Data analysis method, device, apparatus and readable medium

Abstract

Embodiments of the present disclosure provide a data analysis method, apparatus, device, and readable medium. The method includes: determining predicted order data for a target object based on a time series feature representation of the target object's historical order data, where the historical order data indicates at least the time and number of orders generated; determining a supply strategy for the target object based on the time series feature representation and feature representations corresponding to the predicted order data; and determining the supply quantity for the target object based on the supply strategy. This method improves data analysis efficiency and accuracy.

Description

Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of computer technology, and more particularly, to a method, apparatus, device, and computer-readable storage medium for data analysis. Background Art

[0002] Sales of products with large fluctuations are often difficult to predict. This can lead to errors when formulating replenishment strategies for these products. Determining replenishment strategies for these products is a pressing issue. Summary of the Invention

[0003] In a first aspect of the present disclosure, a data analysis method is provided. The method comprises: determining predicted order data for a target object based on a time series feature representation of historical order data of the target object, where the historical order data indicates at least the time and number of orders generated; determining a supply strategy for the target object based on the time series feature representation and feature representations corresponding to the predicted order data; and determining the supply quantity for the target object based on the supply strategy.

[0004] In a second aspect of the present disclosure, a data analysis apparatus is provided. The apparatus includes: a determination module configured to determine predicted order data for a target object based on a temporal feature representation of historical order data of the target object, wherein the historical order data at least indicates an order generation time and a number of orders generated; an execution module configured to determine a supply strategy for the target object based on the temporal feature representation and a feature representation corresponding to the predicted order data; and a supply module configured to determine the supply quantity of the target object based on the supply strategy.

[0005] In a third aspect of the present disclosure, an electronic device is provided. The electronic device includes at least one processing unit; and at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, causing the electronic device to perform the method of the first aspect of the present disclosure.

[0006] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium and can be executed by a processor to perform the method according to the first aspect of the present disclosure.

[0007] In a fifth aspect of the present disclosure, a computer program product is provided, comprising computer-executable instructions, wherein the computer-executable instructions can be executed by a processor to perform the method according to the first aspect of the present disclosure.

[0008] It should be understood that the contents described in the Summary of the Invention are not intended to limit the key features or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The above and other features, advantages and aspects of the various embodiments of the present disclosure will become more apparent hereinafter with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0010] Figure 1 A schematic diagram illustrating an example environment in which embodiments of the present disclosure can be implemented;

[0011] Figure 2 A flowchart showing a process of data analysis according to some embodiments of the present disclosure;

[0012] Figure 3 A schematic diagram illustrating a data analysis process according to some embodiments of the present disclosure;

[0013] Figure 4 A schematic diagram illustrating an example of very stable commodity simulation data according to some embodiments of the present disclosure;

[0014] Figure 5 A schematic diagram illustrating an example of non-stationary commodity simulation data according to some embodiments of the present disclosure;

[0015] Figure 6 A schematic diagram illustrating an example of unsteady commodity simulation turnover data according to some embodiments of the present disclosure;

[0016] Figure 7 A schematic structural block diagram showing an apparatus for data analysis according to certain embodiments of the present disclosure; and

[0017] Figure 8 A block diagram of an electronic device is shown in which one or more embodiments of the present disclosure may be implemented. DETAILED DESCRIPTION

[0018] The following describes embodiments of the present disclosure in more detail with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0019] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, i.e., "including but not limited to." The term "based on" should be understood as "based at least in part on." The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment." The term "some embodiments" should be understood as "at least some embodiments." Other explicit and implicit definitions may be included below.

[0020] It should be noted that the acquisition, storage and application of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0021] It is understandable that before using the technical solutions disclosed in the various embodiments of this disclosure, the type, scope of use, usage scenarios, etc. of the personal information involved in this disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0022] For example, in response to receiving a user's active request, a prompt message is sent to the user to clearly remind the user that the operation requested to be performed will require obtaining and using the user's personal information, so that the user can independently choose whether to provide personal information to the electronic device, application, server or storage medium and other software or hardware that performs the operation of the technical solution of the present disclosure based on the prompt message.

[0023] As an optional but non-limiting implementation, in response to receiving a user's active request, a prompt message may be sent to the user, for example, in the form of a pop-up window, in which the prompt message may be presented in text form. Furthermore, the pop-up window may also include a selection control for the user to select "agree" or "disagree" to provide personal information to the electronic device.

[0024] As used herein, the term "model" can learn the association between corresponding inputs and outputs from training data, so that after training is completed, corresponding outputs can be generated for given inputs. The generation of the model can be based on machine learning technology. Deep learning is a machine learning algorithm that processes inputs and provides corresponding outputs by using multiple layers of processing units. A neural network model is an example of a model based on deep learning. In this article, "model" may also be referred to as "machine learning model", "learning model", "machine learning network" or "learning network", and these terms are used interchangeably in this article.

[0025] A "neural network" is a machine learning network based on deep learning. A neural network is capable of processing inputs and providing corresponding outputs. It typically includes an input layer, an output layer, and one or more hidden layers between the input and output layers. Neural networks used in deep learning applications typically include many hidden layers, thereby increasing the depth of the network. The layers of a neural network are connected in sequence so that the output of the previous layer is provided as input to the next layer, where the input layer receives the input of the neural network and the output of the output layer serves as the final output of the neural network. Each layer of a neural network includes one or more nodes (also called processing nodes or neurons), each of which processes the input from the previous layer.

[0026] Generally speaking, machine learning can be roughly divided into three stages, namely the training stage, the testing stage and the application stage (also called the inference stage). In the training stage, a given model can be trained using a large amount of training data, and the parameter values ​​are continuously updated iteratively until the model can obtain consistent inferences that meet the expected goals from the training data. Through training, the model can be considered to be able to learn the association between input and output (also called input-to-output mapping) from the training data. The parameter values ​​of the trained model are determined. In the testing stage, the test input is applied to the trained model to test whether the model can provide the correct output, thereby determining the performance of the model. The testing stage can sometimes be integrated into the training stage. In the application or inference stage, the trained model can be used to process the actual model input based on the parameter values ​​obtained through training to determine the corresponding model output.

[0027] Traditional replenishment solutions can be implemented using stochastic programming models, single-decision stochastic optimization models, distributed robust replenishment optimization models, and so on. Traditional end-to-end replenishment solutions can be implemented using constructed models. Specifically, models can be constructed based on historical product sales, replenishment lead time (VLT), replenishment cycle, and other factors. The model's loss can also be constructed based on sales forecast accuracy, sales loss from replenishment, availability, and turnover.

[0028] In the context of product sales forecasting, many time series forecasting methods are used, such as the Autoregressive Integrated Moving Average model (ARIMA), the Prophet time series forecasting model, the Long Short Term Memory Network (LSTM), the DeepAR model, and the Temporal Fusion Transformers (TFT).

[0029] Traditional replenishment plans consist of two phases: forecasting and replenishment. During the forecasting phase, daily sales forecasts for the upcoming period are determined based on the forecast results. During the replenishment phase, optimal replenishment parameters are determined based on the sales forecasts, target turnover, and inventory levels.

[0030] Traditional product replenishment solutions present numerous challenges. The first is the uncertainty of product targets. The most straightforward objective in the model used is to maximize revenue, which requires considering both holding costs and stockout costs. However, in practice, these two figures can be difficult to calculate due to factors such as inventory, fulfillment, and returns. In such cases, product turnover and availability can be used as an alternative to calculating holding costs and stockout costs.

[0031] In merchandise sales, replenishment strategies are typically determined based on predetermined targets. For example, a target turnover and availability constraints for a given product, or a constraint where the target turnover is as low as possible, are considered. However, in end-to-end replenishment scenarios, this two-stage replenishment model often presents challenges. For example, for some products, it's difficult to simultaneously meet both turnover and availability requirements. Furthermore, it's difficult to determine targets for volatile products. Furthermore, established targets can be difficult to meet, and so on.

[0032] Furthermore, because the turnover of unstable products often fluctuates widely around the target turnover, the evaluation of these products is unstable. Specifically, for products with unstable sales, not only is sales forecasting difficult, but replenishment assessments are also prone to errors due to objective fluctuations, even if replenishment is accurate. This means that the assessment is unreliable (e.g., unstable turnover assessments).

[0033] In a two-stage model, when measuring the turnover of multiple volatile products, some volatile products may have overstated turnover predictions, while others may have underestimated turnover predictions. In an end-to-end replenishment scenario, if each sample is subjected to a loss designed based on the evaluation results, and the target in the loss is problematic, it will be difficult to train a model with good prediction results.

[0034] In light of this, embodiments of the present disclosure provide a data analysis solution. This solution includes: determining predicted order data for a target object based on a time series feature representation of the target object's historical order data, where the historical order data at least indicates the time and number of orders generated. Determining a supply strategy for the target object based on the time series feature representation and feature representations corresponding to the predicted order data. Determining the supply quantity for the target object based on the supply strategy. This approach improves data analysis efficiency and accuracy.

[0035] Figure 1A schematic diagram of an example environment 100 is shown in which embodiments of the present disclosure can be implemented.

[0036] like Figure 1 As shown, the historical order data 105 corresponding to the target object can be provided to the data analysis system 110 to determine the supply number 125 for the target object. In some embodiments, the target object may indicate a target commodity corresponding to the business. Specifically, the target commodity may indicate a commodity with unstable sales, that is, a commodity whose sales distribution changes over time (excluding promotional days and other activity days). It will be understood that this is merely exemplary and not restrictive. In other examples, the historical order data 105 may at least indicate the order generation time and the order generation number. In more examples, the supply number 125 may indicate the replenishment quantity of the target object. It will be understood that this is merely exemplary and not restrictive.

[0037] The data analysis system 110 may include any computing system with computing capabilities, such as various computing devices / systems, terminal devices, server devices, etc. The terminal device may be any type of mobile terminal, fixed terminal or portable terminal, including a mobile phone, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a media computer, a multimedia tablet, a handheld computer, a portable game terminal, a VR / AR device, a personal communication system (PCS) device, a personal navigation device, a personal digital assistant (PDA), an audio / video player, a digital camera / camcorder, a positioning device, a television receiver, a radio broadcast receiver, an e-book device, a gaming device, or any combination thereof, including accessories and peripherals of these devices or any combination thereof.

[0038] It should be understood that the structure and function of the various elements in the environment 100 are described for illustrative purposes only and do not imply any limitation on the scope of the present disclosure.

[0039] Some embodiments of the present disclosure will be described below with reference to the accompanying drawings.

[0040] Figure 2 FIG2 shows a flow chart of a process 200 for data analysis according to some embodiments of the present disclosure. The process 200 may be implemented at the data analysis system 110. For ease of discussion, reference will be made to FIG200. Figure 1 The process 200 is described with reference to the environment 100 of FIG.

[0041] In block 210 , the data analysis system 110 determines predicted order data for the target object based on the time series feature representation of the historical order data 105 of the target object.

[0042] In some embodiments, the historical order data 105 may at least indicate the time when an order was generated and the number of orders generated. In some embodiments, the time series feature representation of the historical order data 105 may indicate at least one of the sales quantity of the target object and the price of the target object. In other examples, a time series feature extraction may be performed on the historical order data 105 using a time series processing network to obtain a time series feature representation. Specifically, the time series processing network also includes an encoder and a decoder. In more examples, the time series processing network may be determined based on a multilayer perceptron (MLP) and a deep neural network (such as WaveNet). Furthermore, the time series processing network may also be determined based on an MLP and a long short-term memory network (LSTM). It will be understood that this is merely exemplary and not restrictive. In this way, some irrelevant data may be filtered out by extracting the time series features of the historical order data 105, thereby improving the efficiency of object supply.

[0043] In some embodiments, the data analysis system 110 can determine the order data distribution of the target object within a given future time range based on the time series feature representation. In some embodiments, the order data distribution can at least indicate the distribution of predicted sales values ​​of the target object. In some embodiments, the order data distribution can be predicted using quantile prediction or distribution prediction to retain the intermittent information of the time series features. In some embodiments, the data analysis system 110 can determine the predicted order data of the target object based on the probability of the order data distribution within a given future time range. In this way, the volatility of non-stationary objects can be better adapted and the accuracy of object supply can be improved.

[0044] In block 220 , the data analysis system 110 determines a supply strategy for the target object based on the time series feature representation and the feature representation corresponding to the predicted order data.

[0045] In some embodiments, the data analysis system 110 can determine multiple candidate supply strategies for the target object based on the time series feature representation and the feature representation corresponding to the predicted order data. In some embodiments, the data analysis system 110 can determine the supply strategy for the target object from multiple candidate supply strategies. In some embodiments, the data analysis system 110 can determine the supply strategy for the target object from multiple candidate supply strategies based on certain rules (e.g., rules containing business parameters that interact with business needs). In this way, it is possible to be closer to business needs to improve the accuracy and efficiency of object supply.

[0046] In some embodiments, the data analysis system 110 may determine a supply strategy for a target object from multiple candidate supply strategies using a trained supply strategy determination model based on the time series feature representation and the feature representation corresponding to the predicted order data. In some embodiments, the supply strategy determination model may be trained in the following manner. Specifically, the data analysis system 110 may determine a predicted order data sample for the training object based on the time series feature representation of the training object's order data sample. In some embodiments, the data analysis system 110 may determine simulation data corresponding to multiple candidate supply strategy samples for the training object based on the predicted order data sample for the training object. The simulation data indicates the turnover rate and spot rate corresponding to each supply strategy sample. In some embodiments, the data analysis system 110 may determine at least one first supply strategy sample from each candidate supply strategy sample based on the simulation data corresponding to the multiple candidate supply strategy samples using simulation data constraints. In some embodiments, the simulation data constraints may indicate at least one of the following: the spot rate is not higher than a spot rate constraint parameter, the spot rate constraint parameter is determined based on a preset base spot rate and turnover rate, and the turnover rate is not higher than the turnover rate constraint parameter. The turnover rate constraint parameter is determined based on a preset base turnover rate. In some embodiments, the turnover rate constraint parameter may indicate a base turnover rate of 1 / 3. In some embodiments, the turnover rate constraint parameter may be adjusted according to business needs. It is understood that this is merely exemplary and not restrictive.

[0047] In some embodiments, the data analysis system 110 may utilize the supply strategy determination model to determine at least one second supply strategy sample from multiple candidate supply strategy samples based at least on the time series feature representation of the training order data sample and the feature representation corresponding to the predicted order data sample. Furthermore, the data analysis system 110 may train the supply strategy determination model based on the at least one first supply strategy sample and the at least one second supply strategy sample. In this manner, the accuracy of the supply strategy determination model output can be improved through the training process of the supply strategy determination model.

[0048] In some embodiments, the data analysis system 110 may obtain the existing inventory count of the target object. In some embodiments, the data analysis system 110 may determine multiple candidate supply strategies based on the existing inventory count and each of multiple target inventory counts. In this way, multiple candidate supply strategies can be used to adapt to the instability of unstable objects and improve object supply efficiency.

[0049] In some embodiments, the data analysis system 110 may determine multiple target inventory levels using a trained inventory forecasting model based on a time series feature representation and a feature representation corresponding to the predicted order data. In some embodiments, the inventory forecasting model training process includes: the data analysis system 110 may determine a predicted order data sample for the training subject based on the time series feature representation of the training subject's order data sample. In some embodiments, the data analysis system 110 may determine multiple inventory levels corresponding to the training subject using the inventory forecasting model based on the time series feature representation of the order data sample and the predicted order data sample of the training subject. In some embodiments, the data analysis system 110 may determine an inventory turnover rate prediction result for each of the multiple inventory levels within a given time window. Specifically, the inventory turnover ratio (ITO) can measure inventory utilization efficiency within a certain time period (also referred to as a time window) and is defined as the ratio of inventory value to sales cost. That is, a higher inventory turnover ratio indicates faster inventory depletion and more effective inventory management. Furthermore, the turnover performance within a windowed time period can be used to assess the availability constraint rate. In some examples, the inventory turnover rate within a certain time window can be determined based on the total inventory quantity of the replenishment strategy within the window period, the total sales volume of the replenishment strategy within the window period, the sequence number of the time window, the time window selected for calculating the inventory turnover rate, the simulation history duration, and the evaluation time window size. Specifically, the maximum value of the sequence number of the time window can be determined based on the difference between the simulation history duration and the evaluation time window size. It will be understood that this is merely exemplary and non-restrictive.

[0050] In some embodiments, the data analysis system 110 can determine the predicted mean inventory turnover value and the predicted standard deviation inventory turnover value for each inventory quantity in a given time window based on the inventory turnover rate prediction results. Specifically, the mean inventory turnover value can represent the average inventory turnover rate throughout the simulation period and can reflect the average speed of inventory turnover. Furthermore, the mean inventory turnover value can be determined based on the sequence number of the time window, the simulation history duration, the evaluation time window size, and the inventory turnover rate. It will be understood that this is merely exemplary and not restrictive.

[0051] Furthermore, the inventory turnover standard deviation can indicate the volatility of inventory turnover and reflect the stability of inventory management. Specifically, a large inventory turnover standard deviation can indicate that inventory turnover fluctuates significantly across different time windows and that inventory management is relatively unstable. In some examples, the inventory turnover standard deviation can be determined based on the sequence number of the time window, the inventory turnover rate, the mean inventory turnover value, the historical simulation duration, and the evaluation time window size. It should be understood that this is merely exemplary and not restrictive.

[0052] In some embodiments, the data analysis system 110 can determine the actual inventory turnover rate of each inventory quantity in a given time window based on the order data sample of the training object. More importantly, the data analysis system 110 can train the inventory forecasting model based on the inventory turnover mean prediction result, the inventory turnover standard deviation prediction result, the actual inventory turnover rate and the preset inventory turnover rate target value. Specifically, at least one of the turnover mean (itos), spot rate mean (spot_rates), turnover variance (var_ito), basic spot rate constraint (base_min_spot) and basic turnover value (base_ito) of each supply strategy can be input first. In some embodiments, the data analysis system 110 can adjust the spot rate constraint according to the turnover variance (that is, the larger the turnover variance, the smaller the spot rate constraint).

[0053] In some embodiments, the spot rate constraint can be determined by the following method: Min_spot = base_min_spot * (0.75 + 0.25 / (1 + var_ito)). This determination method indicates that the spot rate constraint is greater than or equal to 0.75 times the basic spot rate constraint, and the spot rate constraint is less than the basic spot rate constraint. In some embodiments, the supply strategy that satisfies the spot rate less than the spot rate constraint is discarded. In some embodiments, the supply strategy with a turnover value less than the turnover threshold is discarded. For example, the turnover threshold can be set to 1 / 3 times the basic turnover value. Assuming that the basic turnover value is 9, the turnover threshold is 3. In this case, the supply strategy with a turnover less than the turnover threshold is discarded, that is, the supply strategy with a turnover less than 3 is discarded. It can be understood that this is merely exemplary and not restrictive. More preferably, the supply strategy with the smallest turnover is selected from the supply strategies after the above-mentioned discarding operation is completed, and the supply strategy is labeled as the optimal strategy. In addition, the spot rate constraint and the turnover threshold can be adjusted according to business needs. In this way, it is possible to flexibly adapt to business needs and improve the corresponding supply efficiency and accuracy.

[0054] At block 230 , the data analysis system 110 determines the supply number 125 of the target object based on the supply policy.

[0055] For example, assuming that the supply policy indicates that the supply number 125 is 13, then the supply number 125 of the target object can be determined to be 13. In some embodiments, the supply number 125 indicated by the supply policy can correspond to time. Specifically, the supply policy can indicate that the supply number 125 on the 8th of this month in the future (assuming that the supply policy is determined on the 5th of this month) is 20. It can be understood that this is merely exemplary and not restrictive. In this way, the supply number 125 of the target object can be determined with high accuracy according to the corresponding supply policy, and accurate end-to-end supply of the target object can be achieved.

[0056] Figure 3 FIG2 shows a schematic diagram of a process 300 for data analysis according to some embodiments of the present disclosure. The process 300 may be implemented at the data analysis system 110. For ease of discussion, reference will be made to FIG2 . Figure 1 The process 300 is described with reference to the environment 100 of FIG.

[0057] like Figure 3 As shown, the data analysis system 110 can provide the historical order data 105 of the target object to the time series processing network 315 to determine the time series feature representation 320 of the historical order data 105. In some embodiments, the time series processing network 315 can be determined based on MLP and WaveNet. In other embodiments, the time series processing network 315 can also be determined based on Lstm and MLP. In some embodiments, the data analysis system 110 can provide the time series feature representation 320 to the sales forecasting model 325 to determine the predicted order data 330 of the target object. Specifically, the sales forecasting model 325 can determine the predicted order data 330 through quantile forecasting or distribution forecasting. In some embodiments, the predicted order data 330 can at least indicate sales forecast data for a period of time in the future. In other embodiments, the sales forecasting model 325 can be determined based on MLP and quantile function. It can be understood that this is merely exemplary and not restrictive.

[0058] In some embodiments, the data analysis system 110 may provide the time series feature representation 320 and the predicted order data 330 to the inventory forecasting model 340 to determine a candidate supply strategy set 345. Specifically, the candidate supply strategy set 345 may include multiple ordered candidate supply strategies. For example, assuming that the candidate supply strategy set 345 includes 10 ordered candidate supply strategies, the corresponding inventory forecasting model 340 may include an MLP, where the last layer of the MLP outputs 10 prediction values. The output value of the inventory forecasting model 340 is expressed as follows:

[0059] output=mlp(x) (1)

[0060] Here, output represents the output value of the inventory forecasting model 340, and mlp(x) represents the output value of the last layer of the MLP.

[0061] The formula for converting the negative output value of the inventory forecast model 340 to 0 is as follows:

[0062] output=relu(output) (2)

[0063] Wherein, output represents the output value of the inventory forecasting model 340, relu(output) represents converting the negative output value in the above formula (1) to 0, and the positive output value in the above formula (1) remains unchanged.

[0064] The formula for determining the supply strategy 355 based on the output value is as follows:

[0065]

[0066] Among them, k represents the kth replenishment strategy, n represents the sequence number of the output value, output n Represents an additional non-negative output value to be accumulated with the nth output value.

[0067] In some embodiments, the data analysis system 110 may provide the candidate provisioning strategy set 345 to the provisioning strategy determination model 350 to determine a provisioning strategy 355 for the target object from a plurality of candidate provisioning strategies in the candidate provisioning strategy set 345 .

[0068] Furthermore, during the training phase of the sales forecasting model 325, the data analysis system 110 can determine sales losses 335 based on the predicted order data 330 output by the forecasting network and the actual order data, and adjust the parameters of the sales forecasting model 325 based on the sales forecasting losses 335.

[0069] In some embodiments, during the training phase of the supply strategy determination model 350 , the data analysis system 110 may provide the candidate supply strategy set 345 to the simulation module 360 ​​to perform simulation on each candidate supply strategy and determine the simulation result corresponding to each candidate supply strategy.

[0070] Specifically, the simulation can simulate the commodity procurement, arrival, and sales processes based on the replenishment quantity output by the algorithm, and calculate the data of each stage in the commodity process. The specific simulation process is as follows: the data analysis system 110 sequentially performs the simulation in-transit inventory quantity update, simulation sales quantity update, simulation sales loss calculation, simulation inventory quantity update, simulation order update, simulation arrival update, and simulation in-transit update. More specifically, in the process of simulating order update, the predicted target inventory quantity for the day is obtained, and a judgment is performed on the predicted target inventory quantity. If the day is not an order day or the target inventory quantity is less than or equal to the sum of the simulated inventory quantity and the in-transit inventory quantity, no order is placed. Furthermore, in the process of simulating arrival update, when the replenishment lead time (vendor leadtime, vlt) at time t is v_t, then the simulated arrival quantity at time t+v_t = the simulated arrival quantity at time t+v_t + the simulated order quantity at time t.

[0071] The formula for calculating the simulated in-transit inventory quantity is as follows:

[0072] Simulated in-transit inventory quantity = Simulated in-transit inventory quantity on the previous day - Simulated inventory quantity arriving on the current day

[0073] Number (4)

[0074] The formula for calculating simulated sales volume is as follows:

[0075] Simulated sales = min(real sales, simulated available inventory on the day) (5)

[0077] The formula for calculating the available inventory quantity on the simulation day is as follows:

[0078] Available inventory on the simulated day = simulated inventory on the previous day + simulated inventory on the day of arrival

[0079] Number (6)

[0080] The formula for calculating simulated sales loss is as follows:

[0081] Simulated sales loss = real sales – simulated sales (7)

[0082] The formula for calculating the simulated inventory quantity is as follows:

[0083] Simulated inventory quantity = Simulated inventory quantity of the previous day + Inventory quantity of goods arriving on the current day - Simulated sales quantity

[0084] Quantity (8)

[0085] The target inventory quantity is calculated as follows:

[0086] Target inventory quantity = existing inventory quantity - simulated inventory quantity - in-transit inventory quantity (9)

[0087] The calculation formula of the simulated in-transit quantity is as follows:

[0088] Simulated in-transit quantity = Simulated in-transit quantity + Simulated order quantity (10)

[0089] In some embodiments, the data analysis system 110 can use the simulation module 360 ​​to provide the simulation results corresponding to each candidate supply strategy to the optimal strategy evaluation module 370 to determine the optimal supply strategy and add a label to the optimal supply strategy. In some embodiments, the optimal strategy evaluation module 370 can be configured with multiple rules (such as rules containing business parameters that interact with business needs) and consider the strategy with the best turnover rate as the optimal supply strategy while satisfying the spot rate constraint. Specifically, the label can be expressed as "optimal strategy". It will be understood that this is merely exemplary and not restrictive. In some embodiments, the data analysis system 110 can determine the strategy loss 375 based on the supply strategy 355 output by the supply strategy determination model 350 and the optimal supply strategy output by the optimal strategy evaluation module 370. Adjust the parameters of the supply strategy determination model 350 according to the strategy loss 375.

[0090] In some embodiments, the data analysis system 110 may further train the sales forecasting model 325 based on the target inventory loss 365 and the cumulative target inventory in the simulation results corresponding to each candidate supply strategy output by the simulation module 360. Specifically, the target inventory loss 365 may be determined by a sales loss function, which may be used to optimize the prediction of the target inventory quantity (ti) during the training process. In some embodiments, the sales loss function may balance the spot rate and turnover by using a balance weight between the spot rate and turnover. In other examples, sales loss may be used in place of the spot rate, and cumulative target inventory may be used in place of turnover.

[0091] In some examples, the sales loss function can be determined based on a weight for balancing spot rate and inventory turnover rate (the trade-off between spot rate or sales loss and inventory turnover rate can be controlled), sales loss, mean of turnover, actual turnover rate (generally determined by actual inventory / actual sales in the time window corresponding to the inventory turnover rate), target turnover rate (which can be expected / ideal), standard deviation of turnover, and hyperparameters.

[0092] In some embodiments, the deviation between the predicted inventory turnover rate and the target turnover rate can be measured by calculating the result of (turnover mean - actual turnover rate + target turnover rate) / standard deviation of turnover, and this deviation is standardized so that it is associated with the actual volatility (standard deviation). In some examples, in the sales loss function, the thresholds of different conditions in the loss function can be defined by hyperparameters, that is, when the prediction deviation is small (that is, close to the target), the sales loss is mainly considered. When the prediction deviation is large, the square of the deviation is added as a penalty term to further optimize the prediction. Through these formulas, the model can optimize the predicted target inventory amount on the basis of balancing the spot rate (or sales loss) and the inventory turnover rate, thereby managing inventory more effectively and reducing losses. It can also guide the model to perform more accurate inventory management strategy optimization by comprehensively considering inventory utilization efficiency, volatility and its impact on sales.

[0093] In addition, the data analysis system 110 may first train the sales forecast model 325 based on the sales loss 335 to determine the parameters of the sales forecast model 325. In some embodiments, the data analysis system 110 may train the inventory forecast model 340 based on the target inventory loss 365 to determine the parameters of the inventory forecast model 340. In some embodiments, the analysis system 110 may train the supply strategy determination model 350 based on the strategy loss 375. Additionally, during the training process of the supply strategy determination model 350, as the number of iterations increases, the strategy loss 375 may gradually decrease. When the strategy loss 375 satisfies a loss threshold value, it may be determined that the supply strategy determination model 350 has completed the training process. Specifically, the loss threshold value may be determined or adjusted based on business needs. It will be understood that this is merely exemplary and not restrictive.

[0094] Furthermore, the data analysis system 110 may provide at least one of the sales loss 335, the target inventory loss 365, and the strategy loss 375 to the training strategy module 380 to determine the loss 385. Alternatively, the data analysis system 110 may fine-tune the supply strategy determination model 350 based on the loss 385.

[0095] Continue to refer to Figure 4-Figure 6 When the target object is a commodity, further explanation is given on the simulation comparison between very stable commodities and unstable commodities during the simulation process. Figure 4-Figure 6 In this example, the following assumptions are met: the product satisfies vlt = 5 (i.e., it arrives 5 days after purchase), is replenished every seven days, the point prediction accuracy is 100% (i.e., consistent with the true value), the target turnover for replenishment is 30 days, the time series length is 200 days, and the total sales volume is 100. Furthermore, the turnover measurement window is 30 days, meaning that each evaluation is for a turnover of approximately 30 days.

[0096] Figure 4 A diagram 400 is shown showing an example of very stationary commodity simulation data according to some embodiments of the present disclosure. For ease of discussion, reference will be made to Figure 1 The schematic diagram 400 is described with reference to the environment 100 of FIG.

[0097] like Figure 4 As shown in the figure, for a very stable product, the corresponding daily sales volume is 1, and the sales days are 200. During the simulation of a relatively perfect replenishment of this type of product, it can be found that the final product turnover fluctuates around 30 days, with very small fluctuations. Specifically, the detailed turnover data for a very stable product can be 30.1, 30.03333, 29.9, 30.0, 30.166666, 29.83334, and so on.

[0098] Figure 5 A schematic diagram 500 is shown showing an example of unsteady commodity simulation data according to some embodiments of the present disclosure. For ease of discussion, reference will be made to Figure 1 The schematic diagram 500 is described with reference to the environment 100 of FIG.

[0099] like Figure 5 As shown in the figure, for non-stationary products, the sales volume is set to [0, 0, 4, 0, 0, 0, 0, 0, 0, 4, 0, 0, 0, 0, 0, 8, 0, 0, 4] and repeated 10 times. During the simulation of a relatively perfect replenishment of this type of product, it can be found that the final product turnover fluctuates around 30 days, and the fluctuation is large. Specifically, the detailed turnover data for non-stationary products can be 30.285715, 36.166668, 28.125, 24.444445, 41.166668, 25.666666, and so on.

[0100] Figure 6 A schematic diagram 600 is shown showing an example of unsteady commodity simulation turnover data according to some embodiments of the present disclosure. For ease of discussion, reference will be made to Figure 1 The schematic diagram 600 is described with reference to the environment 100 of FIG.

[0101] like Figure 6As shown, for unstable commodities, turnover fluctuates around 30 days, and the fluctuation is large. Specifically, the turnover fluctuation range is from a value between 22.5 and 25.0 to a value between 40.0 and 42.5. Specifically, the detailed turnover data of unstable commodities can be 30.285715, 30.571428, 36.6666666668, 37.16666828.375, 28.125, 27.875, 27.25, 24.0, 23.88889, 23.777777779, 30.714285, 30.857143, 31.0, 36.83332, 32.0, 32.42857, 38.5, 39.833332, 40.5, 30.75, 31.125 , 30.875, 33.57143, 33.714287, 33.857143, 39.833333322, 34.57143, 41.166668, 41.5, 27.22222221, 30.375, 30.125, 26.444445, 26.11111, 25.666666, 31.285715, 37.037.333332, 28.125, 27.875, 24.7777779, 24.333334, 24.0, etc. In summary, the embodiments of the present disclosure can determine the time series feature representation and the feature representation corresponding to the predicted order data based on the historical order data of the target object, so as to determine the supply strategy of the target object, and determine the supply quantity of the target object according to the supply strategy. In this way, the efficiency and accuracy of object supply can be improved, and end-to-end target object supply can be achieved. It can also better meet business needs by adjusting conditions such as turnover constraints and spot rates.

[0102] The embodiments of the present disclosure also provide corresponding devices for implementing the above methods or processes.

[0103] Figure 7 FIG2 shows a schematic structural block diagram of an apparatus 700 for data analysis according to certain embodiments of the present disclosure. The apparatus 700 may be implemented as or included in the data analysis system 110. Each module / component in the apparatus 700 may be implemented by hardware, software, firmware, or any combination thereof.

[0104] like Figure 7As shown, apparatus 700 includes a determination module 710 configured to determine predicted order data for a target object based on a temporal feature representation of historical order data of the target object, where the historical order data at least indicates the time and number of orders generated. Apparatus 700 also includes an execution module 720 configured to determine a supply strategy for the target object based on the temporal feature representation and a feature representation corresponding to the predicted order data. Apparatus 700 also includes a supply module 730 configured to determine the supply quantity of the target object based on the supply strategy.

[0105] In some embodiments, the determination module 710 is further configured to determine the order data distribution of the target object within a given time range in the future based on the time series feature representation; and determine the predicted order data of the target object based on the order data distribution probability within the given time range in the future.

[0106] In some embodiments, the execution module 720 is further configured to determine multiple candidate supply strategies for the target object based on the time series feature representation and the feature representation corresponding to the predicted order data; and determine the supply strategy for the target object from the multiple candidate supply strategies.

[0107] In some embodiments, the execution module 720 is further configured to determine the supply strategy for the target object from a plurality of candidate supply strategies using a trained supply strategy determination model based on the time series feature representation and the feature representation corresponding to the predicted order data.

[0108] In some embodiments, the execution module 720 is further configured to determine the predicted order data sample of the training object based on the temporal feature representation of the order data sample of the training object; determine the simulation data corresponding to multiple candidate supply strategy samples of the training object based on the predicted order data sample of the training object, the simulation data indicating the turnover rate and spot rate corresponding to each supply strategy sample; determine at least one first supply strategy sample from each candidate supply strategy sample based on the simulation data corresponding to multiple candidate supply strategy samples using the simulation data constraints; determine at least one second supply strategy sample from multiple candidate supply strategy samples using the supply strategy determination model based on at least the temporal feature representation of the order data sample of the training object and the feature representation corresponding to the predicted order data sample; and train the supply strategy determination model based on at least one first supply strategy sample and at least one second supply strategy sample.

[0109] In some embodiments, the simulation data constraint indicates at least one of the following: the spot rate is not higher than the spot rate constraint parameter, the spot rate constraint parameter is determined based on a preset basic spot rate and turnover rate, and the turnover rate is not higher than the turnover rate constraint parameter, and the turnover rate constraint parameter is determined based on a preset basic turnover rate.

[0110] In some embodiments, the supply strategy for the target object is determined from multiple candidate supply strategies, and the execution module 720 is further configured to determine multiple target inventory quantities using a trained inventory forecasting model based on the time series feature representation and the feature representation corresponding to the predicted order data; obtain the existing inventory quantity of the target object; and determine multiple candidate supply strategies based on the existing inventory quantity and each target inventory quantity among the multiple target inventory quantities.

[0111] In some embodiments, the execution module 720 is further configured to determine a predicted order data sample of the training object based on the time series feature representation of the order data sample of the training object; determine a plurality of inventory quantities corresponding to the training object using the inventory forecasting model based on the time series feature representation of the order data sample and the predicted order data sample of the training object; determine the inventory turnover rate forecast result of each of the plurality of inventory quantities in a given time window; determine the inventory turnover mean forecast result and the inventory turnover standard deviation forecast result of each inventory quantity in a given time window based on the inventory turnover rate forecast result; determine the actual inventory turnover rate of each inventory quantity in a given time window based on the order data sample of the training object; and train the inventory forecasting model based on the inventory turnover mean forecast result, the inventory turnover standard deviation forecast result, the actual inventory turnover rate and the preset inventory turnover rate target value.

[0112] The units and / or modules included in the device 700 can be implemented in various ways, including software, hardware, firmware, or any combination thereof. In some embodiments, one or more units and / or modules can be implemented using software and / or firmware, such as machine executable instructions stored on a storage medium. In addition to or as an alternative to machine executable instructions, some or all of the units and / or modules in the device 700 can be implemented at least in part by one or more hardware logic components. By way of example and not limitation, exemplary types of hardware logic components that can be used include field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0113] It should be understood that one or more steps in the above method can be performed by a suitable electronic device or combination of electronic devices. Such an electronic device or combination of electronic devices may include, for example, Figure 1 The data analysis system 110 in FIG.

[0114] Figure 8 8 is a block diagram of an electronic device 800 in which one or more embodiments of the present disclosure may be implemented. Figure 8The illustrated electronic device 800 is merely exemplary and should not be construed as limiting the functionality and scope of the embodiments described herein. Figure 8 The electronic device 800 shown can be used to implement Figure 1 Data analysis system 110 or Figure 7 device 700.

[0115] like Figure 8 As shown, electronic device 800 is in the form of a general electronic device. Components of electronic device 800 may include, but are not limited to, one or more processors or processing units 810, memory 820, storage device 830, one or more communication units 840, one or more input devices 850, and one or more output devices 860. Processing unit 810 may be a real or virtual processor and is capable of performing various processes according to programs stored in memory 820. In a multi-processor system, multiple processing units execute computer-executable instructions in parallel to increase the parallel processing capabilities of electronic device 800.

[0116] The electronic device 800 typically includes a plurality of computer storage media. Such media can be any available media accessible to the electronic device 800, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 820 can be a volatile memory (e.g., registers, cache, random access memory (RAM)), a non-volatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 830 can be a removable or non-removable medium and can include a machine-readable medium, such as a flash drive, a disk, or any other medium that can be used to store information and / or data and can be accessed within the electronic device 800.

[0117] The electronic device 800 may further include additional removable / non-removable, volatile / non-volatile storage media. Figure 8 As shown in FIG, a magnetic disk drive for reading from or writing to a removable, non-volatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to a bus (not shown) by one or more data media interfaces. The memory 820 may include a computer program product 825 having one or more program modules configured to perform various methods or actions of various embodiments of the present disclosure.

[0118] The communication unit 840 implements communication with other electronic devices via a communication medium. Additionally, the functions of the components of the electronic device 800 can be implemented as a single computing cluster or multiple computing machines that can communicate via a communication connection. Thus, the electronic device 800 can use a logical connection with one or more other servers, a network personal computer (PC), or another network node to perform operations in a networked environment.

[0119] The input device 850 may be one or more input devices, such as a mouse, keyboard, or trackball. The output device 860 may be one or more output devices, such as a display, a speaker, or a printer. The electronic device 800 may also communicate with one or more external devices (not shown) through the communication unit 840 as needed, such as a storage device, a display device, or the like, with one or more devices that allow a user to interact with the electronic device 800, or with any device that allows the electronic device 800 to communicate with one or more other electronic devices (e.g., a network card, a modem, etc.). Such communication may be performed via an input / output (I / O) interface (not shown).

[0120] According to an exemplary implementation of the present disclosure, a computer-readable storage medium is provided, on which computer-executable instructions are stored, wherein the computer-executable instructions are executed by a processor to implement the method described above. According to an exemplary implementation of the present disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the method described above.

[0121] Various aspects of the present disclosure are described herein with reference to flowcharts and / or block diagrams of methods, apparatuses, devices, and computer program products implemented according to the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable program instructions.

[0122] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, such that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device is generated that implements the functions / actions specified in one or more blocks in the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, where these instructions cause the computer, programmable data processing device, and / or other device to operate in a specific manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0123] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0124] The flow charts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple implementations of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a part for a module, program segment or instruction, and a part for a module, program segment or instruction comprises one or more executable instructions for realizing the logical function of the specification. In some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be realized by a special hardware-based system that performs the function or action of the specification, or can be realized by a combination of special hardware and computer instructions.

[0125] While various implementations of the present disclosure have been described above, the foregoing description is intended to be illustrative, not exhaustive, and not limited to the disclosed implementations. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described implementations. The terminology used herein is selected to best explain the principles of the implementations, their practical applications, or improvements to existing technologies, or to enable others skilled in the art to understand the various implementations disclosed herein.

Claims

1. A data analysis method, comprising: Determining predicted order data for the target object based on a temporal feature representation of historical order data of the target object, wherein the historical order data at least indicates an order generation time and an order generation quantity; Determining a supply strategy for the target object from a plurality of candidate supply strategies using a trained supply strategy determination model based on the time series feature representation and the feature representation corresponding to the predicted order data; as well as Determining the supply quantity of the target object based on the supply strategy, The supply strategy determination model is trained in the following way: Determining a predicted order data sample of the training object based on a temporal feature representation of the order data sample of the training object; Determining simulation data corresponding to a plurality of candidate supply strategy samples of the training object based on the predicted order data sample of the training object, wherein the simulation data indicates a turnover rate and a spot rate corresponding to each of the supply strategy samples; Based on the simulation data corresponding to the plurality of candidate supply strategy samples, and utilizing the simulation data constraints, determining at least one first supply strategy sample from the respective candidate supply strategy samples; Determining at least one second supply strategy sample from the plurality of candidate supply strategy samples using the supply strategy determination model based at least on the time series feature representation of the order data sample of the training object and the feature representation corresponding to the predicted order data sample; as well as The provisioning strategy determination model is trained based on the at least one first provisioning strategy sample and the at least one second provisioning strategy sample.

2. The method according to claim 1, wherein determining the predicted order data of the target object comprises: Determining, based on the time series feature representation, order data distribution of the target object within a given future time range; as well as Based on the order data distribution probability within the given future time range, the predicted order data of the target object is determined.

3. The method according to claim 1, further comprising: The multiple candidate supply strategies for the target object are determined based on the time series feature representation and the feature representation corresponding to the predicted order data.

4. The method of claim 1 , wherein the simulation data constraint indicates at least one of the following: The spot rate is not higher than a spot rate constraint parameter, and the spot rate constraint parameter is determined based on a preset basic spot rate and the turnover rate. The turnover rate is not higher than a turnover rate constraint parameter, and the turnover rate constraint parameter is determined based on a preset basic turnover rate.

5. The method according to claim 1 , wherein the provisioning policy of the target object is determined from a plurality of candidate provisioning policies, and determining the plurality of candidate provisioning policies comprises: Determining a plurality of target inventory quantities using a trained inventory forecasting model based on the time series feature representation and the feature representation corresponding to the predicted order data; Obtain the existing inventory quantity of the target object; as well as The plurality of candidate supply strategies are determined according to the existing inventory quantity and each target inventory quantity among the plurality of target inventory quantities.

6. The method according to claim 5, wherein the training process of the inventory forecasting model comprises: Determining a predicted order data sample of the training object based on a temporal feature representation of the order data sample of the training object; Determining a plurality of inventory quantities corresponding to the training object using the inventory forecasting model based on the time series feature representation of the order data sample and the predicted order data sample of the training object; Determining an inventory turnover rate forecast result for each of the plurality of inventory quantities in a given time window; Determine, based on the inventory turnover rate forecast result, an inventory turnover mean forecast result and an inventory turnover standard deviation forecast result for each inventory quantity in the given time window; Determining, based on the order data samples of the training object, the actual inventory turnover rate of each inventory quantity in a given time window; as well as The inventory forecasting model is trained based on the inventory turnover mean prediction result, the inventory turnover standard deviation prediction result, the actual inventory turnover rate and a preset inventory turnover rate target value.

7. A data analysis device comprising: A determination module configured to determine predicted order data of the target object based on a temporal feature representation of historical order data of the target object, wherein the historical order data at least indicates an order generation time and an order generation number; an execution module configured to determine a supply strategy for the target object from a plurality of candidate supply strategies using a trained supply strategy determination model based on the time series feature representation and the feature representation corresponding to the predicted order data; as well as A supply module is configured to determine the supply quantity of the target object based on the supply strategy The supply strategy determination model is trained in the following manner: Determining a predicted order data sample of the training object based on a temporal feature representation of the order data sample of the training object; Determining simulation data corresponding to a plurality of candidate supply strategy samples of the training object based on the predicted order data sample of the training object, wherein the simulation data indicates a turnover rate and a spot rate corresponding to each of the supply strategy samples; Based on the simulation data corresponding to the plurality of candidate supply strategy samples, and utilizing the simulation data constraints, determining at least one first supply strategy sample from the respective candidate supply strategy samples; Determining at least one second supply strategy sample from the plurality of candidate supply strategy samples using the supply strategy determination model based at least on the time series feature representation of the order data sample of the training object and the feature representation corresponding to the predicted order data sample; and The provisioning strategy determination model is trained based on the at least one first provisioning strategy sample and the at least one second provisioning strategy sample.

8. An electronic device comprising: at least one processing unit; as well as At least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions causing the electronic device to perform the method according to any one of claims 1 to 6 when executed by the at least one processing unit.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executable by a processor to implement the method according to any one of claims 1 to 6.

10. A computer program product comprising computer executable instructions, characterized in that: The computer executable instructions implement the method according to any one of claims 1 to 6 when executed by a processor.

Images