Object supply prediction method, prediction device and prediction system

Abstract

The invention provides an object supply prediction method, prediction device and prediction system. The prediction method comprises the steps of obtaining related information of a target object, wherein the related information comprises historical time sequence supply data of the target object at a supply end in a historical time period, historical event data of the target object at a use end in the historical time period, and preset event feature data of the target object at the use end in a preset time period; determining predicted supply feature data of the target object in a preset time period at least based on the historical time sequence supply data; determining predicted supply offset feature data of the target object in a preset time period at least based on the historical time sequence supply data, the historical event data and preset event data; and based on the predicted supply feature data and the predicted supply offset feature data, determining predicted time sequence supply data of the target object in a preset time period at the supply end.

Description

Technical Field

[0001] The present disclosure relates to the field of supply prediction, and particularly to a method, apparatus, and system for predicting the supply volume of an object, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] In order to reduce warehousing costs, reduce material waste, and increase material turnover efficiency, the supply side will have a link for estimating the material shipment volume. By estimating the future shipment level of materials, it is possible to know how much material may be shipped at what time point, so that the corresponding amount of materials can be pre-stored in the warehouse, which can avoid blindly hoarding materials in the warehouse and thus effectively reduce the inventory level of the warehouse. However, in industries such as the catering industry and manufacturing, after many materials are shipped from the supply side (such as a warehouse) to the sales side (such as a physical store), they need to be further processed at the sales side to become the final product. When there is a promotion or other situation, there will be a lead time for stocking at the sales side. The existence of the lead time for stocking at the sales side causes the peak of the shipment volume at the supply side to be earlier than the peak brought about by the promotion at the sales side. This shift results in the supply side being unable to accurately estimate the material shipment volume before and after the promotion at the sales side.

[0003] Since the shipment volume at the supply side under promotion is very large, the deviation in the estimated shipment volume caused by advance stocking leads to a series of problems such as difficulties in inventory turnover and increased replenishment costs. The above-mentioned situation where the peak of the shipment is earlier than the holiday or promotion time point, that is, the situation of early shipment, belongs to the problem of kurtosis shift (specifically, kurtosis forward shift). The existence of the kurtosis shift problem leads to an inability to accurately estimate the shipment volume at the supply side, thus affecting the determination of the inventory level of the warehouse at the supply side, and further causing problems such as difficulties in inventory turnover. Therefore, accurately determining the time point of kurtosis shift (i.e., the time point when the situation of kurtosis shift starts) and the kurtosis shift period (i.e., the time period during which the situation of kurtosis shift lasts) is the key to accurate time series prediction.

[0004] The methods described in this section are not necessarily methods that have been previously envisioned or adopted. Unless otherwise specified, no method described in this section should be considered prior art solely because it is included in this section. Similarly, unless otherwise specified, the problems mentioned in this section should not be considered to have been recognized in any prior art. Summary of the Invention

[0005] The present disclosure provides a method, apparatus, and system for predicting the supply volume of an object, an electronic device, a computer-readable storage medium, and a computer program product.

[0006] According to one aspect of the present disclosure, a method for predicting the supply quantity of an object is provided, including: obtaining relevant information of the target object, where the relevant information includes historical sequential supply data of the target object at the supply end within a historical time period, historical event data of the target object at the usage end within the historical time period, and preset event feature data of the target object at the usage end within a preset time period; determining, at least based on the historical sequential supply data, predicted supply feature data of the target object within the preset time period; determining, at least based on the historical sequential supply data, the historical event data, and the preset event data, predicted supply offset feature data of the target object within the preset time period; and determining, based on the predicted supply feature data and the predicted supply offset feature data, predicted sequential supply data of the target object at the supply end within the preset time period.

[0007] According to another aspect of the present disclosure, a prediction system is provided. The prediction system is used for predicting the supply quantity of a target object, and the target object has relevant information, where the relevant information includes: historical sequential supply data of the target object at the supply end within a historical time period, historical event data of the target object at the usage end within the historical time period, and preset event data of the target object at the usage end within a preset time period. The prediction system includes: a sequential prediction module configured to determine, at least based on the historical sequential supply data, predicted supply feature data of the target object within the preset time period; a kurtosis offset prediction module configured to determine, at least based on the historical sequential supply data, the historical event data, and the preset event data, predicted supply offset feature data of the target object within the preset time period; and an output module configured to determine, based on the predicted supply feature data and the predicted supply offset feature data, predicted sequential supply data of the target object at the supply end within the preset time period.

[0008] According to another aspect of the present disclosure, there is provided an apparatus for object supply quantity, including: an acquisition module configured to acquire relevant information of a target object, where the relevant information includes historical sequential supply data of the target object at the supply end within a historical time period, historical event data of the target object at the usage end within the historical time period, and preset event data of the target object at the usage end within a preset time period; a first determination module configured to determine predicted supply characteristic data of the target object within the preset time period based at least on the historical sequential supply data; a second determination module configured to determine predicted supply offset characteristic data of the target object within the preset time period based at least on the historical sequential supply data, the historical event data, and the preset event data; and a third determination module configured to determine predicted sequential supply data of the target object at the supply end within the preset time period based on the predicted supply characteristic data and the predicted supply offset characteristic data.

[0009] According to another aspect of the present disclosure, there is provided an electronic device, including: a memory, a processor, and a computer program stored on the memory, where the processor is configured to execute the computer program to implement the steps of the construction method and / or the prediction method according to the present disclosure.

[0010] According to another aspect of the present disclosure, there is provided a computer-readable storage medium having a computer program stored thereon, where the computer program, when executed by a processor, implements the steps of the construction method and / or the prediction method according to the present disclosure.

[0011] According to another aspect of the present disclosure, there is provided a computer program product including a computer program, where the computer program, when executed by a processor, implements the steps of the construction method and / or the prediction method according to the present disclosure.

[0012] In one or more embodiments of the present disclosure, by determining supply characteristic data within a preset time period based at least on historical sequential data, and predicting supply offset characteristic data within the preset time period in the case of kurtosis offset based on historical sequential data and event data, and then determining predicted sequential data based on the supply characteristic data and the supply offset characteristic data, it is possible to predict the kurtosis offset situation at the supply end, so that the sequential data can still be accurately predicted when the kurtosis offset occurs. In addition, in the case where there is no kurtosis forward shift, the above-mentioned implementation still maintains a good accuracy rate in predicting the sequential data of the target object. Accurately predicting sequential data can reduce the warehousing cost at the supply end, reduce material scrap, and increase the material turnover efficiency.

[0013] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become readily apparent from the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The drawings exemplarily illustrate embodiments and constitute a part of the specification, and are used together with the written description of the specification to explain the exemplary embodiments of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of the claims. In all the drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0015] Figure 1 Shows a structural block diagram of a prediction system according to some exemplary embodiments of the present disclosure;

[0016] Figure 2 Shows a flowchart of a prediction method according to some exemplary embodiments of the present disclosure;

[0017] Figure 3 Shows a prediction result graph using the traditional prediction system TCN;

[0018] Figure 4 Shows a prediction result graph using the prediction system or prediction method of the present disclosure; and

[0019] Figure 5 Shows a structural block diagram of a prediction device according to the present disclosure.

[0020] Figure 6 Shows a block diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] The following describes exemplary embodiments of the present disclosure with reference to the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted below.

[0022] In the present disclosure, unless otherwise specified, the terms "first", "second", etc. are used to describe various elements and are not intended to limit the positional relationship, temporal relationship or importance relationship of these elements. Such terms are only used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of the element, and in certain cases, based on the context description, they may also refer to different instances.

[0023] In the description of the various examples in this disclosure, the terms used are for the purpose of describing specific examples only and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically defined, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one and all possible combinations of the listed items.

[0024] In the related art, for the estimation of the shipment volume of materials, relatively classic linear models (such as Arima, Prophet) or deep learning models (such as Transformer, TCN) are generally used to train the historical material shipment data and date features. If there is a product promotion in the future, the promotion volume will be taken as a feature and incorporated into the model to enable the model to improve the estimated shipment volume of materials during product promotion. However, the size of the lead time for stock preparation depends on factors such as the type of material, the size of the promotion, holidays, and material demand. The current models can only extract the rules of historical shipment volumes and cannot accurately predict the time point when early stock preparation occurs and the size of the lead time for stock preparation. In addition, the size of the promotion, holidays, and material demand among the above-mentioned key factors belong to future information. Current models, such as Prophet, LightGBM, TCN, etc., can only construct lag features and feature stacking when receiving future information, and this method will cause the problem of dimensional explosion, increasing the number of model parameters and training difficulty, resulting in problems such as excessive consumption of computing resources and reduced computing efficiency. In addition, when early stock preparation is carried out at the selling end, the shipment rules at the supply end in the time periods before and after early stock preparation will be quite different from usual. For example, the shipment volume at the supply end at the time point of early stock preparation will be extremely high, but the shipment volume at subsequent time points will gradually decrease. The shipment volume rules at the supply end during the early stock preparation period that occurs at the selling end also need to be learned from history, and rule information needs to be extracted from historical time periods when early stock preparation also occurred, but this kind of information is not much, and both promotion and early shipment are low-probability events. The current shipment volume prediction models extract general historical rules and cannot differentially extract the rules of low-probability events.

[0025] To solve the above problems, the present disclosure provides a prediction method capable of accurately predicting the kurtosis shift at the supply end. By determining the supply characteristic data within a preset time period based at least on historical time series data, and predicting the supply offset characteristic data within the preset time period in the case of a kurtosis shift based on the historical time series data and event data, and then determining the predicted time series data based on the supply characteristic data and the supply offset characteristic data, it is possible to differentially extract the time series characteristic law in the case of a kurtosis shift, so as to accurately predict the kurtosis shift situation at the supply end (including the kurtosis shift time point, the kurtosis shift period, and the time series characteristics within the kurtosis shift period). In addition, in the case where there is no kurtosis advance, the above implementation still maintains a good accuracy rate in predicting the time series characteristic data of the target object.

[0026] In this article, the "supply end" refers to the providing end of the target object. The supply end provides the target object to the "using end", and the target object is used and consumed at the using end. The "target object" may include materials. For example, the material is the raw material stored at the supply end (such as a warehouse), and after it is shipped, it needs to be reprocessed at the using end (i.e., the selling end) to become the final product. In addition, the "target object" may also be an object with a kurtosis shift (specifically, kurtosis advance) problem in prediction, such as live broadcast traffic, bandwidth, electricity, etc. The "time series supply data" refers to the supply volume data of the target object changing over time within a period of time. For example, when the target object is a material, the time series supply data refers to the data of the shipment volume of the material changing over time within a period of time. The "historical time series supply data" refers to the supply volume data (such as shipment volume data) of the target object changing over time before the preset time period. The "predicted time series supply data" refers to the supply volume data (such as shipment volume data) of the target object changing over time within the preset time period. The "historical event data" includes event data used to describe events that can change the time series supply data of the target object occurring at the using end before the preset time period, such as promotional activities, holidays, etc. The "preset event data" includes event data used to describe events that can change the time series supply data of the target object occurring at the using end within the preset time period. For example, when a promotional activity is carried out at the selling end, the selling end will purchase materials from the supply end in advance before the time point of the promotion. The fact that the selling end stocks up in advance before the time point of the promotion is called "stocking up in advance", that is, kurtosis shift (specifically, kurtosis advance). Correspondingly, the supply end will ship in advance in the case of the selling end stocking up in advance, resulting in a peak shipment volume, and this peak shipment volume is ahead of the time point of the promotion. Among them, the "kurtosis shift time point" refers to the starting time point of the kurtosis shift, and the "kurtosis shift period" refers to the time period during which the kurtosis shift occurs.

[0027] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0028] Figure 1 FIG. shows a structural block diagram of a prediction system 1000 according to some exemplary embodiments of the present disclosure. The prediction system 1000 is used to predict the predicted time-series supply data of a target object at a supply end within a preset time period. Among them, the target object at least has historical time-series supply data of the target object at the supply end within a historical time period, historical event data of the target object at the usage end within the historical time period, and preset event data of the target object at the usage end within the preset time period.

[0029] In some embodiments, the historical event data may include at least one of holiday time point features, distance from holiday time length features, and promotional activity features within the historical time period. The preset event data may include at least one of holiday time point features, distance from holiday time length features, and promotional activity features within the preset time period. Among them, the above holiday time point features, distance from holiday time length features, and promotional activity features can all refer to encoded feature data for easy input into machine learning and other models. For example, the holiday time point feature can be a one-hot feature encoded in binary. For example, from May 1st to May 5th is the Labor Day holiday. At this time, a feature L can be constructed to represent Labor Day. The feature L from May 1st to May 5th is 1, and the feature L at other time points is 0. Similarly, one-hot encoded features can be constructed for holidays such as the Mid-Autumn Festival, the Dragon Boat Festival, the National Day, and the Spring Festival. The distance from holiday time length feature can represent the distance time length from a certain time point to a specific holiday. For example, a feature B can be constructed to represent the time length from Labor Day. The feature B from May 1st to May 5th is 0, and the corresponding features B for April 29th and April 30th are 2 and 1 respectively. The corresponding features B for May 6th and May 7th are -1 and -2 respectively. Therefore, feature B>0 represents that the time point is before Labor Day, feature B = 0 represents that the time point is during Labor Day, and feature B<0 represents that the time point is after Labor Day. By analogy, the distance from holiday time length features can be constructed for the Mid-Autumn Festival, the Dragon Boat Festival, the National Day, the Spring Festival, etc. In some examples, if the estimation granularity of the prediction system is not in days, but in weeks or months, the holiday time point features and the distance from holiday time length features can be converted into weeks or months through the following formula (1):

[0030]

[0031] wherein, hol i t is the holiday time point feature or the distance from holiday time length feature of the i-th holiday at time point t, and D is the number of days in a week or the number of days in a certain month.

[0032] The promotion activity characteristics can be the expected promotion quantity characteristics or the number of promotion activities. For example, on January 1st, there are 3 type C activities, 0 type D activities, and 1 type E activity; on January 2nd, there are 0 type C activities, 0 type D activities, and 0 type E activities. Then, the promotion activity characteristics can be constructed as [3, 0, 1] and [0, 0, 0] respectively.

[0033] As Figure 1 shown, the prediction system 1000 may include a time series prediction module 1100, a kurtosis deviation prediction module 1200, and an output module 1300.

[0034] The time series prediction module 1100 is used to extract the historical time series rules of the supply side (e.g., the shipping rules), and is configured to determine the predicted supply characteristic data of the target object within a preset time period based at least on the historical time series supply data.

[0035] The kurtosis deviation prediction module 1200 is used to obtain the kurtosis deviation rules of the target object at the supply side in the case of kurtosis deviation (e.g., advance stockpiling) (including the kurtosis deviation time point, the kurtosis deviation period, and the time series characteristics within the kurtosis deviation period, etc.) (e.g., the advance shipping rules), and can be configured to determine the predicted supply deviation characteristic data of the target object within a preset time period based at least on the historical time series supply data, the historical event data, and the preset event data.

[0036] The output module 1300 is configured to determine the predicted time series supply data of the target object at the supply side within a preset time period based on the predicted supply characteristic data and the predicted supply deviation characteristic data.

[0037] The above embodiments can differentially extract the time series characteristic rules in the case of kurtosis deviation, so as to accurately predict the situation of kurtosis deviation at the supply side (including the kurtosis deviation time point, the kurtosis deviation period, and the time series characteristics within the kurtosis deviation period). In addition, in the case of no kurtosis advance, the above embodiments still maintain a good accuracy rate in predicting the time series supply data of the target object. Accurately predicting the time series data can reduce the warehousing cost at the supply side, reduce material scrapping, and increase the material turnover efficiency.

[0038] Among them, the predicted supply feature data represents the supply volume time series of the target object within a preset time period predicted by the time series prediction module 1100 (i.e., a general, common, or conventional supply volume time series estimated at least based on historical time series supply data, such as a conventional shipment volume series). The supply volume time series can be a data series in which supply volume data is arranged in chronological order, and it is the feature data output by the model in the time series prediction module (i.e., a tensor at each time point within the preset time period, that is, a multi-dimensional array), and can be converted into the supply volume value of the target object within the preset time period through a preset rule. The time series prediction module 1100 may include a second feature encoder 1110 and a second feature decoder 1120. The second feature encoder 1110 is used to at least encode the historical time series supply data to output the historical supply feature encoding of the target object within the historical time period (for example, the historical shipment volume feature encoding, that is, the historical shipment volume pattern). The second feature decoder 1120 is configured to decode the historical supply feature encoding to output the predicted supply feature data of the target object within the preset time period, that is, to generate a corresponding tensor within the preset time period by passing the historical supply feature encoding through the decoder. The structures of the above encoder and decoder can well extract the features of the data and generate corresponding tensors within the preset time period, thereby improving the subsequent prediction accuracy.

[0039] The predicted supply offset feature data represents the supply volume time series of the target object within the kurtosis offset cycle within a preset time period predicted by the kurtosis offset prediction module (for example, the shipment volume series after considering early shipments), which is the feature data output by the model in the kurtosis offset prediction module (i.e., a tensor at each time point within the preset time period, that is, a multi-dimensional array), and can be converted into the early supply volume value of the target object within the preset time period through a preset rule. The kurtosis offset prediction module 1200 may include: a first feature encoder 1210 and a first feature decoder 1220. Among them, the first feature encoder 1210 is configured to at least encode the historical time series supply data and historical event data to output the historical supply offset feature encoding of the target object within the historical time period, for example, the historical early shipment pattern in the case of an early stock preparation period. The first feature decoder 1220 is configured to decode the historical supply offset feature encoding to output the first supply offset feature data of the target object within the preset time period (that is, to generate a corresponding tensor within the preset time period by passing the historical supply offset feature encoding through the decoder) for determining the predicted supply offset feature data. As Figure 1 shown, the first feature decoder may include, for example, a plurality of fully connected layers 1221 connected to the first feature encoder 1210 to generate corresponding tensors F within the preset time periods T1, T2, T3, T4 T1 、F T2 、F T3 、FT4 The structure of the above codec can well extract the features of the data and generate corresponding tensors within a preset time period, thereby improving the subsequent prediction accuracy.

[0040] In some embodiments, the above first feature encoder, first feature decoder, second feature encoder, and second feature decoder may use any existing encoder and decoder structures, such as TCN, LSTM, Transformer, etc.

[0041] The above structure with a double-layer encoder can well extract the general law of time-series supply data and the time-series offset law during the kurtosis offset period (for example, the law of early shipment volume), so that the model can still give a very high prediction accuracy when the kurtosis offset occurs.

[0042] In some embodiments, the kurtosis offset prediction module 1200 may further include: an event feature extraction model 1230. The event feature extraction model 1230 is configured to at least extract features from preset event data to output second supply offset feature data of the target object within a preset time period for determining the predicted supply offset feature data. Thus, efficient extraction of future information can be achieved, helping the prediction system to more accurately determine the situation of kurtosis offset (for example, the early stocking cycle and the early stocking time point). In some examples, the event feature extraction model 1230 may include a transposed convolution model, which has a structure opposite to that of causal convolution. For example, as Figure 1 shown, by placing the pooling layer 1231 at the end of the matrix, transposed convolution can be achieved. Assuming the historical time point T0, through convolution of, for example, 3 layers (C1, C2, C3), the last layer can contain information of, for example, 4 future time points (T1, T2, T3, T4). The transposed convolution module can achieve a longer future receptive field by increasing the number of convolution layers or using dilated convolution. Since the TCN model has been proven to be efficient in extracting historical information, a transposed convolution similar to TCN can also efficiently extract future information, helping the model to better predict the early stocking time point and the early stocking cycle according to the preset event information.

[0043] In some embodiments, the kurtosis offset prediction module 1200 may further include a first dimensionality compression model 1260. The first dimensionality compression model is configured to compress the preset event data and output the compressed preset event data to the event feature extraction model 1230. Since the preset event data obtained through the above encoding is sparse features and may include many 0s, it can be compressed by the first dimensionality compression model to obtain the feature data that the event feature extraction model can use, and the occupancy of computing resources can be reduced, improving the computing efficiency. The first dimensionality compression model 1260 may include a fully connected layer or other models capable of compressing data dimensions. Among them, the fully connected layer can be implemented, for example, by the following formula (2):

[0044] M = [F a F b F c × w + b (2)

[0045] where M is the compressed feature data, F a is the holiday time point feature, F b is the feature of the time length from the holiday, F c is the event activity feature (e.g., promotion activity feature), and w and b are the parameters of the fully connected layer.

[0046] In some embodiments, the kurtosis offset prediction module 1200 may further include: a first fusion layer 1240. The first fusion layer is configured to fuse (e.g., stack or splice) the first supply offset feature data and the second supply offset feature data. In addition, the kurtosis offset prediction module 1200 may further include: a second dimensionality compression model 1250. The second dimensionality compression model 1250 is configured to compress the fused first supply offset feature data and second supply offset feature data to output the predicted supply offset feature data. Among them, the second dimensionality compression model can also adopt a fully connected layer or other models capable of compressing data dimensions. The above implementation can fuse the future information features and the temporal offset rules to obtain more accurate predicted supply offset feature data. Compressing the fused first supply offset feature data and second supply offset feature data is convenient for further fusion with the predicted supply feature data later, avoiding the data dimension of the model from being too large and causing dimensional explosion, thereby reducing the occupancy of computing resources and improving the computing efficiency.

[0047] In some embodiments, the output module 1300 may include: a second fusion layer 1310. The second fusion layer is configured to fuse (e.g., superimpose or splice) the predicted supply feature data and the predicted supply offset feature data. The output module 1300 may further include: an output layer 1320. The output layer 1320 is configured to output predicted temporal supply data based on the fused predicted supply feature data and the predicted supply offset feature data, so as to obtain the predicted temporal supply data by considering both the temporal feature pattern and the temporal feature pattern in the case of kurtosis offset.

[0048] Figure 2 FIG. shows a flowchart of a prediction method 2000 according to some exemplary embodiments of the present disclosure. The prediction method may be implemented, for example, by a prediction system 1000. The following describes each step of the prediction method 2000 in detail in conjunction with Figures 1 to 2 , The prediction method 2000 may include: Step S201, obtaining relevant information of the target object, where the relevant information includes historical temporal supply data of the target object at the supply end during a historical time period, historical event data of the target object at the usage end during a historical time period, and preset event data of the target object at the usage end during a preset time period; Step S202, determining predicted supply feature data of the target object during the preset time period based at least on the historical temporal supply data; Step S203, determining predicted supply offset feature data of the target object during the preset time period based at least on the historical temporal supply data, historical event data, and preset event data; and Step S204, determining predicted temporal supply data of the target object at the supply end during the preset time period based on the predicted supply feature data and the predicted supply offset feature data.

[0049] In step S201, relevant information of the target object is obtained.

[0050] In some embodiments, the target object may be a material. The historical temporal supply data is historical shipment volume data, and correspondingly, the predicted temporal supply data is predicted shipment volume data. The preset event data may include at least one of holiday time point features, holiday distance features, and event activity features during the preset time period. The historical event data includes at least one of holiday time point features, holiday distance features, and event activity features during the historical time period. It should be understood here that the features of the target object, historical event data, preset event data, holiday time point features, holiday distance features, and event activity features are the same as those of the corresponding terms described in Figure 1 , and for the sake of brevity, they will not be described in detail here.

[0051] In step S202, predicted supply feature data of the target object during the preset time period is determined based at least on the historical temporal supply data.

[0052] Among them, the predicted supply characteristic data is the supply quantity characteristic data predicted at least based on the historical time-series supply data (that is, the general, common, or conventional supply quantity time series estimated at least based on the historical time-series supply data. For example, the conventional shipment volume series), that is, the conventional supply quantity characteristic data in the preset time period inferred based on the historical time-series law (which can be a specific supply quantity value or a code related to the supply quantity value, and can be converted into a supply quantity value through certain rules). In some examples, at least the historical time-series supply data can be input into the time-series prediction module 1100 to determine the predicted supply characteristic data of the target object within the preset time. At this time, the predicted supply characteristic data represents the supply quantity time series (such as the shipment volume series) of the target object predicted by the time-series prediction module within the preset time period, which is the characteristic data output by the model in the time-series prediction module and can be converted into the supply quantity value of the target object within the preset time period through preset rules. Alternatively, the change law of the time series can also be determined based on the historical time-series supply data in other ways, and then the predicted supply characteristic data can be determined according to this law.

[0053] In step S203, determine the predicted supply offset characteristic data of the target object within the preset time period at least based on the historical time-series supply data, historical event data, and preset event data.

[0054] Since there is a kurtosis offset situation of the target object (for example, there is a situation of advance stock preparation at the selling end of the shipment of materials), it is necessary to consider the event characteristics that cause the kurtosis offset, so as to more accurately estimate the predicted supply offset characteristic data (such as advance shipment characteristic data, including the advance shipment time point, advance shipment cycle, and shipment volume characteristics within the advance shipment cycle) in the case of kurtosis offset.

[0055] Among them, the predicted supply deviation feature data is the supply deviation feature data predicted based on at least historical time-series supply data, historical event data, and event preset event data, that is, the supply volume feature data within the peak deviation cycle in the preset time period inferred considering the peak deviation situation (it can be a specific supply volume value or a code related to the supply volume value, which can be converted into a supply volume value through certain rules). In some embodiments, at least the historical time-series supply data, historical event data, and event preset event data can be input into the kurtosis deviation prediction module 1200 to determine the predicted supply deviation feature data of the target object within the preset time period. The predicted supply deviation feature data represents the supply volume time series of the target object within the peak deviation cycle in the preset time period predicted by the kurtosis deviation prediction module (for example, the shipment volume sequence after considering early shipment), which is the feature data output by the model in the kurtosis deviation prediction module, and can be converted into the early shipment volume value of the target object within the peak deviation cycle in the preset time period through preset rules.

[0056] In step S204, based on the predicted supply feature data and the predicted supply deviation feature data, determine the predicted time-series supply data of the target object at the supply end within the preset time period.

[0057] In some examples, the predicted supply feature data and the predicted supply deviation feature data can be superimposed as the predicted time-series supply data of the target object at the supply end within the preset time period. Alternatively, the predicted supply feature data and the predicted supply deviation feature data can also be fused (for example, spliced), and the fused predicted supply feature data and predicted supply deviation feature data are input into the output layer 1320 to obtain the predicted time-series supply data.

[0058] The above embodiments can differentially extract the time-series feature laws in the case of kurtosis deviation, so as to accurately predict the situation of kurtosis deviation at the supply end (including the kurtosis deviation time point, kurtosis deviation cycle, and time-series features within the kurtosis deviation cycle). In addition, in the case of no kurtosis advance, the above embodiments still maintain a good accuracy rate in predicting the time-series supply data of the target object. Accurately predicting time-series data can reduce the warehousing cost at the supply end, reduce material scrap, and increase the material turnover efficiency.

[0059] In some embodiments, step S202 of determining the predicted supply feature data of the target object within a preset time period based at least on the historical time-series supply data may include: at least performing a feature encoding operation on the historical time-series supply data (for example, inputting it into the second feature encoder 1110) to obtain the historical supply feature encoding of the target object within the historical time period; and performing a feature decoding operation on the historical supply feature encoding (for example, inputting it into the second feature decoder 1120) to obtain the predicted supply feature data of the target object within the preset time period. The above operations of feature encoding and feature decoding can well extract the features of the data and generate corresponding tensors within the preset time period, so as to more accurately and efficiently determine the supply feature data within the preset time period.

[0060] In some embodiments, the relevant information further includes the historical usage data of the target object at the usage end within the historical time period. Wherein, when the target object is a material, the historical usage data can be obtained by using the sales volume of the products manufactured from the material at the selling end and the usage coefficient. At this time, step S202 of determining the predicted supply feature data of the target object within a preset time period based at least on the historical time-series supply data may include: determining the predicted supply feature data of the target object within the preset time period based at least on the historical time-series supply data and the historical usage data. Specifically, for example, at least inputting the historical time-series supply data and the historical usage data into the second feature encoder 1110 to obtain the historical supply feature encoding of the target object within the historical time period. Inputting the historical supply feature encoding into the second feature decoder 1120 to obtain the predicted supply feature data of the target object within the preset time period. The above implementation manner also considers the historical usage data at the usage end (that is, the consumption situation of the target object at the usage end) when determining the time-series law, so as to more accurately determine the time-series law and more accurately predict the supply feature data within the preset time period.

[0061] In some embodiments, the relevant information may further include time feature data, which is obtained by performing a Fourier transform on the time encoding. The Fourier transform can help the model learn the historical fluctuations and trends. The Fourier transform formula is, for example, one of the following formulas (3) and (4):

[0062]

[0063]

[0064] Wherein, T is the period, and an appropriate period can be selected according to the seasonality of the target object, such as a material, or the estimation granularity; x is the time encoding.

[0065] At this time, step S202, determining the predicted supply characteristic data of the target object within a preset time period based on at least historical sequential supply data may include: determining the predicted supply characteristic data of the target object within a preset time period based on at least historical sequential supply data and time characteristic data. Specifically, for example, at least inputting the historical sequential supply data and time characteristic data into the second feature encoder 1110 to obtain the historical supply characteristic encoding of the target object within the historical time period. Inputting the historical supply characteristic encoding into the second feature decoder 1120 to obtain the predicted supply characteristic data of the target object within the preset time period. When determining the sequential rule in the above implementation manner, time characteristic data is also considered, which helps to learn the historical fluctuations and trends.

[0066] In some examples, step S202, determining the predicted supply characteristic data of the target object within a preset time period based on at least historical sequential supply data may include: determining the predicted supply characteristic data of the target object within a preset time period based on historical sequential supply data, historical usage data, and time characteristic data. When determining the sequential rule in the above implementation manner, historical usage data at the usage end (i.e., the consumption situation of the target object at the usage end) and time characteristic data are also considered, so as to more accurately determine the sequential rule and help to learn the historical fluctuations and trends, so as to more accurately predict the supply characteristic data within the preset time period.

[0067] In some embodiments, step S203, determining the predicted supply offset characteristic data of the target object within a preset time period based on at least historical sequential supply data, historical event data, and preset event data includes: determining the first supply offset characteristic data of the target object within a preset time period based on at least historical sequential supply data and historical event data (i.e., characteristic data representing the kurtosis offset time point, kurtosis offset period, supply volume within the kurtosis offset period, etc.); determining the second supply offset characteristic data of the target object within a preset time period based on at least the preset event data (i.e., characteristic data representing the kurtosis offset time point, kurtosis offset period, etc.); and generating the predicted supply offset characteristic data based on the first supply offset characteristic data and the second supply offset characteristic data. The above implementation manner can consider historical sequential supply data and historical event data when predicting the supply offset characteristic data of the target object to obtain the sequential offset rule (e.g., the early shipment rule) when kurtosis offset (i.e., early stockpiling) occurs, so as to differentially extract the rules of small probability events and better predict the sequential data in the case of kurtosis offset. In addition, specifically extract future information characteristics for future information such as preset event data to avoid the problem of feature dimension explosion when receiving future information, thereby avoiding increasing the number of model parameters and training difficulty.

[0068] In some embodiments, determining the first supply offset feature data of a target object within a preset time period based at least on historical time-series supply data and historical event data includes: at least performing a feature encoding operation on the historical time-series supply data and the historical event data (for example, inputting them into the first feature encoder 1210) to obtain the historical supply offset feature encoding of the target object within the historical time period; and performing a feature decoding operation on the historical supply offset feature encoding (for example, inputting it into the first feature decoder 1220) to obtain the first supply offset feature data of the target object within the preset time period. The above operations of feature encoding and feature decoding can well extract the features of the data and generate corresponding tensors within the preset time period, so as to more accurately and efficiently determine the supply feature data within the preset time period.

[0069] In some embodiments, determining the first supply offset feature data of a target object within a preset time period based at least on historical time-series supply data and historical event data may further include: before performing the feature encoding operation on the historical event data, performing a dimension compression operation on the historical event data (for example, inputting it into the first dimension compression model 1260) to obtain the compressed historical event data. By compressing the historical event data, the requirements of the feature encoder for the input data dimension are met, so as to improve the running speed of the model and the calculation efficiency.

[0070] In some embodiments, step S203, determining the first supply offset feature data of a target object within a preset time period based at least on historical time-series supply data and historical event data includes: determining the first supply offset feature data based on the historical time-series supply data, historical usage data, and historical event data. Specifically, for example, inputting the historical time-series supply data, historical usage data, and historical event data into the first feature encoder 1210 to obtain the historical supply offset feature encoding of the target object within the historical time period; and inputting the historical supply offset feature encoding into the first feature decoder 1220 to obtain the first supply offset feature data of the target object within the preset time period. When determining the time-series offset rule, the above implementation manner also considers the historical usage data at the usage end (that is, the consumption situation of the target object at the usage end), so as to more accurately determine the time-series offset rule to predict the predicted time-series supply data within the preset time period.

[0071] In some embodiments, determining the second supply offset feature data of the target object within a preset time period based at least on preset event data includes: at least performing a feature extraction operation on the preset event data (for example, inputting it into the event feature extraction model 1230) to obtain the second supply offset feature data (that is, the feature data characterizing the kurtosis offset time point, kurtosis offset period, etc.). The above implementation manner can obtain the event features that affect the supply offset, so as to more accurately and efficiently determine the supply data within the preset time period.

[0072] In some embodiments, determining the second supply offset feature data of the target object within a preset time period based at least on preset event data may further include: before performing the feature extraction operation on the preset event data, performing a dimension compression operation on the preset event data (for example, inputting it into the first dimension compression model 1260) to obtain the compressed preset event data. By compressing the preset event data, the requirements of the event feature extraction model 1230 for the input data dimension are met, so as to improve the running speed of the model and the calculation efficiency.

[0073] In some embodiments, the relevant information further includes the demand data of the target object at the usage end within a preset time period. The demand data of the target object within a preset time period can be provided by the usage end or obtained through a prediction model. At this time, determining the second supply offset feature data of the target object within a preset time period based at least on preset event data includes: determining the second supply offset feature data based on the preset event data and the demand data. Specifically, for example, inputting the preset event data and the demand data into the event feature extraction model 1230 to obtain the second supply offset feature data. The above further considering the future demand data can more accurately determine the future information features.

[0074] In some embodiments, generating the predicted supply offset feature data based on the first supply offset feature data and the second supply offset feature data may include: fusing the first supply offset feature data and the second supply offset feature data (for example, inputting the first supply offset feature data and the second supply offset feature data into the first fusion layer 1240); performing a dimension compression operation on the fused first supply offset feature data and second supply offset feature data (for example, inputting it into the second dimension compression model 1250) to obtain the predicted supply offset feature data. The above implementation manner can fuse the future information features and the time series offset law to obtain more accurate predicted supply offset feature data. Compressing the fused first supply offset feature data and second supply offset feature data is convenient for further fusing with the predicted supply feature data later, avoiding the data dimension of the model from being too large, causing dimensional explosion, thereby reducing the occupation of computing resources and improving the calculation efficiency.

[0075] In some embodiments, step S204, determining the predicted sequential supply data of the target object at the supply end within a preset time period based on the predicted supply characteristic data and the predicted supply offset characteristic data may include: fusing the predicted supply characteristic data and the predicted supply offset characteristic data (for example, inputting the predicted supply characteristic data and the predicted supply offset characteristic data into the second fusion layer 1310); and for the fused predicted supply characteristic data and predicted supply offset characteristic data, performing time shift on the predicted supply characteristic data by using the predicted supply offset characteristic data (for example, inputting the fused predicted supply characteristic data and predicted supply offset characteristic data into the output layer 1320) to obtain the predicted sequential supply data, so as to obtain the predicted sequential supply data by considering both the sequential characteristic law and the sequential characteristic law in the case of kurtosis offset.

[0076] Figure 3 The figure shows the prediction result graph using the traditional prediction system TCN; Figure 4 The figure shows the prediction result graph using the prediction system 1000 and the prediction method 2000 of the present disclosure. Figure 3 It can be seen that the predicted material shipment volume at the supply end almost coincides with the predicted material demand volume at the sales end, without considering the advance stock preparation time point and the advance stock preparation period. Therefore, there is a kurtosis offset from the actual material shipment volume, and it is impossible to accurately predict the material shipment volume at the supply end in the case of advance stock preparation. From Figure 4 It can be seen that the prediction system 1000 and the prediction method 2000 of the present disclosure can make the predicted material shipment volume almost coincide with the actual material shipment volume, and there is a kurtosis offset from the predicted material demand volume at the sales end. The advance stock preparation time point and the advance stock preparation period are considered, so as to achieve accurate prediction of the material shipment volume at the supply end in the case of advance stock preparation.

[0077] Figure 5The structural block diagram of a prediction device 5000 according to an exemplary embodiment of the present disclosure is shown. The prediction device 5000 may include: an acquisition module 501 configured to acquire relevant information of a target object, where the relevant information includes historical sequential supply data of the target object at the supply end within a historical time period, historical event data of the target object at the usage end within a historical time period, and preset event data of the target object at the usage end within a preset time period; a first determination module 502 configured to determine predicted supply feature data of the target object within a preset time period based at least on the historical sequential supply data; a second determination module 503 configured to determine predicted supply offset feature data of the target object within a preset time period based at least on the historical sequential supply data, historical event data, and preset event data; and a third determination module 504 configured to determine predicted sequential supply data of the target object at the supply end within a preset time period based on the predicted supply feature data and the predicted supply offset feature data.

[0078] The above embodiments can differentially extract the temporal feature law in the case of kurtosis offset, so as to accurately predict the kurtosis offset situation at the supply end (including the kurtosis offset time point, kurtosis offset period, and temporal features within the kurtosis offset period). In addition, in the case where there is no kurtosis forward shift, the above embodiments still maintain a good accuracy rate in predicting the sequential supply data of the target object. Accurately predicting the sequential data can reduce the warehousing cost at the supply end, reduce material scrapping, and increase the material turnover efficiency.

[0079] In some embodiments, the first determination module 502 may include: a second feature encoding module configured to perform at least a feature encoding operation on the historical sequential supply data to obtain historical supply feature encoding of the target object within a historical time period; and a second feature decoding module configured to perform a feature decoding operation on the historical supply feature encoding to obtain predicted supply feature data of the target object within a preset time period. The above operations of feature encoding and feature decoding can well extract the features of the data and generate corresponding tensors within a preset time period, so as to more accurately and efficiently determine the supply feature data within a preset time period.

[0080] In some embodiments, the relevant information further includes historical usage data of the target object within a historical time period at the usage end. Wherein, when the target object is a material, the historical usage data can be obtained from the sales volume of the products manufactured using the material at the sales end and the usage coefficient. At this time, the first determination module 502 may include: a fourth determination module configured to determine the predicted supply characteristic data of the target object within a preset time period based at least on the historical sequential supply data and the historical usage data. When determining the sequential pattern in the above implementation manner, the historical usage data at the usage end (i.e., the consumption situation of the target object at the usage end) is also considered, so as to more accurately determine the sequential pattern and more accurately predict the supply characteristic data within the preset time period.

[0081] In some embodiments, the relevant information may further include time characteristic data, which is obtained by performing a Fourier transform on the time encoding. At this time, the first determination module 502 may include: a fifth determination module configured to determine the predicted supply characteristic data of the target object within a preset time period based at least on the historical sequential supply data and the time characteristic data. When determining the sequential pattern in the above implementation manner, the time characteristic data is also considered, which helps to learn the historical fluctuations and trends.

[0082] In some examples, the first determination module 502 may include: a sixth determination module configured to determine the predicted supply characteristic data of the target object within a preset time period based on the historical sequential supply data, the historical usage data, and the time characteristic data. When determining the sequential pattern in the above implementation manner, the historical usage data at the usage end (i.e., the consumption situation of the target object at the usage end) and the time characteristic data are also considered, so as to more accurately determine the sequential pattern and help to learn the historical fluctuations and trends, and more accurately predict the supply characteristic data within the preset time period.

[0083] In some embodiments, the second determination module 503 includes: a seventh determination module configured to determine first supply offset feature data of the target object within a preset time period (i.e., feature data representing the kurtosis offset time point, kurtosis offset period, supply quantity within the kurtosis offset period, etc.) at least based on historical sequential supply data and historical event data; an eighth determination module configured to determine second supply offset feature data of the target object within a preset time period (i.e., feature data representing the kurtosis offset time point, kurtosis offset period, etc.) at least based on preset event data; and a generation module configured to generate predicted supply offset feature data based on the first supply offset feature data and the second supply offset feature data. The above embodiments can consider historical sequential supply data and historical event data when predicting the supply offset feature data of the target object to obtain the sequential offset rule (e.g., early shipment rule) when kurtosis offset (i.e., early stockpiling) occurs, so as to differentially extract the rules of small probability events and better predict sequential data in the case of kurtosis offset. In addition, future information features are separately extracted for future information such as preset event data to avoid the problem of feature dimension explosion when receiving future information, thereby avoiding increasing the number of model parameters and training difficulty.

[0084] In some embodiments, the seventh determination module includes: a first feature encoding module configured to perform feature encoding operations on at least the historical sequential supply data and historical event data (e.g., inputting them into the first feature encoder 1210) to obtain historical supply offset feature encodings of the target object within a historical time period; and a first feature decoding module configured to perform feature decoding operations on the historical supply offset feature encodings (e.g., inputting them into the first feature decoder 1220) to obtain first supply offset feature data of the target object within a preset time period. The above operations of feature encoding and feature decoding can well extract the features of the data and generate corresponding tensors within a preset time period, so as to more accurately and efficiently determine the supply feature data within the preset time period.

[0085] In some embodiments, the seventh determination module may further include: a first dimension compression module configured to perform dimension compression operations on the historical event data (e.g., inputting them into the first dimension compression model 1260) before performing feature encoding operations on the historical event data to obtain compressed historical event data. By compressing the historical event data, the requirements of the feature encoder for the input data dimension are met, so as to improve the running speed of the model and improve the calculation efficiency.

[0086] In some embodiments, the second determination module 503 includes: a ninth determination module configured to determine first supply offset feature data based on historical sequential supply data, historical usage data, and historical event data. When determining the sequential offset rule, the above-described embodiment also considers the historical usage data at the usage end (i.e., the consumption situation of the target object at the usage end), so as to more accurately determine the sequential offset rule to predict the predicted sequential supply data within a preset time period.

[0087] In some embodiments, the eighth determination module includes: a feature extraction module configured to perform at least feature extraction operations on preset event data to obtain second supply offset feature data (i.e., feature data characterizing the kurtosis offset time point, kurtosis offset period, etc.). The above-described embodiment can obtain event features that affect the supply offset, so as to more accurately and efficiently determine the supply data within a preset time period.

[0088] In some embodiments, the eighth determination module may further include: a first dimension compression module configured to perform dimension compression operations on the preset event data before performing feature extraction operations on the preset event data to obtain compressed preset event data. By compressing the preset event data, the requirements for the input data dimension of the event feature extraction model are met, so as to improve the running speed of the model and improve the calculation efficiency.

[0089] In some embodiments, the relevant information further includes the demand data of the target object at the usage end within a preset time period. The demand data of the target object within a preset time period can be provided by the usage end or obtained through a prediction model. At this time, the eighth determination module includes: a tenth determination module configured to determine second supply offset feature data based on the preset event data and the demand data. The above further consideration of future demand data can more accurately determine future information features.

[0090] In some embodiments, the generation module may include: a first fusion module configured to fuse the first supply offset feature data and the second supply offset feature data; a second dimension compression module configured to perform dimension compression operations on the fused first supply offset feature data and second supply offset feature data to obtain predicted supply offset feature data. The above-described embodiment can fuse future information features and sequential offset rules to obtain more accurate predicted supply offset feature data. Compressing the fused first supply offset feature data and second supply offset feature data is convenient for further fusion with the predicted supply feature data later, avoiding excessive data dimensions of the model and causing dimensional explosion, thereby reducing the occupation of computing resources and improving the calculation efficiency.

[0091] In some embodiments, the third determination module may include: a second fusion module configured to fuse the predicted supply feature data and the predicted supply offset feature data; and a time shift module configured to, for the fused predicted supply feature data and predicted supply offset feature data, perform a time shift on the predicted supply feature data by using the predicted supply offset feature data to obtain predicted time-series supply data, so as to obtain the predicted time-series supply data by taking into account both the time-series feature pattern and the time-series feature pattern in the case of kurtosis offset.

[0092] It should be understood that Figure 5 each module of the device 5000 shown in Figure 2 may correspond to each step in the prediction method 2000 described with reference to

[0093] Accordingly, the operations, features, and advantages described above for the prediction method 2000 also apply to the device 5000 and the modules included therein. For the sake of brevity, certain operations, features, and advantages are not described herein again.

[0093] According to another aspect of the present disclosure, there is also provided an electronic device, including: a memory, a processor, and a computer program stored on the memory, wherein the processor is configured to execute the computer program to implement the steps of method 2000.

[0094] According to another aspect of the present disclosure, there is also provided a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of method 2000.

[0095] According to another aspect of the present disclosure, there is also provided a computer program product including a computer program, wherein the computer program, when executed by a processor, implements the steps of method 2000.

[0096] Referring to Figure 6 , a block diagram of an electronic device 6000 that may be the present disclosure will now be described, which is an example of a hardware device applicable to various aspects of the present disclosure. The electronic device may be different types of computer devices, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0097] Figure 6 A block diagram of an electronic device according to an embodiment of the present disclosure is shown. As Figure 6As shown, the electronic device 6000 may include at least one processor 601, a working memory 602, an I / O device 604, a display device 605, a storage device 606, and a communication interface 607 that can communicate with each other via a system bus 603.

[0098] The processor 601 can be a single processing unit or multiple processing units, and all processing units can include a single or multiple computing units or multiple cores. The processor 601 can be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuits, and / or any device that manipulates signals based on operation instructions. The processor 601 can be configured to obtain and execute computer-readable instructions stored in the working memory 602, the storage device 605, or other computer-readable media, such as the program code of the operating system 602a, the program code of the application 602b, etc.

[0099] The working memory 602 and the storage device 606 are examples of computer-readable storage media for storing instructions that are executed by the processor 601 to implement the various functions described above. The working memory 602 can include both volatile and non-volatile memories (e.g., RAM, ROM, etc.). In addition, the storage device 606 can include a hard disk drive, a solid-state drive, removable media, including external and removable drives, memory cards, flash memory, floppy disks, optical discs (e.g., CD, DVD), storage arrays, network-attached storage, storage area networks, etc. The working memory 602 and the storage device 606 can both be collectively referred to as memory or computer-readable storage media in this article, and can be non-transitory media capable of storing computer-readable, processor-executable program instructions as computer program code, and the computer program code can be executed by the processor 601 as a specific machine configured to implement the operations and functions described in the examples in this article.

[0100] The I / O device 604 can include an input device and / or an output device. The input device can be any type of device that can input information into the electronic device 6000, and can include, but is not limited to, a mouse, a keyboard, a touch screen, a trackpad, a trackball, a joystick, a microphone, and / or a remote control. The output device can be any type of device that can present information, and can include, but is not limited to, a video / audio output terminal, a vibrator, and / or a printer.

[0101] The communication interface 607 allows the electronic device 6000 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks, and may include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth™ device, an 802.11 device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0102] The application 602b in the working register 602 can be loaded and executed to perform the various methods and processes described above, such as Figure 2 the steps S201 - S204 in. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 6000 via the storage device 606 and / or the communication interface 607. When the computer program is loaded and executed by the processor 601, one or more steps of the prediction method described above can be performed.

[0103] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuit systems, field - programmable gate arrays (FPGA), application - specific integrated circuits (ASIC), application - specific standard products (ASSP), systems - on - chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special - purpose or general - purpose programmable processor that receives data and instructions from a storage system, at least one input device, and at least one output device, and transmits the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0104] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to the processor or controller of a general - purpose computer, a special - purpose computer, or other programmable data - processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowchart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand - alone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0105] In the context of this disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0106] In order to provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0107] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of a communication network include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0108] A computing system can include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship between the client and the server is generated by computer programs running on the respective computers and having a client-server relationship with each other.

[0109] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and no limitations are imposed herein.

[0110] Although embodiments or examples of the present disclosure have been described with reference to the accompanying drawings, it should be understood that the above methods, systems, and devices are merely exemplary embodiments or examples, and the scope of the present invention is not limited by these embodiments or examples, but is only defined by the authorized claims and their equivalent scope. Various elements in the embodiments or examples can be omitted or replaced by their equivalent elements. In addition, the steps can be executed in an order different from that described in the present disclosure. Further, the various elements in the embodiments or examples can be combined in various ways. Importantly, with the evolution of technology, many of the elements described herein can be replaced by equivalent elements that emerge after the present disclosure.

Claims

1. A method for predicting the supply quantity of an object, comprising: obtaining relevant information of the target object, where the relevant information includes: historical sequential supply data of the target object at the supply end within a historical time period, historical event data of the target object at the usage end within the historical time period, and preset event data of the target object at the usage end within a preset time period; determining, at least based on the historical sequential supply data, the predicted supply characteristic data of the target object within the preset time period; determining, at least based on the historical sequential supply data, the historical event data, and the preset event data, the predicted supply offset characteristic data of the target object within the preset time period; and determining, based on the predicted supply characteristic data and the predicted supply offset characteristic data, the predicted sequential supply data of the target object at the supply end within the preset time period.

2. The prediction method according to claim 1, wherein determining, at least based on the historical sequential supply data, the historical event data, and the preset event data, the predicted supply offset characteristic data of the target object within the preset time period includes: determining, at least based on the historical sequential supply data and the historical event data, the first supply offset characteristic data of the target object within the preset time period; determining, at least based on the preset event data, the second supply offset characteristic data of the target object within the preset time period; and generating the predicted supply offset characteristic data based on the first supply offset characteristic data and the second supply offset characteristic data.

3. The prediction method according to claim 2, wherein determining, at least based on the historical sequential supply data and the historical event data, the first supply offset characteristic data of the target object within the preset time period includes: performing at least a feature encoding operation on the historical sequential supply data and the historical event data to obtain the historical supply offset feature encoding of the target object within the historical time period; and performing a feature decoding operation on the historical supply offset feature encoding to obtain the first supply offset characteristic data of the target object within the preset time period.

4. The prediction method according to claim 2, wherein determining, at least based on the preset event data, the second supply offset characteristic data of the target object within the preset time period includes: performing at least a feature extraction operation on the input of the preset event data to obtain the second supply offset characteristic data.

5. The prediction method according to claim 2, wherein generating the predicted supply offset characteristic data based on the first supply offset characteristic data and the second supply offset characteristic data includes: fusing the first supply offset characteristic data and the second supply offset characteristic data; performing a dimension compression operation on the fused first supply offset characteristic data and second supply offset characteristic data to obtain the predicted supply offset characteristic data.

6. The prediction method according to claim 1, wherein Determining the predicted supply feature data of the target object within the preset time period based at least on the historical sequential supply data includes: Performing at least a feature encoding operation on the historical sequential supply data to obtain the historical supply feature encoding of the target object within the historical time period; and Performing a feature decoding operation on the historical supply feature encoding to obtain the predicted supply feature data of the target object within the preset time period.

7. The prediction method according to any one of claims 1 to 6, wherein, Determining the predicted sequential supply data of the target object at the supply end within the preset time period based on the predicted supply feature data and the predicted supply offset feature data includes: Fusing the predicted supply feature data and the predicted supply offset feature data; and For the fused predicted supply feature data and predicted supply offset feature data, shifting the predicted supply feature data in time using the predicted supply offset feature data to obtain the predicted sequential supply data.

8. The prediction method according to any one of claims 2 to 5, wherein, The relevant information further includes the demand data of the target object at the usage end within the preset time period, and wherein, determining the second supply offset feature data of the target object within the preset time period based at least on the preset event data includes: determining the second supply offset feature data based on the preset event data and the demand data, and / or wherein the relevant information further includes the historical usage data of the target object at the usage end within the historical time period, and wherein, determining the predicted supply feature data of the target object within the preset time period based at least on the historical sequential supply data includes: determining the predicted supply feature data of the target object within the preset time period based at least on the historical sequential supply data and the historical usage data, and wherein, determining the first supply offset feature data of the target object within the preset time period based at least on the historical sequential supply data and the historical event data includes: determining the first supply offset feature data based on the historical sequential supply data, the historical usage data, and the historical event data, and / or wherein the relevant information further includes time feature data, the time feature data is obtained by performing a Fourier transform on time encoding, and wherein, determining the predicted supply feature data of the target object within the preset time period based at least on the historical sequential supply data includes: determining the predicted supply feature data of the target object within the preset time period based at least on the historical sequential supply data and the time feature data.

9. A prediction system, wherein, The prediction system is used for predicting the supply quantity of a target object, and the target object has relevant information, where the relevant information includes: historical time-series supply data of the target object at the supply end within a historical time period, historical event data of the target object at the usage end within the historical time period, and preset event data of the target object at the usage end within a preset time period, and the prediction system includes: A time-series prediction module configured to determine prediction supply characteristic data of the target object within the preset time period based at least on the historical time-series supply data; A kurtosis deviation prediction module configured to determine prediction supply deviation characteristic data of the target object within the preset time period based at least on the historical time-series supply data, the historical event data, and the preset event data; and An output module configured to determine predicted time-series supply data of the target object at the supply end within the preset time period based on the prediction supply characteristic data and the prediction supply deviation characteristic data.

10. A prediction device for the supply quantity of an object comprising: An acquisition module configured to acquire relevant information of a target object, where the relevant information includes historical time-series supply data of the target object at the supply end within a historical time period, historical event data of the target object at the usage end within the historical time period, and preset event data of the target object at the usage end within a preset time period; A first determination module configured to determine prediction supply characteristic data of the target object within the preset time period based at least on the historical time-series supply data; A second determination module configured to determine prediction supply deviation characteristic data of the target object within the preset time period based at least on the historical time-series supply data, the historical event data, and the preset event data; and A third determination module configured to determine predicted time-series supply data of the target object at the supply end within the preset time period based on the prediction supply characteristic data and the prediction supply deviation characteristic data.

11. An electronic device comprising: A memory, a processor, and a computer program stored on the memory, wherein the processor is configured to execute the computer program to implement the steps of the prediction method according to any one of claims 1-8.

12. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of the prediction method according to any one of claims 1-8.

Images