WO2018188402A1

WO2018188402A1 - Method and apparatus for predicting product demand

Info

Publication number: WO2018188402A1
Application number: PCT/CN2018/074769
Authority: WO
Inventors: 陈新杰; 赵志洪; 齐泉; 王巍
Original assignee: 华为技术有限公司
Priority date: 2017-04-12
Filing date: 2018-01-31
Publication date: 2018-10-18
Also published as: CN108694460A; CN108694460B

Abstract

A method and apparatus for predicting product demand, which at least resolve the problem with current methods for predicting product demand whereby the beneficial results for companies cannot be maximized. Said method comprises: the apparatus for predicting product demand acquiring product demand parameters (S201); the apparatus for predicting product demand inputting the product demand parameters into a pre-trained demand prediction model and generating the predicted volume of demand for the product in the following stage, wherein the pre-trained demand prediction model is acquired on the basis of asymmetric loss function training, the asymmetric loss function being the different functions predicting the loss from one more product and for predicting the loss from one less product (202).

Description

Product demand prediction method and device

The present application claims the priority of the Chinese Patent Application, the entire disclosure of which is hereby incorporated by reference.

Technical field

The present application relates to the field of artificial intelligence technology, and in particular, to a product demand prediction method and apparatus.

Background technique

Product demand forecasting is a key part of business operations and is used to guide the production and stocking of the company. Excessive product demand forecasts can lead to excessive inventory and increased inventory cost risk. Too small demand forecasts will result in low order fulfillment rates and reduced customer satisfaction. Therefore, reasonable product demand forecasting is particularly important.

At present, the mainstream product demand forecasting method is to establish a time series model and a predictive factor regression model based on historical demand and predictors related to future product demand, and output predicted values of future product demand. However, the predicted value of future product demand obtained by this method may deviate from the actual value of future product demand, so that the benefit of the enterprise cannot be maximized.

Therefore, how to provide a reasonable forecasting method for product demand makes the predicted value of future product demand closer to the actual value of future product demand, thus maximizing the benefit of the enterprise, which is an urgent problem to be solved.

Summary of the invention

The embodiments of the present application provide a product demand prediction method and device to solve at least the problem that the current product demand prediction method cannot maximize the benefit of the enterprise.

To achieve the above objective, the embodiment of the present application provides the following technical solutions:

In a first aspect, a method for predicting a product demand is provided, the method comprising: obtaining a demand parameter of a product; inputting a demand parameter of the product into a pre-trained demand forecasting model, and generating a predicted demand quantity of the product in a next stage, wherein The pre-trained demand forecasting model is based on an asymmetric loss function training that predicts that the loss caused by one more product is different from the predicted loss caused by one product. That is to say, the embodiment of the present application considers that in actual operation, it is inconsistent to predict that the loss caused by one product is less than the loss caused by predicting one product. For example, for products that need to purchase materials overseas, as a result of the need to add emergency air freight, etc., the loss caused by the lack of one product may be greater than the loss caused by one more product. For products that can be purchased at any time in the surrounding area, if the material cost is high, the loss caused by one more product may be greater than the loss caused by the lack of one product. Therefore, the embodiment of the present application predicts the demand of the next stage product generated according to the demand forecasting model trained in the scenario that the loss caused by predicting one more product is less than the predicted loss of one product. It is closer to the actual value of future product demand, which can maximize the benefits of the company.

In a possible design, the method further comprises: obtaining a defect loss ratio of the product, and training a plurality of training parameters of the demand prediction model of the product, wherein the defect loss ratio is used to represent a missing product The ratio of the loss to the loss caused by one more product; according to the plurality of training parameters and the ratio of the missing loss, the demand prediction model of the product is trained by minimizing the asymmetric loss function, and the pre-trained demand prediction model is obtained. That is to say, when performing the demand prediction model training, the embodiment of the present application not only considers the case that the loss caused by predicting one product is different from the loss caused by predicting one product, and also considers the loss caused by the lack of one product and one more. The ratio of the loss caused by the product, and the trained demand forecasting model obtained under the premise of the minimum loss, so the trained demand forecasting model is reasonable, and the next stage product generated based on the trained demand forecasting model is The predicted demand is closer to the actual value of future product demand, and the loss is the smallest, which can maximize the benefits of the enterprise.

In a possible design, the demand parameters of the product include: the actual demand of the product at the current stage and the actual demand of the product in the previous stage of the current stage; the plurality of training data includes: the historical stage of the product Actual demand and forecast demand. That is to say, in the specific implementation, the future demand of the product can be predicted based on the historical demand of the product and the predicted demand.

In one possible design, the demand forecasting model includes:

Indicates the predicted demand for the product in the t-th stage, y _t-1 represents the actual demand for the product in the t-1th stage, and y _t-2 represents the actual demand for the product in the t-2th stage.

α is a model factor. Based on this demand forecasting model, the future demand for the product can be predicted.

In one possible design, the asymmetric loss function includes:

among them,

Indicates that the value of i from 1 to t is summed;

Express when

The value is 1, otherwise it is 0; W indicates the ratio of the loss of the product. Based on the asymmetric loss function, it is possible to make predictions that the loss caused by one product is inconsistent with the prediction of the loss caused by one product.

In a possible design, the ratio of the loss of the product is obtained, including: when the current stage is not the initial stage, according to the out-of-stock status of the product in the previous stage of the current stage and the inventory status of the product in the current stage, The current period of the product's lack of loss ratio, the out-of-stock status includes out of stock or no out of stock; the inventory status includes less inventory, moderate inventory or inventory; when the current stage is the initial stage, the pre-configured The initial loss-to-loss ratio of the product is determined as the ratio of the current loss at the current stage. Based on this scheme, the product's defect-to-loss ratio can be flexibly adjusted to make the product in a more reasonable product state.

In a possible design, according to the out-of-stock status of the product in the previous stage of the current stage and the inventory status of the product in the current stage, the defect-to-loss ratio of the product at the current stage is determined, including: the previous one in the current stage. In the current stage, the out-of-stock status of the product is out of stock. When the inventory status of the product is low in the current stage, it is determined that the current loss ratio of the product is the first value, and the first value is a positive real number; In the previous stage, the out-of-stock status of the product is out of stock. When the inventory status of the product is moderate in the current stage, it is determined that the current loss ratio of the product is the second value, and the second value is a positive real number. In the previous stage of the current stage, the out-of-stock status of the product is out of stock. At the current stage, when the inventory status of the product is too much inventory, it is determined that the current loss ratio of the product is 0; the previous one in the current stage In the stage, the out-of-stock status of the product is no shortage. At the current stage, when the inventory status of the product is less inventory, it is determined that the current loss ratio of the product is the third value. The third value is a positive real number; in the previous stage of the current stage, the out-of-stock status of the product is no out-of-stock. At the current stage, the inventory status of the product is moderately in stock, and the current loss ratio of the product is determined to be 0. In the previous stage of the current stage, the out-of-stock status of the product is no out-of-stock. At the current stage, the inventory status of the product is too much inventory, and the current loss ratio of the product is determined to be the fourth value, the fourth value. Is a negative real number greater than -1. Based on the scheme, the out-of-stock status of the product stage can be shifted to the previous stage of the current stage to be out of stock, and the inventory status of the current stage product is a more reasonable product status with moderate inventory.

In one possible design, the initial loss-to-loss ratio for different products is configured as follows: For all products, the same initial loss-to-loss ratio is configured. Based on this method, the initial loss-to-loss ratio of the product can be configured simply and quickly.

In one possible design, the initial loss-to-loss ratio of different products is configured by configuring a preset initial loss-to-loss ratio for each of the preset proportions of products; Attributes of each of the proportioned products and attributes of each of the products other than the predetermined proportion of products, establishing an optimal proximity model for each of the predetermined proportions of the products, wherein The product in the optimal proximity model includes the product of the preset ratio of products that is closest to each of the products of the preset ratio; according to the optimal proximity model, the product of the preset ratio Each product other than the configuration has an initial loss-to-loss ratio, wherein the cost-to-loss ratio of each product in the optimal proximity model is the same. Based on this method, the initial loss-to-loss ratio of the product can be configured more accurately.

In a second aspect, a product demand forecasting device is provided, the product demand forecasting device having the function of implementing the above method. This function can be implemented in hardware or in hardware by executing the corresponding software. The hardware or software includes one or more modules corresponding to the functions described above.

A third aspect provides a product demand prediction apparatus, including: a processor, a memory, a bus, and a communication interface; the memory is configured to store a computer execution instruction, and the processor is connected to the memory through the bus, when the product demand prediction device In operation, the processor executes the computer-executed instructions stored by the memory to cause the product demand forecasting device to perform the product demand forecasting method of any of the first aspects above.

A fourth aspect, a computer readable storage medium for storing computer program instructions for use in a product demand forecasting apparatus, wherein when executed on a computer, causes the computer to perform any of the above first aspects Product demand forecasting method.

In a fifth aspect, there is provided a computer program product comprising instructions which, when run on a computer, cause the computer to perform the product demand prediction method of any of the above first aspects.

For the technical effects brought by any one of the second aspect to the fifth aspect, refer to the technical effects brought by different design modes in the first aspect, and details are not described herein again.

DRAWINGS

1 is a schematic structural diagram of hardware of a product demand prediction apparatus according to an embodiment of the present application;

2 is a schematic flowchart of a method for predicting a product demand according to an embodiment of the present application;

FIG. 3 is a schematic flowchart of a method for training a product demand prediction model according to an embodiment of the present application;

4 is a schematic diagram of transfer of a product state of a product according to an embodiment of the present application;

FIG. 5 is a schematic structural diagram of a product demand forecasting apparatus according to an embodiment of the present application;

FIG. 6 is a schematic structural diagram of another product demand prediction apparatus according to an embodiment of the present application.

detailed description

For ease of understanding, a brief description of some of the concepts related to the embodiments of the present application is first given for reference, as follows:

1) Asymmetric loss function

The asymmetric loss function is a function that assumes that the loss caused by predicting one more product is different from the loss caused by predicting that one product is less.

2), the ratio of loss to loss

The loss-to-loss ratio is used to characterize the ratio of the loss caused by the absence of one product to the loss caused by one more product. The exemplary loss-to-loss ratio can be expressed by the symbol W. Among them, you can define an asymmetric loss function, so that:

When W>0, the loss caused by the lack of one product is greater than the loss caused by one more product.

When W=0, the loss caused by the lack of one product is consistent with the loss caused by one more product.

When W < 0, the loss caused by the lack of one product is less than the loss caused by one more product.

3) Time series model

The time series model is a model for predicting the future using historical events.

4), the optimal close model (k-Nearest Neighbor, KNN)

KNN is based on the corresponding features, looking for models of the same product with the closest features.

5) Demand forecasting model

The demand forecasting model is a model that predicts product demand based on the demand parameters of the product.

The technical solutions in the embodiments of the present application will be described below with reference to the accompanying drawings in the embodiments of the present application. In the description of the present application, unless otherwise stated, "/" means the meaning of or, for example, A/B may represent A or B; "and/or" herein is merely an association describing the associated object. The relationship indicates that there may be three kinds of relationships, for example, A and/or B, which may indicate that there are three cases where A exists separately, A and B exist at the same time, and B exists separately. In addition, in the description of the present application, "a plurality" means two or more than two.

As shown in FIG. 1 , a hardware structure diagram of a product demand forecasting apparatus 10 provided by an embodiment of the present application includes at least one processor 101 , a communication bus 102 , a memory 103 , and at least one communication interface 104 . .

The processor 101 can be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more programs for controlling the execution of the program of the present application. integrated circuit.

Communication bus 102 can include a path for communicating information between the components described above.

The communication interface 104 uses a device such as any transceiver for communicating with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Networks (WLAN), etc. .

The memory 103 can be a Read-Only Memory (ROM) or other type of static storage device that can store static information and instructions, a Random Access Memory (RAM) or other type that can store information and instructions. The dynamic storage device can also be an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Compact Disc Read-Only Memory (CD-ROM) or other optical disc storage, and a disc storage device. (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or can be used to carry or store desired program code in the form of instructions or data structures and can be Any other media accessed, but not limited to this. The memory can exist independently and be connected to the processor via a bus. The memory can also be integrated with the processor.

The memory 103 is used to store application code for executing the solution of the present application, and is controlled by the processor 101 for execution. The processor 101 is configured to execute the application code stored in the memory 103, thereby implementing the product demand prediction method in the embodiment of the present application.

In a particular implementation, as an embodiment, processor 101 may include one or more CPUs, such as CPU0 and CPU1 in FIG.

In a particular implementation, as an embodiment, product demand forecasting device 10 may include multiple processors, such as processor 101 and processor 108 in FIG. Each of these processors can be a single-CPU processor or a multi-core processor. A processor herein may refer to one or more devices, circuits, and/or processing cores for processing data, such as computer program instructions.

In a specific implementation, as an embodiment, the product demand forecasting device 10 may further include an output device 105 and an input device 106. Output device 105 is in communication with processor 101 and can display information in a variety of ways. For example, the output device 105 can be a liquid crystal display (LCD), a light emitting diode (LED) display device, a cathode ray tube (CRT) display device, or a projector. Wait. Input device 106 is in communication with processor 101 and can accept user input in a variety of ways. For example, input device 106 can be a mouse, keyboard, touch screen device, or sensing device, and the like.

The above product demand forecasting device 10 can be a general purpose device or a dedicated device. In a specific implementation, the product demand prediction device 10 may be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, an embedded device, or a map. A device of similar structure in 1. The embodiment of the present application does not limit the type of the product demand prediction device 10.

As shown in FIG. 2 , a schematic flowchart of a product requirement prediction method provided by an embodiment of the present application includes the following steps:

S201. The product demand forecasting device obtains a demand parameter of the product.

Exemplarily, the demand parameter of the product may include: the actual demand quantity of the product in the current stage and the actual demand quantity of the product of the previous stage of the current stage; the plurality of training data may include: the actual demand quantity and the predicted demand quantity of the product in the historical stage. Of course, the demand parameter of the product may also include related factors such as holidays, and the plurality of training data may also include related factors such as holidays, which are not specifically limited in the embodiment of the present application.

S202. The product demand forecasting device inputs the demand parameter of the product into the pre-trained demand forecasting model to generate the predicted demand quantity of the product in the next stage, wherein the pre-trained demand forecasting model is obtained based on the asymmetric loss function training. The asymmetric loss function is a function that predicts that the loss caused by one more product is different from the loss predicted by one product.

The pre-trained demand forecasting model in the embodiment of the present application is a model for predicting product demand according to the demand parameter of the product, and is obtained based on the asymmetric loss function training.

Exemplarily, the demand forecasting model can be as shown in formula (1):

among them,

Indicates the predicted demand for the product in the t-th phase, y _t-1 represents the actual demand for the product in the t-1th stage, and y _t-2 represents the actual demand for the product in the t-2th stage.

α is a model factor.

Exemplarily, the asymmetric loss function can be as shown in equation (2):

among them,

Indicates that the value of i from 1 to t is summed;

Express when

The value is 1, otherwise it is 0; W is the ratio of the product's loss to loss.

Specifically, the process of training the demand prediction model based on the asymmetric loss function will be elaborated in the following embodiments, and details are not described herein again.

That is to say, the embodiment of the present application considers that in actual operation, it is inconsistent to predict that the loss caused by one product is less than the loss caused by predicting one product. For example, for products that need to purchase materials overseas, as a result of the need to add emergency air freight, etc., the loss caused by the lack of one product may be greater than the loss caused by one more product. For products that can be purchased at any time in the surrounding area, if the material cost is high, the loss caused by one more product may be greater than the loss caused by the lack of one product. Therefore, the embodiment of the present application predicts the demand of the next stage product generated according to the demand forecasting model trained in the scenario that the loss caused by predicting one more product is less than the predicted loss of one product. It is closer to the actual value of future product demand, which can maximize the benefits of the company.

The action of the product demand forecasting device in the above steps S201 and S202 can be performed by the processor 101 in the product demand forecasting device 10 shown in FIG. 1 by calling the application code stored in the memory 103, which is not used in this embodiment of the present application. Any restrictions.

Optionally, as shown in FIG. 3, the training method of the demand prediction model provided by the embodiment of the present application includes the following steps:

S301. The product demand forecasting device acquires a defect loss ratio of the product, and a plurality of training parameters of the demand prediction model for training the product.

Among them, the loss-to-loss ratio is used to characterize the ratio of the loss caused by the lack of one product to the loss caused by one more product. For details, refer to the above explanation of the concept, and details are not described herein again.

S302. The product demand forecasting device obtains the pre-trained demand forecasting model by minimizing the asymmetric loss function training product demand forecasting model according to the plurality of training parameters and the missing loss ratio.

Exemplarily, when the demand prediction model is as shown in formula (1), and the asymmetric loss function is as shown in formula (2), assuming that the current phase is the fourth phase, that is, t=3 in formula (2), The process of training the demand forecasting model can be as follows:

among them,

Hypothesis

then

and then,

Since the values of y ₁ , y ₂ , y ₃ and W are known, the asymmetry loss function is minimized to obtain the value of α. By incorporating the value of α into equation (1), a trained demand forecasting model can be obtained.

Of course, the embodiment of the present application is merely an exemplary example of the product demand forecasting device training demand forecasting model. Of course, other equipments may train the demand forecasting model and then provide the product demand forecasting device for use. This example does not specifically limit this.

When performing the demand prediction model training, the embodiment of the present application not only considers the case that the loss caused by predicting one product is different from the loss caused by predicting one product, and also considers the loss caused by the lack of one product and the loss caused by one more product. The ratio of the training, and the trained demand forecasting model obtained under the premise of the minimum loss, so the trained demand forecasting model is reasonable, based on the predicted demand of the next stage product generated by the trained demand forecasting model. It is closer to the actual value of future product demand and has the least loss, which can maximize the benefits of the company.

The action of the product demand forecasting device in the above steps S301 and S302 can be performed by the processor 101 in the product demand forecasting device 10 shown in FIG. 1 by calling the application code stored in the memory 103, which is not used in the embodiment of the present application. Any restrictions.

Optionally, in the product requirement prediction method provided by the embodiment of the present application, the product demand prediction device obtains the ratio of the loss-to-loss ratio of the product, which may specifically include:

When the current stage is not the initial stage, the product demand forecasting device determines the defect loss ratio of the current stage product according to the out-of-stock status of the product in the previous stage of the current stage and the inventory quantity status of the current stage product, and the shortage status includes the shortage. The goods are either out of stock; the inventory status includes less inventory, moderate inventory or more inventory.

When the current phase is the initial phase, the product demand forecasting device determines the initial loss-to-loss ratio of the pre-configured product as the current-stage loss-to-loss ratio.

The inventory status of the current stage product is defined according to the current inventory quantity of the product relative to the current inventory quantity of other products. For example, for different products, according to the inventory quantity, the inventory quantity below 30% is defined as less inventory, the inventory quantity is defined as 30%~70%, the inventory is moderate, and the inventory quantity is above 70%. There are many inventories, and the embodiment of the present application does not specifically limit how the inventory status is divided.

Among them, the out-of-stock status of the previous stage of the current stage and the inventory status of the current stage of the product can constitute six product statuses, as follows:

a), the out-of-stock status of the product in the previous stage of the current stage is out of stock, and the inventory status of the product in the current stage is less inventory;

b), the out-of-stock status of the product in the previous stage of the current stage is out of stock, and the inventory status of the product in the current stage is moderately in stock;

c), the out-of-stock status of the product in the previous stage of the current stage is out of stock, and the inventory status of the product in the current stage is more in stock;

d), the out-of-stock status of the product in the previous stage of the current stage is not out of stock, and the inventory status of the product at the current stage is less inventory;

e), the out-of-stock status of the product in the previous stage of the current stage is no shortage of goods, and the inventory status of the products in the current stage is moderately in stock;

f) The out-of-stock status of the product in the previous stage of the current stage is that there is no out-of-stock status. At the current stage, the inventory status of the product is more in stock.

The above six product states can be simply expressed as: a (with, less), b (with, moderate), c (with, more), d (none, less), e (none, moderate), f (none, many).

Among them, as shown in Figure 4, the most ideal product status is product status e (none, moderate), that is, the out-of-stock status of the product in the previous stage of the current stage is not out of stock, and the inventory status of the current stage is moderately stocked. . Therefore, in the embodiment of the present application, when the current stage is not the initial stage, if the product status of the current stage is not the product status e, it should be considered that the product status e can be reached in the next stage of the current stage. That is, as shown in Figure 4, the remaining product states should tend to shift to product state e.

Based on this, optionally, when the asymmetric loss function is the formula (2), the product demand forecasting device determines the current stage of the loss in the current stage according to the out-of-stock status of the previous stage of the current stage and the inventory status of the current stage product. Specific, can include:

When the product status is a (yes, less), it is determined that the ratio of the loss of the current stage product is the first value, and the first value is a positive real number, that is, W>0.

When the product status is b (yes, moderate), it is determined that the ratio of the loss of the product in the current stage is the second value, and the second value is a positive real number, that is, W>0.

When the product status is c (yes, many), it is determined that the current loss ratio of the product in the current stage is 0, that is, W=0.

When the product status is d (none, less), it is determined that the current loss ratio of the product in the current stage is a third value, and the third value is a positive real number, that is, W>0;

When the product status is e (none, moderate), it is determined that the current loss ratio of the product is 0.

When the product status is f (none, many), it is determined that the current loss ratio of the product in the current stage is the fourth value, and the fourth value is a negative real number greater than -1, that is, -1<W<0.

Based on this scheme, the ratio of the loss to the loss of the product can be flexibly adjusted, so that the system is in a more reasonable product state.

The rationality of the defect loss ratio W of the current stage product determined based on the above method is verified below.

According to formula (2), when W>0, the loss caused by the lack of one product is greater than the loss caused by one more product, indicating that the predicted amount of the product should be increased. In the embodiment of the present application, when the product status is a (yes, less), b (yes, moderate), and d (none, less), the determined W is W>0. Obviously, when the product status is a (yes, less), b (with, moderate), and d (none, less), it is reasonable to increase the predicted amount of the product.

According to formula (2), when W=0, the loss caused by the lack of one product is consistent with the loss caused by more than one product, indicating that the current forecast of the product is reasonable, without increasing the forecast of the product, nor reducing the product. Forecast amount. In the embodiment of the present application, when the product status is c (yes, many) and e (none, moderate), the determined W is W=0. Obviously, when the product status is c (yes, more) and e (none, moderate), it is reasonable that the predicted amount of the product does not need to be increased or decreased.

According to formula (2), when -1<W<0, the loss caused by the lack of one product is less than the loss caused by one more product, indicating that the predicted amount of the product should be reduced. In the embodiment of the present application, when the product status is f (none, many), the determined W is -1<W<0. Obviously, when the product status is f (none, many), it is reasonable to reduce the predicted amount of the product.

Exemplarily, assume that the first value = the second value = the third value = 1, that is, the loss caused by the lack of one product is double the loss caused by one more product; the fourth value = -0.5, that is, the lack of a product If the loss is half of the loss caused by one more product, the current-to-defect ratio of the current stage can be determined according to the state transition matrix as shown in Table 1, as follows:

Table I

Among them, in Table 1, the corresponding row is found according to the state of the product, and the ratio of the loss loss corresponding to the column with the loss-to-loss ratio of 1 in the row is the ratio of the loss of the current stage product to be determined.

For example, if the product status is a (yes, less), then as shown in Table 1, the ratio of the loss-to-loss ratio corresponding to the column with the loss-to-loss ratio of 1 in the row of the product status is 1, that is, the current stage product. The ratio of loss to loss is 1, that is, the loss caused by the lack of one product is double the loss caused by one more product, and the predicted amount of the product should be increased.

Or, for example, if the product status is b (yes, moderate), as shown in Table 1, the ratio of the loss-to-loss ratio corresponding to the column with the loss-to-loss ratio of 1 in the row of the product state is 1, that is, the current The loss-to-loss ratio of the stage product is 1, that is, the loss caused by the lack of one product is double the loss caused by one more product, and the predicted amount of the product should be increased.

Or, for example, if the product status is c (yes, more), as shown in Table 1, the ratio of the loss-to-loss ratio corresponding to the column with the loss-to-loss ratio of 1 in the row of the product state is 0, that is, the current The loss-to-loss ratio of the stage products is 0, that is, the loss caused by the lack of one product is consistent with the loss caused by one more product, and the predicted quantity of the product does not need to be increased or reduced.

Or, for example, if the product status is d (none, less), as shown in Table 1, the ratio of the loss-to-loss ratio corresponding to the column with the loss-to-loss ratio of 1 in the row of the product state is 1, that is, the current The loss-to-loss ratio of the stage product is 1, that is, the loss caused by the lack of one product is double the loss caused by one more product, and the predicted amount of the product should be increased.

Or, for example, if the product status is e (none, moderate), as shown in Table 1, the ratio of the loss-to-loss ratio corresponding to the column with the loss-to-loss ratio of 1 in the row of the product state is 0, that is, the current The loss-to-loss ratio of the stage products is 0, that is, the loss caused by the lack of one product is consistent with the loss caused by one more product, and the predicted quantity of the product does not need to be increased or reduced.

Or, for example, if the product status is f (none, many), as shown in Table 1, the ratio of the loss-to-loss ratio corresponding to the column with the loss-to-loss ratio of 1 in the row of the product state is -0.5, that is, At the current stage, the product's loss-to-loss ratio is -0.5, that is, the loss caused by the lack of one product is half of the loss caused by one more product, and the predicted amount of the product should be reduced.

Optionally, in the embodiment of the present application, the initial loss/loss ratio of different products may be configured as follows:

First, the cold start mode: that is, for all products, the same initial loss ratio is configured.

For example, for all products, an initial value of W=0 can be directly assigned, that is, the loss caused by the lack of one product is consistent with the loss caused by one more product; or, for all products, an initial value of W=1 can be directly assigned. That is, the loss caused by the lack of one product is double the loss caused by one more product; and so on. The initial loss-to-loss ratio may be an empirical value, or may be determined according to the attribute of the product. The embodiment of the present application does not specifically limit the initial loss-to-loss ratio. Among them, the attributes of the product may specifically include the number of suppliers, product prices, inventory levels, and the like.

Based on this method, the initial loss-to-loss ratio of the product can be configured simply and quickly.

Second, the hot start mode: that is, each of the preset proportions of the products is respectively configured with a preset initial loss ratio; the properties of each of the products according to the preset ratio and the pre The property of each product other than the proportioned product establishes an optimal closeness model for each of the preset proportions of products, wherein the optimal immediate model includes products other than the preset proportion of products The product with the closest product property to each of the products in the preset ratio; according to the optimal close model, the initial loss-to-loss ratio is configured for each product other than the preset ratio product, wherein the optimal immediate vicinity For the introduction of the model, refer to the above-mentioned concept explanation part, and details are not described herein again. In the embodiment of the present application, the defect loss ratio of each product in the optimal proximity model is the same. The attributes of the product may specifically include the number of suppliers, product prices, inventory levels, and the like.

For example, 10% of the products can be selected by the business experts to give the value of the loss-to-loss ratio W(i); furthermore, the optimal close-up model (KNN) can be established based on the attributes of 10% of the products and the attributes of the other 90% of the products. ), that is, looking for the most similar products in terms of attributes. Furthermore, for an additional 90% of the products, the initial loss-to-loss ratio can be configured according to the optimal close proximity model. Wherein, assuming that the product A in the other 90% of the products belongs to the optimal immediate model of the product B in the 10% product, the initial loss-to-loss ratio of the product A = the initial loss-to-loss ratio of the product B.

Based on this method, the initial loss-to-loss ratio of the product can be configured more accurately.

The solution provided by the embodiment of the present application is mainly introduced from the perspective of the product demand forecasting method for executing the product demand forecasting method. It can be understood that, in order to implement the above functions, the above-mentioned demand forecasting device includes a hardware structure and/or a software module corresponding to each function. Those skilled in the art will readily appreciate that the present application can be implemented in a combination of hardware or hardware and computer software in combination with the elements and algorithm steps of the various examples described in the embodiments disclosed herein. Whether a function is implemented in hardware or computer software to drive hardware depends on the specific application and design constraints of the solution. A person skilled in the art can use different methods to implement the described functions for each particular application, but such implementation should not be considered to be beyond the scope of the present application.

The embodiment of the present application may divide the function module by the product requirement prediction device according to the above method example. For example, each function module may be divided according to each function, or two or more functions may be integrated into one processing module. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of the module in the embodiment of the present application is schematic, and is only a logical function division, and the actual implementation may have another division manner.

For example, in the case of dividing the respective functional modules by using the respective functions, FIG. 5 shows a possible structural diagram of the product requirement prediction device 50 involved in the above embodiment, including: an obtaining module 501 and a generating module 503. The obtaining module 501 is configured to support the product demand forecasting device 50 to perform step S201 in FIG. 2; the generating module 503 is configured to support the product demand forecasting device 40 to perform step S202 in FIG. 2.

Optionally, as shown in FIG. 5, the product requirement prediction apparatus 50 provided by the embodiment of the present application further includes a training module 502. The obtaining module 501 is further configured to support the product demand forecasting device 50 to perform step S301 in FIG. 3; the training module 502 is configured to support the product demand forecasting device 40 to perform step S302 in FIG. 3.

Optionally, the obtaining 401 module is specifically used to: when the current stage is not the initial stage, determine the current period of the loss-to-loss ratio according to the out-of-stock status of the product in the previous stage of the current stage and the inventory status of the current stage product. The status of the goods includes out of stock or no out of stock; the inventory status includes less inventory, moderate inventory or large inventory; when the current stage is the initial stage, the initial loss ratio of the pre-configured product is determined as the current stage of the shortfall. Loss ratio.

Further, the module 401 is specifically used for: the out-of-stock status of the product in the previous stage of the current stage is out of stock, and the current inventory status of the product is less inventory, and the ratio of the loss of the product in the current stage is determined to be the first The first value is the positive real number; in the previous stage of the current stage, the out-of-stock status of the product is out of stock. When the inventory status of the current stage is the stock is moderate, the ratio of the loss-to-deposit ratio of the current stage is determined to be the second. The value, the second value is a positive real number; in the previous stage of the current stage, the out-of-stock status of the product is out of stock. When the inventory status of the current stage is too much inventory, it is determined that the current loss ratio of the product is 0; In the previous stage of the current stage, the out-of-stock status of the product is no out-of-stock. When the inventory status of the product in the current stage is low, the ratio of the loss of the product in the current stage is determined to be the third value, and the third value is the positive real number; The out-of-stock status of the product in the previous stage of the stage is that there is no shortage of goods. When the inventory status of the current stage is the inventory is moderate, the loss of the product at the current stage is determined. The ratio is 0; in the previous stage of the current stage, the out-of-stock status of the product is no out-of-stock. When the inventory status of the current stage is too much inventory, it is determined that the current loss ratio of the product is the fourth value, and the fourth value is A negative real number greater than -1.

Optionally, as shown in FIG. 5, the product requirement prediction device 50 further includes a configuration module 504. The configuration module 504 is configured to configure initial loss-to-loss ratios of different products by configuring the same initial loss-to-loss ratio for all products.

Alternatively, the configuration module 504 is configured to configure an initial loss-to-loss ratio of different products by configuring a preset initial loss-to-loss ratio for each of the products of the preset ratio; The attributes of each product in the product and the attributes of each product other than the preset proportion of products, the optimal closeness model is established for each of the preset proportions of the products, wherein the optimal closeness model The product including the product of the preset ratio is the closest to each of the products in the preset ratio; according to the optimal close model, each product is configured for the preset ratio The initial loss-to-loss ratio, in which the loss-to-loss ratio of each product in the optimal immediate model is the same.

All the related content of the steps involved in the foregoing method embodiments may be referred to the functional descriptions of the corresponding functional modules, and details are not described herein again.

In the case where the respective functional modules are divided in an integrated manner, FIG. 6 shows a possible structural diagram of the product demand prediction device 60 involved in the above embodiment. As shown in FIG. 6, the product demand forecasting device 60 includes a processing module 601. The processing module 601 is configured to perform operations performed by the obtaining module 501, the training module 502, the generating module 503, and the configuration module 504 in FIG. For details, refer to the embodiment of the embodiment shown in FIG. 5, and details are not described herein again.

In this embodiment, the product demand forecasting device is presented in the form of dividing each functional module corresponding to each function, or the product demand forecasting device is presented in a form that divides each functional module in an integrated manner. A "module" herein may refer to a particular ASIC, circuitry, processor and memory that executes one or more software or firmware programs, integrated logic circuitry, and/or other devices that provide the functionality described above. In a simple embodiment, those skilled in the art will appreciate that the product demand forecasting device 50 or the product demand forecasting device 60 may take the form shown in FIG. For example, the acquisition module 501, the training module 502, the generation module 503, and the configuration module 504 in FIG. 5 can be implemented by the processor 101 and the memory 103 of FIG. Specifically, the obtaining module 501, the training module 502, the generating module 503, and the configuration module 504 can be executed by calling the application code stored in the memory 103 by the processor 101, which is not limited in this embodiment. Alternatively, for example, the processing module 601 in FIG. 6 may be implemented by the processor 101 and the memory 103 of FIG. 1. Specifically, the processing module 601 may be executed by the processor 101 calling the application code stored in the memory 103. The embodiment of the present application does not impose any limitation on this.

The product demand prediction device provided by the embodiment of the present invention can be used to perform the foregoing product requirement prediction method. Therefore, the technical effects that can be obtained by reference to the foregoing method embodiments are not described herein.

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using a software program, it may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in accordance with embodiments of the present invention are generated in whole or in part. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be from a website site, computer, server or data center Transmission to another website site, computer, server or data center via wired (eg coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (eg infrared, wireless, microwave, etc.). The computer readable storage medium can be any available media that can be accessed by a computer or a data storage device that includes one or more servers, data centers, etc. that can be integrated with the media. The usable medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium (such as a solid state disk (SSD)) or the like.

Although the present application has been described herein in connection with the various embodiments, those skilled in the art can Other variations of the disclosed embodiments are achieved. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. A single processor or other unit may fulfill several of the functions recited in the claims. Certain measures are recited in mutually different dependent claims, but this does not mean that the measures are not combined to produce a good effect.

Although the present invention has been described in connection with the specific features and embodiments thereof, it is obvious that various modifications and combinations may be made without departing from the spirit and scope of the application. Accordingly, the description and drawings are to be regarded as It will be apparent to those skilled in the art that various modifications and changes can be made in the present application without departing from the scope of the application. Thus, it is intended that the present invention cover the modifications and variations of the present invention.

Claims

A product demand forecasting method, characterized in that the method comprises:

Obtain the demand parameters of the product;

Entering the demand parameter of the product into the pre-trained demand forecasting model to generate a predicted demand quantity of the product in the next stage, wherein the pre-trained demand forecasting model is obtained based on the asymmetric loss function training. The asymmetric loss function is a function that predicts that the loss caused by one more product is different from the loss caused by predicting that one product is less.
The method of claim 1 further comprising:

Obtaining a ratio of a loss-to-loss ratio of the product, and training a plurality of training parameters of a demand prediction model of the product, wherein the loss-to-loss ratio is used to characterize a loss caused by the lack of one product and a loss caused by one more product ratio;

And pre-training the demand prediction model by training the demand prediction model of the product by minimizing the asymmetric loss function according to the plurality of training parameters and the loss-to-loss ratio.
The method according to claim 2, wherein the demand parameter of the product comprises: an actual demand amount of the product in the current stage and an actual demand quantity of the product in the previous stage of the current stage;

The plurality of training data includes: an actual demand quantity and a predicted demand quantity of the product in the historical stage.
The method of claim 3, wherein the demand prediction model comprises:

Indicates the predicted demand for the product in the t-th phase, y t-1 represents the actual demand of the product in the t-1th stage, and y t-2 represents the actual demand of the product in the t-2th stage. ,
α is a model factor.
The method of claim 4 wherein said asymmetric loss function comprises:

among them,
Indicates that the value of i from 1 to t is summed;
Express when
The value is 1, otherwise it is 0; W represents the ratio of the loss of the product.
The method according to any one of claims 2 to 5, wherein the obtaining a loss-to-loss ratio of the product comprises:

When the current stage is not the initial stage, determining the defect loss ratio of the product in the current stage according to the out-of-stock status of the product in the previous stage of the current stage and the inventory status of the product in the current stage. The out-of-stock status includes out-of-stock or out-of-stock; the inventory status includes less inventory, moderate inventory, or large inventory;

When the current phase is the initial phase, the initial loss-to-loss ratio of the pre-configured product is determined as the ratio of the loss of the product of the current phase.
The method according to claim 6, wherein said determining said current stage according to said stock-out status of said product in a previous stage of said current stage and said stock quantity status of said product in said current stage The ratio of loss to loss of the product, including:

In the previous stage of the current stage, the out-of-stock status of the product is out of stock, and in the current stage, when the inventory quantity status of the product is less inventory, determining the ratio of the loss-to-loss ratio of the product in the current stage is a first value, the first value being a positive real number;

In the previous stage of the current stage, the out-of-stock status of the product is out of stock, and when the inventory status of the product in the current stage is moderately in stock, the ratio of the loss of the products in the current stage is determined. a second value, the second value being a positive real number;

In the previous stage of the current stage, the out-of-stock status of the product is out of stock, and in the current stage, when the inventory quantity status of the product is in stock, determining the ratio of the loss-to-loss ratio of the product in the current stage is 0;

In the previous stage of the current stage, the out-of-stock status of the product is no out-of-stock. In the current stage, when the inventory quantity status of the product is less inventory, the ratio of the loss-to-loss ratio of the product in the current stage is determined as a third value, the third value being a positive real number;

In the previous stage of the current stage, the out-of-stock status of the product is no shortage, and the inventory status of the product in the current stage is that the inventory is moderate, and the ratio of the loss of the products in the current stage is determined. Is 0;

In the previous stage of the current stage, the out-of-stock status of the product is no out-of-stock. When the inventory status of the product in the current stage is a large inventory, it is determined that the defect loss ratio of the product in the current stage is A fourth value, the fourth value being a negative real number greater than -1.
The method according to claim 6 or 7, wherein the initial loss-to-loss ratio of the different products is configured by: configuring the same initial loss-to-loss ratio for all products;

Alternatively, configure the initial loss-to-loss ratio for different products as follows:

Pre-set initial loss-to-loss ratios are configured for each of the preset proportions of products;

Establishing the most for each of the preset proportions of products according to the attributes of each of the preset proportions of products and the attributes of each of the products other than the preset proportions of products Preferably, the optimal proximity model includes a product in the product other than the preset ratio of products that is closest to each of the predetermined proportions of products;

According to the optimal proximity model, an initial loss-to-loss ratio is configured for each product other than the predetermined proportion of products, wherein each product in the optimal proximity model has the same loss-to-loss ratio.
A product demand prediction device, the device comprising: an acquisition module and a generation module;

The obtaining module is configured to acquire a demand parameter of the product;

The generating module is configured to input a demand parameter of the product into a pre-trained demand forecasting model, and generate a predicted demand quantity of the product in a next stage, wherein the pre-trained demand forecasting model is based on an asymmetric Obtained by the loss function training, the asymmetric loss function is a function that predicts that the loss caused by one more product is different from the loss caused by predicting one less product.
The device according to claim 9, wherein said device further comprises a training module;

The obtaining module is further configured to acquire a ratio of a loss-to-loss ratio of the product, and a plurality of training parameters for training a demand prediction model of the product, wherein the ratio of the missing loss is used to represent a loss caused by the lack of a product The ratio of losses caused by one more product;

The training module is configured to train the demand prediction model of the product by minimizing the asymmetric loss function according to the plurality of training parameters and the loss ratio, to obtain the pre-trained demand prediction model .
The apparatus according to claim 10, wherein the demand parameter of the product comprises: an actual demand amount of the product in the current stage and an actual demand quantity of the product in the previous stage of the current stage;

The plurality of training data includes: an actual demand quantity and a predicted demand quantity of the product in the historical stage.
The device according to claim 11, wherein the demand prediction model comprises:

Indicates the predicted demand for the product in the t-th phase, y t-1 represents the actual demand of the product in the t-1th stage, and y t-2 represents the actual demand of the product in the t-2th stage. ,
α is a model factor.
The apparatus of claim 12 wherein said asymmetric loss function comprises:

among them,
Indicates that the value of i from 1 to t is summed;
Express when
The value is 1, otherwise it is 0; W represents the ratio of the loss of the product.
The device according to any one of claims 10 to 13, wherein the obtaining module is specifically configured to:

When the current stage is not the initial stage, determining the defect loss ratio of the product in the current stage according to the out-of-stock status of the product in the previous stage of the current stage and the inventory status of the product in the current stage. The out-of-stock status includes out-of-stock or out-of-stock; the inventory status includes less inventory, moderate inventory, or large inventory;

When the current phase is the initial phase, the initial loss-to-loss ratio of the pre-configured product is determined as the ratio of the deficit loss of the current phase.
The device according to claim 14, wherein the obtaining module is specifically configured to:

In the previous stage of the current stage, the out-of-stock status of the product is out of stock, and in the current stage, when the inventory quantity status of the product is less inventory, determining the ratio of the loss-to-loss ratio of the product in the current stage is a first value, the first value being a positive real number;

In the previous stage of the current stage, the out-of-stock status of the product is out of stock, and when the inventory status of the product in the current stage is moderately in stock, the ratio of the loss of the products in the current stage is determined. a second value, the second value being a positive real number;

In the previous stage of the current stage, the out-of-stock status of the product is out of stock, and in the current stage, when the inventory quantity status of the product is in stock, determining the ratio of the loss-to-loss ratio of the product in the current stage is 0;

In the previous stage of the current stage, the out-of-stock status of the product is no out-of-stock. In the current stage, when the inventory quantity status of the product is less inventory, the ratio of the loss-to-loss ratio of the product in the current stage is determined as a third value, the third value being a positive real number;

In the previous stage of the current stage, the out-of-stock status of the product is no shortage, and the inventory status of the product in the current stage is that the inventory is moderate, and the ratio of the loss of the products in the current stage is determined. Is 0;

In the previous stage of the current stage, the out-of-stock status of the product is no out-of-stock. When the inventory status of the product in the current stage is a large inventory, it is determined that the defect loss ratio of the product in the current stage is A fourth value, the fourth value being a negative real number greater than -1.
The device according to claim 14 or 15, wherein the device further comprises a configuration module;

The configuration module is configured to configure an initial loss-to-loss ratio of different products by configuring the same initial loss-to-loss ratio for all products;

Alternatively, the configuration module is configured to configure an initial loss ratio of different products by:

Pre-set initial loss-to-loss ratios are configured for each of the preset proportions of products;

Establishing the most for each of the preset proportions of products according to the attributes of each of the preset proportions of products and the attributes of each of the products other than the preset proportions of products Preferably, the optimal proximity model includes a product in the product other than the preset ratio of products that is closest to each of the predetermined proportions of products;

According to the optimal proximity model, an initial loss-to-loss ratio is configured for each product other than the predetermined proportion of products, wherein each product in the optimal proximity model has the same loss-to-loss ratio.
A product demand forecasting device, comprising: a processor, a memory, a bus, and a communication interface;

The memory is configured to store a computer execution instruction, the processor is connected to the memory through the bus, and when the product demand prediction device is in operation, the processor executes the computer execution instruction stored in the memory, The product demand forecasting device is configured to perform the product demand forecasting method according to any one of claims 1-8.