity of the enterprise, that is, the production quantity ≤ P. The procurement quantity cannot exceed the remaining supply capacity of the supplier, that is, Q ≤ S. The inventory quantity cannot be negative, that is, I ≥ 0.

[0066] The inventory quantity I' in the next stage = I + P + Q - market demand; the remaining production capacity P' in the next stage = P - production quantity; the remaining supply capacity S' of the supplier in the next stage = S - Q. For example, in the first stage, if the purchase quantity Q = 300 units, the production quantity is 800 units (not exceeding the production capacity), and the market demand is 1200 units, then the inventory quantity I' in the next stage = 500 + 800 + 300 - 1200 = 400 units, the remaining production capacity P' = 1000 - 800 = 200 units, and the remaining supply capacity S' of the supplier = 800 - 300 = 500 units.

[0067] Aiming to minimize the in - warehouse storage time of products, the in - warehouse storage time of products is proportional to the inventory quantity. For example, if the inventory quantity in the first stage is 500 units and the in - warehouse storage time is 1 month, then the contribution of the objective function value in this stage is 500×1 = 500. Calculate the objective function value for each stage, and select the purchase quantity that minimizes the objective function value as the optimal decision for the current stage. After completing the calculations for 6 stages, obtain the optimal purchase time points and purchase quantities from the initial stage to the final stage through backtracking. The backtracking result is to purchase 300 units, 200 units, and 300 units respectively in the 1st, 3rd, and 5th months, at which time the total in - warehouse storage time of the products is the smallest.

[0068] Based on the optimized warehouse layout, the overall volume utilization rate of the warehouse is 6%, and the space resource consumption index is set to 1 - 6% = 94% (the smaller the index value, the lower the space resource consumption). The total in - warehouse storage time of the products obtained through the dynamic programming algorithm is 1500 months (each mobile phone stored in the warehouse for 1 month is recorded as 1 month), and the time resource consumption index is set to 1500÷(6×1200 + 500)≈0.2 (the total demand is the expected demand for 6 months plus the initial inventory, and the smaller the index value, the lower the time resource consumption). The weight of the space resource consumption index is 0.6, and the weight of the time resource consumption index is 0.4. Then the minimum warehouse resource consumption = 0.6×94% + 0.4×0.2 = 0.644. By continuously adjusting the weights and optimizing the algorithm parameters, the minimum warehouse resource consumption can be further reduced.

[0069] Step S300: Use the moving average method to process the historical sales data and predict the future sales volume of the warehouse products; combine the future sales volume with the existing inventory and the in - transit quantity to calculate the replenishment quantity.

[0070] Specifically, determine the moving average period and match it with the seasonal period of the product; the moving average period is n. For the historical sales data from 1 to n months, calculate the moving average values in sequence; extrapolate the moving average values to predict the sales volume for the (n + 1)-th month; as time goes by, compare the predicted future sales volume with the actual sales data; calculate the prediction error using the mean absolute error to measure the deviation degree between the predicted value and the actual value. When the deviation degree exceeds the set threshold, adjust the moving average period.

[0071] Furthermore, the first calculation formula for the replenishment quantity is RQ = FS - CI - ITQ, where RQ is the replenishment quantity, FS is the future sales volume, CI is the existing inventory quantity, and ITQ is the in-transit quantity; for products with historical sales data fluctuations exceeding the set threshold, introduce a safety inventory TQ and use the second calculation formula for the replenishment quantity, specifically: RQ = FS - CI - ITQ - TQ.

[0072] In a specific embodiment, for Model A and Model B, due to their seasonal characteristics, set the moving average period n to 3 months to match the seasonal period. For Model C, since its sales are stable, set the moving average period n to 6 months. Calculate the moving average values and predict the future sales volume.

[0073] The actual sales data for the 13th month are as follows: Model A: 1800 units, Model B: 1400 units, Model C: 980 units. The mean absolute error (MAE) is calculated as follows: Model A: |1833.33 - 1800| = 33.33 units, Model B: |1433.33 - 1400| = 33.33 units, Model C: |951.67 - 980| = 28.33 units. Set the deviation degree threshold to 50 units. Since the prediction errors of Model A, B, and C do not exceed the threshold, the current moving average period is not adjusted for the time being. If the prediction error exceeds the threshold in subsequent months, for example, the predicted sales volume of Model A in the 14th month is 1900 units and the actual sales are 1800 units, with an error of |1900 - 1800| = 100 units > 50 units, then consider shortening the moving average period, such as adjusting the moving average period of Model A from 3 months to 2 months, and recalculate the moving average values and the predicted sales volume.

[0074] Current time:

[0075] Model A: Existing inventory quantity CI = 300 units, in-transit quantity ITQ = 200 units;

[0076] Model B: Existing inventory quantity CI = 250 units, in-transit quantity ITQ = 150 units;

[0077] Model C: Existing inventory quantity CI = 180 units, in-transit quantity ITQ = 120 units.

[0078] The historical sales data of Model A fluctuates beyond the set threshold (for example, the sales data fluctuates by more than 20% in some months), and a safety stock of TQ = 100 units is introduced. At this time, the replenishment quantity of Model A uses the second calculation formula: RQ = 1833.33 - 300 - 200 - 100 = 1233.33 ≈ 1233 units.

[0079] Step S400: Obtain logistics data, determine the replenishment time according to the logistics transportation time, and adjust the replenishment quantity according to the transportation capacity; based on the replenishment time and replenishment quantity, use the dynamic programming algorithm to obtain the optimal delivery frequency, match the optimal delivery frequency with the inventory, and obtain the inventory usage plan.

[0080] Specifically, establish a data sharing interface with the logistics supplier to extract the average transportation time of various products from the supplier's shipping location to the warehouse; obtain the transportation capacity of different transportation tools from the logistics supplier, including the maximum single shipment volume and the transportation frequency per unit time;

[0081] Determine the replenishment time according to the logistics transportation time, and establish a replenishment time model. The formula is as follows: Among them, RT is the replenishment time, in days, ADFS is the future average daily sales volume, obtained by dividing the future sales volume FS by the number of days in the current month, LTT is the logistics transportation time, and SLT is the safety lead time, representing the additional time reserved by the user; when the logistics transportation is delayed, the replenishment time is advanced accordingly according to the actual delay situation; when the sales speed increases or decreases, recalculate the time to expect to sell out the existing inventory and in-transit quantity, and then adjust the replenishment time;

[0082] Compare the calculated replenishment quantity with the transportation capacity of the logistics supplier. When the replenishment quantity exceeds the maximum single shipment volume of the transportation tool, split the replenishment quantity.

[0083] Furthermore, based on the replenishment time, divide the inventory management cycle into several stages, define the state variables of each stage, including the current inventory level, the remaining replenishment quantity, and the number of completed deliveries; determine the decision variable of each stage as the delivery frequency, specifically, decide whether to make a delivery and the number of deliveries in this stage; based on the changes in inventory, replenishment operations, and the delivery process, establish a state transition equation; with the goal of minimizing inventory backlog and out-of-stock risks, construct an objective function; determine the initial state, specifically, the inventory, remaining replenishment quantity, and number of deliveries in the first stage; set the boundary conditions for each stage, specifically, the upper and lower limits of the delivery frequency, the upper limit is determined according to the transportation capacity, and the lower limit is determined according to the need to ensure the normal turnover of inventory;

[0084] Starting from the first stage, according to the state transition equation, objective function, and boundary conditions, calculate the next-stage state and objective function values corresponding to different delivery frequency decisions in the current state; select the delivery frequency that minimizes the objective function value as the optimal decision for the current stage, take the state of the next stage as the new current state, and enter the calculation of the next stage; when the calculations for all stages are completed, obtain the optimal delivery frequency from the initial stage to the final stage through the backtracking process;

[0085] Establish an association model between the optimal delivery frequency and inventory, coordinate inventory and delivery time nodes, and thus obtain an inventory usage plan.

[0086] In a specific embodiment, adjust the replenishment quantity according to the transportation capacity. For example: Model A: The calculated replenishment quantity RQ is 1333 units (rounded from 1333.33 units before). The maximum single-vehicle transportation volume is 300 units. 1333÷300 = 4 remainder 133. So, the replenishment quantity is split into 5 transports. In the first 4 transports, each transport is 300 units, and the last transport is 133 units.

[0087] Use the dynamic programming algorithm to determine the optimal delivery frequency. Taking Model A as an example, the inventory management cycle is 30 days, and it is divided into 15 stages according to the replenishment time of 2 days. State variables for each stage: Current inventory level I: For example, in the first stage, I = 300 units (the existing initial inventory). Remaining replenishment quantity R: In the first stage, R = 1333 units (the calculated replenishment quantity). Number of completed deliveries D: In the first stage, D = 0. Decision variable delivery frequency x: The value range is 0 (no delivery), 1 (one delivery), 2 (two deliveries), etc. Boundary conditions: Upper limit of delivery frequency: The truck can transport 2 times a day, and the upper limit of delivery frequency within 2 days is 4 times. Considering actual operations, it is set to 3 times. Lower limit of delivery frequency: To ensure normal inventory turnover, it is set to 1 time.

[0088] State transition equation: Next-stage inventory level I' = I+(x×300)-(ADFS×2) (each delivery is 300 units, and here 2 is the stage time of 2 days); Remaining replenishment quantity R' = R-(x×300); Number of completed deliveries D' = D + x. Objective function: Aiming to minimize inventory backlog and shortage risks, set the inventory backlog cost coefficient as w1 = 0.6 and the shortage cost coefficient as w2 = 0.8. The objective function Z = w1×max(0, I')+w2×max(0, R') (here the max function is used to determine the cost in case of inventory backlog or shortage). Starting from the first stage, calculate the next-stage state and objective function values corresponding to different delivery frequencies x, select the delivery frequency that minimizes the objective function value as the optimal decision for the current stage, and enter the calculation of the next stage. After 15 stages of calculation and backtracking, the optimal delivery frequency from the initial stage to the final stage is 3 times every 2 days.

[0089] For Model A, the delivery frequency is 3 times every 2 days, with each delivery being about 300 units. One day before each delivery, the warehouse prepares the space and manpower for receiving goods. For example, on the first day, a storage area that can accommodate 300 mobile phones is arranged, and the loading and unloading personnel are allocated. Model A has a high delivery frequency and adopts the first-in, first-out (FIFO) inventory management strategy to ensure that the mobile phones that enter the warehouse first are sold first, reducing the risk of inventory backlog.

[0090] As Figure 2 As shown in the system structure diagram of an inventory optimization system based on multi-source data of the present invention, the present invention provides an inventory optimization system based on multi-source data, including:

[0091] Data acquisition and prediction module: including: a data acquisition unit and a demand prediction unit; wherein, the data acquisition unit acquires production data and sales data, and fuses them to obtain multi-source data, and the demand prediction unit uses the multi-source data to train an LSTM neural network to predict the future market demand;

[0092] Target solving module: including: an initial setting unit and an optimization algorithm solving unit; wherein, the initial setting unit sets the objective function and constraint conditions respectively in terms of space resource consumption and time resource consumption based on the future market demand, and the optimization algorithm solving unit solves the corresponding objective function to calculate the minimum warehouse resource consumption;

[0093] Replenishment quantity calculation module: including: a moving average processing unit and a replenishment quantity calculation unit; wherein, the moving average processing unit uses the moving average method to process the historical sales data to predict the future sales volume of the products in the warehouse; the replenishment quantity calculation unit combines the future sales volume with the existing inventory quantity and the quantity in transit to calculate the replenishment quantity;

[0094] Inventory planning module: including: a replenishment quantity adjustment unit, a dynamic programming solving unit and an inventory planning unit; wherein, the replenishment quantity adjustment unit acquires logistics data, determines the replenishment time according to the logistics transportation time, and adjusts the replenishment quantity according to the transportation capacity; the dynamic programming solving unit uses the dynamic programming algorithm to obtain the optimal delivery frequency based on the replenishment time and the replenishment quantity, and the inventory planning unit matches the optimal delivery frequency with the inventory to obtain an inventory usage plan.

[0095] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

Claims

1. An inventory optimization method based on multi-source data, characterized in that It includes the following steps: Obtain production data and sales data, fuse them to get multi-source data, and use the multi-source data to train an LSTM neural network to predict future market demand; Based on the future market demand, set the objective function and constraints respectively in terms of spatial resource consumption and time resource consumption, solve the corresponding objective function, and calculate the minimum warehouse resource consumption; Apply the moving average method to process historical sales data and predict the future sales volume of warehouse products; combine the future sales volume with the existing inventory and in-transit quantity to calculate the replenishment quantity; Obtain logistics data, determine the replenishment time according to the logistics transportation time, and adjust the replenishment quantity according to the transportation capacity; Based on the replenishment time and replenishment quantity, use the dynamic programming algorithm to obtain the optimal delivery frequency, and match the optimal delivery frequency with the inventory to obtain the inventory usage plan.

2. The inventory optimization method based on multi-source data according to claim 1, wherein, The obtaining production data and sales data, fusing them to get multi-source data, and using the multi-source data to train an LSTM neural network to predict future market demand includes: Obtain production data and sales data, use the product ID as the associated field to ensure the alignment of production data and sales data in terms of time, create a data table in the relational database, and fuse the production data and sales data to get multi-source data; Set the input layer, hidden layer, and output layer of the LSTM neural network, determine the parameters of the LSTM neural network, arrange the multi-source data in chronological order, divide it into training set, validation set, and test set according to a certain proportion, and construct a data sequence; perform forward propagation and backward propagation, and conduct the evaluation and adjustment of the LSTM neural network; the user inputs the production data and sales data of the product into the trained LSTM neural network to obtain the predicted future market demand, specifically the production capacity limit, supply chain constraints, and order fulfillment rate.

3. An inventory optimization method based on multi-source data according to claim 1, characterized in that, The setting the objective function and constraints respectively in terms of spatial resource consumption and time resource consumption based on the future market demand, solving the corresponding objective function, and calculating the minimum warehouse resource consumption includes: The warehouse resource consumption includes spatial resource consumption and time resource consumption; in terms of spatial resource consumption, on the premise of meeting the future market demand, reduce the occupancy of warehouse space by products, apply the optimization layout algorithm, and optimize the warehouse layout according to the occupied space and sales speed of products to reduce spatial resource consumption; in terms of time resource consumption, according to the future market demand and sales data, arrange the procurement time through the time optimization algorithm to determine the optimal procurement time point and reduce the in-warehouse storage time of products; Based on the optimized warehouse layout, obtain the spatial resource consumption index by calculating the utilization rate of the total warehouse volume; convert the in-warehouse storage time of the product corresponding to the optimal procurement time point into a time length index to obtain the time resource consumption index; based on the spatial resource consumption index and time resource consumption index, obtain the minimum warehouse resource consumption through weighted summation.

4. The inventory optimization method based on multi-source data according to claim 3, wherein The applying the optimization layout algorithm to optimize the warehouse layout according to the occupied space and sales speed of products to reduce spatial resource consumption includes: Obtain the occupied space and sales speed of the product. The optimization layout algorithm is used to select the genetic algorithm. The objective function of the genetic algorithm is to reduce the consumption of space resources. The constraint conditions of the genetic algorithm include product correlation constraints, storage condition constraints, and channel and handling equipment constraints. Among them, the product correlation constraint means that products with correlation are placed in adjacent positions to reduce the picking path and time. The storage condition constraint represents the storage condition requirements of the product. For products with the same storage condition requirements, they are arranged in the same storage area. The channel and handling equipment constraint means the setting of reserved channels in the warehouse and the operating space of handling equipment; In the genetic algorithm, determine the size of the initial population, set the maximum number of iterations, crossover probability, and mutation probability; divide the warehouse into several storage areas, each area corresponding to a gene position of the chromosome, and the value of the gene position represents the product number stored in that area; during the iteration process, generate a new layout plan through crossover and mutation operations, and decode the new plan to obtain the actual warehouse layout; according to the objective function and constraint conditions of the genetic algorithm, calculate the fitness value for each layout plan, and the fitness value reflects the pros and cons of the layout plan in reducing the consumption of space resources; continuously generate new populations through selection, crossover, and mutation operations; evaluate the optimal layout plan generated by the genetic algorithm, make adjustments and optimizations, and implement the optimal layout plan.

5. The inventory optimization method based on multi-source data according to claim 3, wherein According to future market demands and sales data, arrange the procurement time through the time optimization algorithm to determine the best procurement time point and reduce the in-warehouse storage time of products, including: The time optimization algorithm selects the dynamic programming algorithm. According to the procurement time range and the time distribution of future market demands, divide the procurement time into several stages; define the state of each stage, including the current inventory quantity, remaining production capacity, and remaining supply capacity of the supplier; determine the decision variable of each stage, specifically the procurement quantity; establish a state transition equation according to production, supply, and sales situations; with the goal of minimizing the in-warehouse storage time of products, construct an objective function, and calculate the sum of the products of the inventory quantity of each stage and the in-warehouse storage time of the products; the constraint conditions of the dynamic programming algorithm include market demand constraints, production capacity constraints, supply capacity constraints, and non-negative inventory constraints; Determine the initial state, including the initial inventory quantity, the initial production capacity of the enterprise, and the initial supply capacity of the supplier, and set the objective function value of the initial stage to 0; starting from the first stage, according to the state transition equation, objective function, and constraint conditions, calculate the next-stage state and objective function value corresponding to different procurement quantities in the current state; select the procurement quantity that minimizes the objective function value as the optimal decision for the current stage, take the next-stage state as the new current state, and enter the calculation of the next stage; when all stages of calculation are completed, obtain the optimal procurement time point from the initial stage to the final stage through the backtracking process; starting from the optimal procurement time point of the final stage, gradually backtrack to the initial stage according to the state transition equation and decision records to determine the best procurement time point and procurement quantity of each stage.

6. The inventory optimization method based on multi-source data according to claim 1, wherein Using the moving average method to process historical sales data and predict the future sales volume of warehouse products, including: Determine the moving average period and match it with the product's seasonal cycle; the moving average period is n. For the historical sales data from 1 to n months, calculate the moving average values in sequence; extrapolate the moving average values to predict the sales volume for the (n + 1)-th month; as time goes by, compare the predicted future sales volume with the actual sales data; calculate the prediction error using the mean absolute error to measure the deviation between the predicted value and the actual value. When the deviation exceeds the set threshold, adjust the moving average period.

7. The inventory optimization method based on multi-source data according to claim 1, wherein Combining the future sales volume with the existing inventory and in-transit quantity to calculate the replenishment quantity, including: The first calculation formula for the replenishment quantity is RQ = FS - CI - ITQ, where RQ is the replenishment quantity, FS is the future sales volume, CI is the existing inventory, and ITQ is the in-transit quantity; for products with historical sales data fluctuations exceeding the set threshold, introduce a safety stock TQ and use the second calculation formula for the replenishment quantity, specifically: RQ = FS - CI - ITQ - TQ.

8. An inventory optimization method based on multi-source data according to claim 1, characterized in that Obtaining logistics data, determining the replenishment time based on the logistics transportation time, and adjusting the replenishment quantity according to the transportation capacity, including: Establish a data sharing interface with the logistics supplier to extract the average transportation time of various products from the supplier's shipping location to the warehouse; obtain the transportation capacity of different transportation tools from the logistics supplier, including the maximum single shipment volume and the transportation frequency per unit time; Determine the replenishment time according to the logistics transportation time, and establish a replenishment time model. The formula is as follows: Among them, RT is the replenishment time, in days; ADFS is the average daily future sales volume, obtained by dividing the future sales volume FS by the number of days in the current month; LTT is the logistics transportation time; SLT is the safety lead time, representing the additional time reserved by users. When the logistics transportation is delayed, the replenishment time is advanced accordingly according to the actual delay situation. When the sales speed increases or decreases, recalculate the time required to sell out the existing inventory and the in-transit quantity, and then adjust the replenishment time. Compare the calculated replenishment quantity with the transportation capacity of the logistics supplier. When the replenishment quantity exceeds the maximum single shipment volume of the transportation tool, split the replenishment quantity.

9. An inventory optimization method based on multi-source data according to claim 1, characterized in that Based on the replenishment time and replenishment quantity, use the dynamic programming algorithm to obtain the optimal delivery frequency, and match the optimal delivery frequency with the inventory to obtain the inventory usage plan, including: Based on the replenishment time, divide the inventory management cycle into several stages, define the state variables for each stage, including the current inventory level, remaining replenishment quantity, and completed delivery times; determine the decision variable for each stage as the delivery frequency, specifically deciding whether to conduct a delivery and the number of deliveries in this stage; establish a state transition equation based on the changes in inventory, replenishment operations, and delivery processes; construct an objective function with the goal of minimizing inventory backlog and out-of-stock risks; determine the initial state, specifically the inventory, remaining replenishment quantity, and delivery times in the first stage; set the boundary conditions for each stage, specifically the upper and lower limits of the delivery frequency. The upper limit is determined according to the transportation capacity, and the lower limit is determined according to the need to ensure normal inventory turnover; Starting from the first stage, calculate the next-stage state and objective function values corresponding to different delivery frequency decisions in the current state according to the state transition equation, objective function, and boundary conditions; select the delivery frequency that minimizes the objective function value as the optimal decision for the current stage, take the next-stage state as the new current state, and enter the calculation of the next stage; when all stages of the calculation are completed, obtain the optimal delivery frequency from the initial stage to the final stage through the backtracking process; Establish an association model between the optimal delivery frequency and inventory, coordinate the inventory and delivery time nodes, and thus obtain an inventory usage plan.

10. An inventory optimization system based on multi-source data, which uses an inventory optimization method based on multi-source data described in any one of claims 1-9, characterized in that, Including: Data acquisition and prediction module: including: data acquisition unit and demand prediction unit; among them, the data acquisition unit acquires production data and sales data, fuses them to obtain multi-source data, and the demand prediction unit uses the multi-source data to train an LSTM neural network to predict future market demand; Objective solution module: including: initial setting unit and optimization algorithm solution unit; among them, the initial setting unit sets the objective function and constraints respectively in terms of space resource consumption and time resource consumption based on future market demand, and the optimization algorithm solution unit solves the corresponding objective function to calculate the minimum warehouse resource consumption; Replenishment quantity calculation module: including: moving average processing unit and replenishment quantity calculation unit; among them, the moving average processing unit uses the moving average method to process historical sales data to predict the future sales volume of warehouse products; the replenishment quantity calculation unit combines the future sales volume with the existing inventory and in-transit quantity to calculate the replenishment quantity; Inventory planning module: including: replenishment quantity adjustment unit, dynamic programming solution unit and inventory planning unit; among them, the replenishment quantity adjustment unit obtains logistics data, determines the replenishment time according to the logistics transportation time, and adjusts the replenishment quantity according to the transportation capacity; the dynamic programming solution unit uses the dynamic programming algorithm to obtain the optimal delivery frequency based on the replenishment time and replenishment quantity, and the inventory planning unit matches the optimal delivery frequency with the inventory to obtain an inventory usage plan.

Images