Demand data prediction method and device, electronic device, and storage medium

Abstract

The present disclosure provides a demand data prediction method and device, electronic device, and storage medium, which relate to the field of data processing. The demand data prediction method includes: collecting target order data and target order related data; obtaining a pre-built short-term prediction model, and inputting the target order data into the short-term prediction model to obtain a short-term demand data prediction result; building a long-term prediction model based on the target order data, the target order related data, and the short-term demand data prediction result, and calling the target prediction time data, using the target prediction time data as the input data of the long-term prediction model, and the long-term prediction model outputs the long-term demand data prediction result; the short-term demand data prediction result and the long-term demand data prediction result are merged to obtain the target demand data prediction result. The technical solution of the embodiment of the present disclosure can expand the order data source and improve the accuracy of the demand data prediction result.

Description

Background Art

[0002] With the advent of the big data era, more and more businesses rely on big data analysis, and then predict the relevant demand data of the business based on the results of big data analysis, and make reasonable layout of the industries corresponding to the business based on the prediction results of relevant demand data.

[0003] In the relevant demand data prediction methods, either the data source of the demand data is relatively single, or only short-term demand data can be predicted, while the accuracy of the prediction results of long-term demand data is relatively low.

[0004] Therefore, how to provide a high-accuracy and universal demand data forecasting method is of great practical significance for business planning.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0006] The purpose of the embodiments of the present disclosure is to provide a demand data prediction method, a demand data prediction device, an electronic device and a computer-readable storage medium, thereby overcoming the problem of low accuracy of demand data prediction at least to a certain extent.

[0007] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0008] According to a first aspect of an embodiment of the present disclosure, a demand data forecasting method is provided, comprising: collecting target order data and target order-related data; obtaining a pre-built short-term forecasting model, and inputting the target order data into the short-term forecasting model to obtain a short-term demand data forecasting result; building a long-term forecasting model based on the target order data, the target order-related data, and the short-term demand data forecasting result, and calling the target forecasting time data, and outputting the long-term demand data forecasting result according to the target forecasting time data based on the long-term forecasting model; fusing the short-term demand data forecasting result and the long-term demand data forecasting result to obtain a target demand data forecasting result.

[0009] In some example embodiments of the present disclosure, based on the aforementioned scheme, the collection of target order data and target order associated data includes: acquiring historical order data and historical order associated data; cleaning the historical order data, and extracting the target order data from the cleaned historical order data; calculating the correlation coefficient between the historical order associated data and the target order data, and sorting the correlation coefficient to obtain weight data of the historical order associated data; and filtering the target order associated data corresponding to the target order data from the historical order associated data according to the weight data.

[0010] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the short-term prediction model includes a first type of time series model, a second type of time series model, and a third type of time series model, and the obtaining of a pre-built short-term prediction model and inputting the target order data into the short-term prediction model to obtain a short-term demand data prediction result includes: obtaining the pre-built first type of time series model, the second type of time series model, and the third type of time series model; inputting the target order data into the first type of time series model, the second type of time series model, and the second type of time series model, respectively, to obtain a first short-term prediction result, a second short-term prediction result, and a third short-term prediction result; integrating the first short-term prediction result, the second short-term prediction result, and the third short-term prediction result to obtain the short-term demand data prediction result.

[0011] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the long-term prediction model is constructed based on the target order data, the target order-related data, and the short-term demand data prediction results, including: obtaining a pre-constructed initial long-term prediction model, and training the initial long-term prediction model based on the target order data, the target order-related data, and the short-term demand data prediction results to obtain model parameters of the trained initial long-term prediction model; and constructing the long-term prediction model based on the model parameters.

[0012] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the model parameters include a first model parameter and a second model parameter, and the initial long-term prediction model is trained based on the target order data, the target order-related data, and the short-term demand data prediction results to obtain the model parameters of the trained initial long-term prediction model, including: determining the short-term real demand data corresponding to the target order data, and screening the key order-related data with slow data changes from the target order-related data; performing linear regression curve fitting on the target order data and the target order-related data to obtain a linear regression model, and inputting the key order-related data into the linear regression model to obtain the first model parameter; calling the loss function of the initial long-term prediction model, and inputting the short-term demand data prediction results and the short-term real demand data into the loss function to obtain the second model parameter, and using the first model parameter and the second model parameter as the model parameters of the trained initial long-term prediction model.

[0013] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the long-term demand data prediction result is output according to the target prediction time data based on the long-term prediction model, including: dividing the target prediction time data into stages to determine the long-term prediction duration data in the target prediction time data; inputting the long-term prediction duration data into the long-term prediction model to obtain the long-term demand data prediction result.

[0014] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the short-term demand data forecast result and the long-term demand data forecast result are merged to obtain the demand data forecast result, including: determining the short-term time granularity data of the short-term demand data forecast result, and the long-term time granularity data of the long-term demand data forecast result; performing time granularity uniform processing on the short-term time granularity data and the long-term time granularity data to obtain standard time granularity data; performing granularity conversion on the short-term demand data forecast result and the long-term demand data forecast result based on the standard time granularity data, and fusing the short-term demand data forecast result and the long-term demand data forecast result after granularity conversion to obtain the target demand data forecast result.

[0015] According to the second aspect of an embodiment of the present disclosure, a demand data forecasting device is provided, comprising: a data collection module, used to collect target order data and target order-related data; a short-term demand data forecasting result determination module, used to obtain a pre-built short-term forecasting model, and input the target order data into the short-term forecasting model to obtain a short-term demand data forecasting result; a long-term demand data forecasting result determination module, used to fit the target order data, the target order-related data, and the short-term demand data forecasting result to obtain a long-term forecasting model, and call the target forecasting time data, and output the long-term demand data forecasting result based on the long-term forecasting model according to the target forecasting time data; a forecasting result fusion module, used to fuse the short-term demand data forecasting result and the long-term demand data forecasting result to obtain a target data forecasting result.

[0016] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the data collection module includes a data preprocessing unit, and the data preprocessing unit is used to obtain historical order data and historical order associated data; clean the historical order data, and extract the target order data from the cleaned historical order data; calculate the correlation coefficient between the historical order associated data and the target order data, and sort the correlation coefficient to obtain weight data of the historical order associated data; and filter the target order associated data corresponding to the target order data from the historical order associated data according to the weight data.

[0017] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the short-term demand data forecast result determination module includes a short-term demand data forecast result determination unit, and the short-term demand data forecast result determination unit is used to obtain the pre-built first type time series model, the second type time series model, and the third type time series model; input the target order data into the first type time series model, the second type time series model, and the second type time series model respectively to obtain a first short-term forecast result, a second short-term forecast result, and a third short-term forecast result; and integrate the first short-term forecast result, the second short-term forecast result, and the third short-term forecast result to obtain the short-term demand data forecast result.

[0018] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the long-term demand data prediction result determination module includes a long-term prediction model construction unit, and the long-term prediction model construction unit is used to obtain a pre-constructed initial long-term prediction model, and train the initial long-term prediction model based on the target order data, the target order associated data, and the short-term demand data prediction results to obtain model parameters of the trained initial long-term prediction model; and construct the long-term prediction model based on the model parameters.

[0019] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the long-term demand data prediction result determination module also includes a model parameter determination unit, the model parameter determination unit is used to determine the short-term real demand data corresponding to the target order data, and to filter the key order-related data with slow data changes from the target order-related data; perform linear regression curve fitting on the target order data and the target order-related data to obtain a linear regression model, and input the key order-related data into the linear regression model to obtain the first model parameters; determine the loss function of the initial long-term prediction model, and input the short-term demand data prediction result and the short-term real demand data into the loss function to obtain the second model parameters, and use the first model parameters and the second model parameters as the model parameters of the trained initial long-term prediction model.

[0020] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the long-term demand data forecast result determination module also includes a long-term demand data forecast result determination unit, and the long-term demand data forecast result determination unit is used to perform stage division processing on the target prediction time data, and determine the long-term prediction duration data in the target prediction time data; input the long-term prediction duration data into the long-term prediction model to obtain the long-term demand data prediction result.

[0021] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the prediction data fusion module includes a prediction result fusion unit, and the prediction result fusion unit is used to determine the short-term time granularity data of the short-term demand data prediction result, and the long-term time granularity data of the long-term demand data prediction result; perform time granularity uniform processing on the short-term time granularity data and the long-term time granularity data to obtain standard time granularity data; perform granularity conversion on the short-term demand data prediction result and the long-term demand data prediction result based on the standard time granularity data, and fuse the short-term demand data prediction result and the long-term demand data prediction result after granularity conversion to obtain the target demand data prediction result.

[0022] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device, comprising: a processor; and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the demand data prediction method described in any one of the above items is implemented.

[0023] According to a fourth aspect of an embodiment of the present disclosure, there is provided a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for predicting demand data according to any one of the above items is implemented.

[0024] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0025] The demand data prediction method in the example embodiment of the present disclosure collects target order data and target order associated data; obtains a pre-built short-term prediction model, and inputs the target order data into the short-term prediction model to obtain a short-term demand data prediction result; constructs a long-term prediction model based on the target order data, the target order associated data, and the short-term demand data prediction result, and outputs the long-term demand data prediction result through the long-term prediction model; fuses the short-term demand data prediction result and the long-term demand data prediction result to obtain the target demand data prediction result. On the one hand, the target order data and the target order associated data can be used as the data source of the target demand data prediction result, avoiding the inability to perform long-term demand data prediction due to low business volume; on the other hand, the short-term demand data prediction result can be involved in the long-term demand data prediction process, and a long-term prediction model can be constructed based on the target order data, the target order associated data, and the short-term demand data prediction result, avoiding the large cumulative prediction error of the long-term prediction model due to excessive concentration of data sources, improving the prediction accuracy of the long-term prediction model, and thus improving the accuracy of the target demand data prediction result.

[0026] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0028] Figure 1 A schematic diagram schematically shows a demand data forecasting method process according to some embodiments of the present disclosure;

[0029] Figure 2 A schematic diagram schematically shows a data preprocessing method flow according to some embodiments of the present disclosure;

[0030] Figure 3 A schematic diagram schematically shows a process flow of a short-term demand data forecasting method according to some embodiments of the present disclosure;

[0031] Figure 4 A schematic diagram schematically illustrates a long-term prediction model building architecture according to some embodiments of the present disclosure;

[0032] Figure 5 A schematic diagram schematically shows a flow chart of a method for determining model parameters according to some embodiments of the present disclosure;

[0033] Figure 6 A schematic diagram schematically shows a prediction result fusion method process according to some embodiments of the present disclosure;

[0034] Figure 7 A schematic diagram schematically shows a demand data prediction device according to some embodiments of the present disclosure;

[0035] Figure 8 A schematic diagram schematically shows a structure of a computer system of an electronic device according to some embodiments of the present disclosure;

[0036] Fig. 9 A schematic diagram of a computer-readable storage medium according to some embodiments of the present disclosure is schematically shown.

[0037] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0038] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be more comprehensive and complete and will fully convey the concept of the example embodiments to those skilled in the art.

[0039] In addition, the described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring the various aspects of the present disclosure.

[0040] Furthermore, the drawings are only schematic illustrations and are not necessarily drawn to scale. The block diagrams shown in the drawings are only functional entities and do not necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0041] In this exemplary embodiment, a demand data prediction method is first provided, and the demand data prediction method can be applied to a server. Figure 1 The following is a schematic diagram of a demand data forecasting method process according to some embodiments of the present disclosure. Figure 1 As shown, the demand data prediction method may include the following steps:

[0042] In step S110, target order data and target order associated data are collected;

[0043] In step S120, a pre-built short-term forecasting model is obtained, and the target order data is input into the short-term forecasting model to obtain a short-term demand data forecast result;

[0044] In step S130, a long-term prediction model is constructed based on the target order data, the target order associated data, and the short-term demand data prediction result, and the target prediction time data is called, and the long-term demand data prediction result is output according to the target prediction time data based on the long-term prediction model;

[0045] In step S140, the short-term demand data prediction result and the long-term demand data prediction result are fused to obtain the target demand data prediction result.

[0046] According to the demand data prediction method in this example embodiment, on the one hand, the target order data and the target order-related data can be used as the data source of the target demand data prediction results, thereby avoiding the inability to perform long-term demand data prediction due to low business volume; on the other hand, the short-term demand data prediction results can be involved in the long-term demand data prediction process, and a long-term prediction model can be constructed based on the target order data, the target order-related data, and the short-term demand data prediction results, thereby avoiding the large cumulative prediction error of the long-term prediction model due to excessive concentration of data sources, improving the prediction accuracy of the long-term prediction model, and thus improving the accuracy of the target demand data prediction results.

[0047] Next, the demand data prediction method in this example embodiment will be further described.

[0048] In step S110, target order data and target order associated data are collected.

[0049] In an example embodiment of the present disclosure, the target order data may refer to data obtained after pre-processing order data related to the city to be predicted. For example, the express business data may be order data of the city to be predicted and an assisting prediction city with the same administrative level as the city to be predicted. The target order data may be data obtained by filtering abnormal order data in the order data of the city to be predicted and the city to be predicted. The target order data may also be data obtained by backfilling abnormal orders in the order data of the city to be predicted and the city to be predicted. Of course, the target order data may also be data obtained by performing other processing on the order data of the city to be predicted, and this example embodiment does not specifically limit this.

[0050] The target order-related data may refer to data obtained after preprocessing order-related data related to the city to be predicted. For example, the order-related data may be the consumer price index, the number of permanent residents, the number of broadband access users, and the gross domestic product (GDP) of the city to be predicted and the assisting prediction city. The target order-related data may be data obtained by calculating weight data of each order-related data and filtering the data from the order-related data according to each weight data. The target order-related data may also be data obtained by filtering the determined order-related data according to the degree of correlation between each order-related data and the target order data. Of course, the target order-related data may also be data obtained by processing other order-related data related to the city to be predicted. This embodiment does not specifically limit this.

[0051] With the rapid development of the express delivery industry, it is necessary to adjust logistics sites and routes in a timely manner. Most logistics site and route adjustments rely on long-term forecasts of the city's express delivery business volume. This can be done by obtaining historical order data and historical order-related data of the city to be predicted and the city assisting in the prediction, and pre-processing the obtained historical order data and historical order-related data to obtain target order data and target order-related data, thereby avoiding the situation where a single city has too little historical order data to make long-term forecasts. At the same time, the order-related data that affects the city's order data is included in the city's demand data prediction data source, reducing the possibility of relying solely on historical order data for target demand data prediction, resulting in large deviations in the target demand data prediction results.

[0052] In step S120, a pre-built short-term forecasting model is obtained, and the target order data is input into the short-term forecasting model to obtain a short-term demand data forecasting result.

[0053] In an example embodiment of the present disclosure, the short-term prediction model may refer to a model for predicting short-term demand data of a city to be predicted. For example, the short-term prediction model may be a time series model for predicting short-term demand data of a city to be predicted, such as an exponential smoothing model. The short-term prediction model may also be a time series model for predicting short-term demand data of a city to be predicted, such as a year-on-year model. The short-term prediction model may be a time series model for predicting short-term demand data of a city to be predicted, such as a moving average autoregressive model. Of course, the short-term prediction model may also be other models for predicting short-term demand data of a city to be predicted, and this example embodiment does not specifically limit this.

[0054] Preferably, the target order data of the city to be predicted can be input into multiple trained short-term prediction models, and the prediction results output by each short-term prediction model can be integrated to obtain the short-term demand data prediction results of the city to be predicted, thereby improving the stability and accuracy of the short-term demand data prediction results.

[0055] For example, the target order data of the city to be predicted can be input into the exponential smoothing model, the year-on-year model, and the moving average autoregression model respectively to obtain the first prediction result output by the exponential smoothing model, the second prediction result output by the year-on-year model, and the third prediction result output by the moving average autoregression model. Then, the average of the first prediction result, the second prediction result, and the third prediction result can be calculated, and the average value can be used as the short-term demand data prediction result of the city to be predicted. Of course, other operations can also be performed on the first prediction result, the second prediction result, and the third prediction result to improve the stability and accuracy of the short-term demand data prediction results.

[0056] In step S130, a long-term prediction model is constructed based on the target order data, the target order associated data, and the short-term demand data prediction results, and the target prediction time data is called, and the long-term demand data prediction results are output based on the long-term prediction model.

[0057] In an example embodiment of the present disclosure, the long-term prediction model may refer to a prediction model obtained by performing life cycle curve fitting on the target order data, target order associated data, and short-term demand data prediction results of the predicted city. For example, the long-term prediction model may be a deterministic prediction model obtained by performing life cycle curve fitting on the target order data, target order associated data, and short-term demand data prediction results of the predicted city. The long-term prediction model may also be a random prediction model obtained by performing life cycle curve fitting on the target order data, target order associated data, and short-term demand data prediction results of the predicted city. Of course, the long-term prediction model may refer to other prediction models obtained by performing life cycle curve fitting on the target order data, target order associated data, and short-term demand data prediction results of the predicted city. This example embodiment does not specifically limit this.

[0058] The target prediction time data may refer to the time data preset to be input into the long-term prediction model. For example, the target prediction time data may be the year time span data preset to be input into the long-term prediction model from the current year. The target prediction time data may also be the month time span data preset to be input into the long-term prediction model from the current month. Of course, the target prediction time data may also be other time span data preset to be input into the long-term prediction model. This embodiment does not specifically limit this.

[0059] In general, the express business data (order data and order-related data) of a city shows a trend of first increasing, then decreasing and gradually tending to zero. The growth process of the express business data of the city to be predicted can be regarded as a process that changes continuously over time, and the life cycle curve is fitted for the express business data of the city to build a long-term prediction model. However, directly fitting the life cycle curve with the city's order data may result in a large deviation of the fitting curve, resulting in poor long-term prediction results. Therefore, different data sets can be used to fit the model parameters of the long-term prediction model. For example, the target order-related data and the target order data, as well as the target order data, the short-term demand data prediction results, and the short-term real demand data corresponding to the short-term demand data prediction results can be used to fit the model parameters of the long-term prediction model respectively, so as to improve the accuracy of the model parameters of the long-term prediction model, thereby improving the accuracy of the long-term demand data prediction results of the city to be predicted.

[0060] In step S140, the short-term demand data prediction result and the long-term demand data prediction result are fused to obtain the target demand data prediction result.

[0061] In an example embodiment of the present disclosure, the target demand data prediction result may refer to the prediction data obtained based on the fused short-term prediction results of the demand data and the long-term prediction results of the demand data. For example, the short-term prediction results of the demand data may be the short-term order volume prediction results of the city to be predicted, and the long-term prediction results of the demand data may be the long-term order volume prediction results of the city to be predicted. The target demand data prediction result may be the prediction data obtained based on the fused short-term order volume prediction results and the long-term order volume prediction results of the city to be predicted. Of course, the short-term demand data prediction results may also be the short-term prediction results of other demand data of the city to be predicted, and the long-term demand data prediction results may also be the long-term prediction results of other demand data of the city to be predicted. The target demand data prediction result may also be the prediction data obtained based on the fused short-term prediction results of the demand data and the long-term prediction results of the demand data. This example embodiment does not make any special limitation on this.

[0062] Optionally, a target prediction time period related to the target demand data of the city to be predicted can be set. For example, the target prediction time period can be time span data from the current time, and the target prediction time period can be divided into short-term prediction time data and long-term prediction time data. For example, the target prediction time period can be 10 years, and the short-term prediction time data can be future time data for the first 3 years of the target prediction time period on the time axis, and the long-term prediction time data can be time data for the last 7 years of the target prediction time period on the time axis. Then, the short-term demand data prediction results 3 years later output by the short-term prediction model can be used as the basis for the long-term demand data prediction results 4 to 10 years from the current time, based on the short-term demand data prediction results 3 years later obtained by the long-term prediction model.

[0063] Since the long-term demand data forecast results may be based on a larger time granularity such as yearly granularity, while the short-term demand data forecast results may be based on a smaller time granularity such as monthly granularity, the long-term demand data forecast results may be split into monthly granularity. For example, the monthly granularity demand data forecast results corresponding to the long-term demand data forecast results may be estimated based on the proportion of monthly granularity historical order data one year earlier than the current time. Thus, the short-term demand data forecast results and the long-term demand data forecast results after monthly granularity splitting are spliced ​​to obtain the future demand data for each month in the next 10 years.

[0064] Figure 2 The following is a schematic diagram showing a data preprocessing method flow according to some embodiments of the present disclosure. Figure 2 As shown, the data preprocessing method may include the following steps:

[0065] In step S210, historical order data and historical order associated data are obtained;

[0066] In step S220, the historical order data is cleaned, and target order data is extracted from the cleaned historical order data;

[0067] In step S230, the correlation coefficient between the historical order associated data and the target order data is calculated, and the correlation coefficients are sorted to obtain weight data of the historical order associated data;

[0068] In step S240, target order associated data corresponding to the target order data is obtained by screening the historical order associated data according to the weight data.

[0069] Among them, historical order data may refer to historical express business data related to the city to be predicted. For example, the historical order data may be the historical express business data of the city to be predicted and cities with the same administrative level as the city to be predicted. The historical order data may also be the historical express business data of the city to be predicted and cities with the same economic development speed as the city to be predicted. Of course, the historical order data may also be other historical express business data related to the city to be predicted, and this embodiment does not make any special limitation to this.

[0070] The historical order-related data may refer to the influencing factor data of the historical order data. For example, the historical order-related data may be the regional influencing factor data corresponding to the historical order data of the city to be predicted and the city to assist in prediction, such as the gross regional product (GDP). The historical order-related data may also be the humanistic influencing factor data corresponding to the historical order data of the city to be predicted and the city to assist in prediction, such as the number of permanent residents. The historical order-related data may also be the technological level influencing factor data corresponding to the historical order data of the city to be predicted and the city to assist in prediction, such as the number of broadband access users and online retail sales. Of course, the historical order-related data may also be other influencing factor data corresponding to the historical order data of the city to be predicted and the city to assist in prediction, such as the city's administrative level. This embodiment does not make any special restrictions on this. Among them, the city to be predicted may refer to a city with the same economic strength or administrative level as the city to be predicted, and this embodiment does not make any special restrictions on this.

[0071] Weight data may refer to data used to measure the degree of influence of each historical order-related data on the target order data. For example, weight data may be data determined based on a weight coefficient of each historical order-related data and the target order data and used to measure the degree of influence of each historical order-related data on the target order data.

[0072] The historical order data and historical order-related data of the city to be predicted and the city assisting in the prediction can be obtained from the official website of the National Bureau of Statistics or the provincial statistics bureau of the city, and the historical order data and historical order-related data obtained with the city to be predicted are cleaned, such as performing anomaly detection on the historical order data related to the city to be predicted, mainly removing the historical order information with repeated records from the historical order data, and fitting the abnormally large order data of different categories with the empirical curve, setting the quantile to remove the abnormal historical order data exceeding the threshold corresponding to the quantile, and for the abnormal demand data caused by some extreme events, such as the sharp decline in the number of express delivery business orders in February caused by the epidemic, we backfill the historical order data in February to determine the target order data from the historical order data related to the city to be predicted. Among them, the backfill operation of the historical order data can adopt the interpolation method and the historical proportion method, for example, the historical data before and after can be used for fitting, and linear interpolation or polynomial interpolation can be performed, or the prediction can be based on the proportion of historical data in the same period and the current data, which is not specifically limited in this embodiment.

[0073] At the same time, the correlation coefficients between each historical order-related data and the target order data can be calculated respectively, and the weights of each historical order-related data can be determined according to the correlation coefficients of each historical order-related data, so as to screen out order-related data with larger weights from the historical order-related data related to the city to be predicted as the target order-related data, and use the target order-related data in the fitting of the long-term prediction model to improve the prediction accuracy of the long-term prediction model.

[0074] Figure 3 The following is a schematic diagram of a short-term demand data forecasting method process according to some embodiments of the present disclosure. Figure 3 As shown, the short-term demand data forecasting method may include the following steps:

[0075] In step S310, a pre-built first type time series model, a second type time series model, and a third type time series model are obtained;

[0076] In step S320, the target order data is respectively input into the first type time series model and the second type time series model to obtain a first short-term prediction result, a second short-term prediction result, and a third short-term prediction result;

[0077] In step S330, the first short-term prediction result, the second short-term prediction result, and the third short-term prediction result are integrated to obtain a short-term demand data prediction result.

[0078] Among them, the first type of time series model can refer to an exponential smoothing model. For example, the first type of time series model can be a single exponential model, and the first type of time series model can also be a double exponential model. Of course, the first type of time series model can also refer to other exponential smoothing models such as a triple exponential model. This embodiment does not make any special limitations on this.

[0079] The second type of time series model may refer to a prediction model with a time granularity ratio characteristic. For example, the second type may be a year-on-year model with a time granularity ratio characteristic, and the second type may be a year-on-year model with a time granularity ratio characteristic. Of course, the second type of time series model may also be other models with a time granularity ratio characteristic, and this embodiment does not make any special limitation to this.

[0080] The third type of time series model may refer to a time series model with autoregressive averaging characteristics. For example, the third type of time series model may be an AR model (auto regression model) with autoregressive averaging characteristics. The third type of time series model may also be an MA model (moving average model) with autoregressive averaging characteristics. Of course, the third type of time series model may also be other models with autoregressive averaging characteristics such as an ARMA model (auto regression moving average model). This embodiment does not specifically limit this.

[0081] Using the target order data of the city to be predicted and the assisting prediction city as the input data source of the short-term model avoids the situation where the order data of a single city is limited and the advantages of the short-term prediction model cannot be brought into play, thereby improving the utilization rate of the short-term prediction model; at the same time, the first output result of the first type of time series model, the second output result of the second type of time series model, and the third output result of the third type of time series model can be integrated and processed. For example, the first output result of the first type of time series model, the second output result of the second type of time series model, and the third output result of the third type of time series model are averaged, and the average of the three is used as the short-term demand data prediction result of the city to be predicted, which can improve the stability and accuracy of the short-term demand data prediction result of the city to be predicted.

[0082] In an example embodiment of the present disclosure, a pre-built initial long-term prediction model can be obtained, and the initial long-term prediction model can be trained based on target order data, target order-related data, and short-term demand data prediction results to obtain model parameters of the trained initial long-term prediction model, and a long-term prediction model can be constructed based on the model parameters.

[0083] Among them, the initial long-term prediction model may refer to a long-term prediction model obtained based on life cycle curve fitting. For example, the initial long-term prediction model may be a deterministic long-term prediction model obtained based on life cycle curve fitting. Of course, the initial long-term prediction model may also be a stochastic long-term prediction model obtained based on life cycle curve fitting. This embodiment does not make any special limitation on this.

[0084] The model parameters may refer to parameters corresponding to the logistic regression function of the initial long-term prediction model trained by performing life cycle curve fitting on the target order data, the target order-related data, and the short-term demand data prediction results. For example, the model parameters may be equation coefficients corresponding to the logistic regression function of the initial long-term prediction model trained by performing life cycle curve fitting on the target order data, the target order-related data, and the short-term demand data prediction results. The model parameters may also be function equations corresponding to the logistic regression function of the initial long-term prediction model trained by performing life cycle curve fitting on the target order data, the target order-related data, and the short-term demand data prediction results. Of course, the model parameters may also be other parameters corresponding to the logistic regression function of the initial long-term prediction model trained by performing life cycle curve fitting on the target order data, the target order-related data, and the short-term demand data prediction results. This embodiment does not make any special limitations on this.

[0085] Since the growth trend of the city's order data conforms to the law of the "S"-shaped life cycle curve, which is to increase first, then decrease and gradually tend to zero, the growth process of the city's demand data can be regarded as a process that changes continuously over time. According to the relationship between demand data (Q(t)) and time (t), it can be determined that the demand data per unit time can be expressed by the following relational expression:

[0086]

[0087] Solving the calculation relation expression (1), we can obtain:

[0088]

[0089] The above relational expression (2) can be transformed to obtain:

[0090]

[0091] in, K represents the upper limit of Q(t), a is the scale parameter, which represents the dispersion of the distribution curve; b is the location parameter, which represents the degree of center deviation of the distribution curve; when The value of t is b, and the relational expression (3) can be used as the logistic regression function of the initial long-term prediction model.

[0092] From the above calculation expression, it can be seen that the model parameters a, b, and K in the initial long-term prediction model satisfy the relationship expression between a, b, and K in the life cycle curve. Furthermore, the initial long-term prediction model can be trained based on the target order data, target order association data, and short-term demand data prediction results of the city to be predicted, and the long-term prediction model can be constructed based on the model parameters of the trained initial long-term prediction model.

[0093] Figure 4 A schematic diagram of a long-term prediction model construction architecture according to some embodiments of the present disclosure is schematically shown; the long-term prediction model construction architecture 400 mainly includes a data fusion module 410, a model parameter estimation module 420, and a model parameter integration module 430.

[0094] Among them, the data fusion module 410 is mainly used to fuse the target order data and the target order related data, so that the fused target order data constitutes the target scalar data set and the target order related data constitutes the regression variable data set, and constructs a linear regression model through data fitting. At the same time, it is also used to fuse the target order data and the short-term prediction results, so that the fused target order data and the short-term prediction results are substituted into the loss function of the trained initial long-term prediction model to estimate the model parameters. The model parameter estimation module 420 is mainly used to estimate the parameter K in the logistic regression function of the trained initial long-term prediction model according to the fused target order data and the target order related data. At the same time, it can also estimate a and b in the logistic regression function of the trained initial long-term prediction model according to the fused target order data and the short-term prediction results. The model parameter integration module 430 is mainly used to integrate the logistic regression function of the initial long-term training model and the model parameters obtained after training the initial long-term prediction model based on the target order data, the target order related data, and the short-term prediction result data, and construct the logistic regression function of the long-term prediction model.

[0095] The target order data and the target order-related data can be fused to estimate a model parameter of the initial long-term prediction model that has been trained. The target order data and the short-term demand data prediction results can also be fused to estimate the other two model parameters of the initial long-term prediction model that has been trained. This avoids the problem of low accuracy in model parameter estimation due to excessive data concentration. At the same time, by incorporating the short-term demand data prediction results into the long-term prediction model construction process, the data source for constructing the long-term prediction model is expanded, avoiding the problem of being unable to perform long-term prediction due to fewer data sources, and improving the utilization rate of the target order data, the target order-related data, and the short-term demand data prediction results.

[0096] Figure 5The following is a schematic diagram of a model parameter determination method flow according to some embodiments of the present disclosure. Figure 5 As shown, the model parameter determination method may include the following steps:

[0097] In step S510, short-term real demand data corresponding to the target order data is determined, and key order-related data with slow data changes are screened from the target order-related data;

[0098] In step S520, a linear regression curve is fitted on the target order data and the target order associated data to obtain a linear regression model, and the key order associated data is input into the linear regression model to obtain a first model parameter;

[0099] In step S530, the loss function of the initial long-term prediction model is called, and the short-term demand data prediction results and the short-term actual demand data are input into the loss function to obtain the second model parameters, and the first model parameters and the second model parameters are used as the model parameters of the trained initial long-term prediction model.

[0100] Among them, the short-term real demand data may refer to the demand data obtained by aggregating the target order data with the same time granularity as the short-term demand data forecast result. For example, the time granularity of the short-term demand data forecast result is the monthly granularity, and the short-term real demand data may be the demand data obtained by aggregating the target order data of the same month. The time granularity of the short-term demand data forecast result may also be the annual granularity, and the short-term real demand data may also be the demand data obtained by aggregating the target order data of the same year. Of course, the short-term demand data forecast result may also be other time granularities, and the short-term real demand data may also be other demand data with the same time granularity as the short-term demand data forecast result obtained by aggregating the target order data. This embodiment does not make any special limitation on this.

[0101] The key order-related data may refer to the order-related data that has entered a stable development period obtained by screening out the target order-related data. For example, the key order-related data may be the permanent population data that has entered a stable development period obtained by screening out the target order-related data. The key order-related data may also be the consumer price index that has entered a stable development period obtained by screening out the target order-related data. Of course, the key order-related data may also be other order-related data that has entered a stable development period obtained by screening out the target-related order data, such as the total retail sales of consumer goods. This embodiment does not specifically limit this.

[0102] The first model parameter may refer to the model parameter of the linear regression model obtained by fitting the linear regression curve of the target order data and the target association data, and the second model parameter may refer to the model parameter obtained by combining the first model parameter and inputting the short-term real demand data and the short-term demand data prediction results into the loss function of the trained initial long-term prediction model.

[0103] The loss function may refer to a function used to measure the prediction accuracy of the model. For example, the loss function may be a square error function used to measure the prediction accuracy of the model. The loss function may also be a mean square error function used to measure the prediction accuracy of the model. The loss function may also be a mean absolute error function used to measure the prediction accuracy of the model. Of course, the loss function may also be a function used by other users to measure the prediction accuracy of the model. This embodiment does not specifically limit this.

[0104] For example, to predict the long-term future demand data of City A, it is necessary to build a long-term prediction model, and then train the initial long-term prediction model based on the target order data, target order related data, and short-term demand data prediction results. In the early stage of model training, it is necessary to fuse the target order data and target order related data, as well as the short-term demand data prediction results and short-term real demand data. For example, the target order data of City A Y = (y1, y2, ..., y t ), and the target order data Y'=(y1',y2',...,y t ') constitutes the target scalar data set, and the target order association data X1=(x 11 ,x 12 ,,...,x 1t ),X2=(x 21 ,x 22 ,,...,x 2t ),....,X a =(x a1 ,x a2 ,,...,x at ), and the target order data of city B X1'=(x 11 ',x 12 ,,...,x 1t '),X2'=(x 21 ',x 22 ',,...,x 2t '),....,X a '=(x a1 ',x a2 ',,...,x at') constitutes a regression variable data set, and a linear regression model is constructed by data fitting. The model function of the linear regression model is shown in relational expression (4):

[0105] y=d1x1+d2x2+...+d a x a +d0 (4)

[0106] By detecting the target order-related data related to city A, it is determined that the key order-related data of city A entering the stable development period is X * =(X1 * ,X2 * ,...,X a * ), X * =(X1 * ,X2 * ,...,X a * ) is substituted into relational expression (4) to obtain the first model parameter, which can be expressed by relational expression (5):

[0107] K=d1X1 * +d2X2 * +...+d a X a * +d0 (5)

[0108] Among them, d1, d2, d a , d0 is the coefficient of the regression function of the linear regression model, and the number of coefficients of the linear regression model is the same as the number of target order associated data.

[0109] At the same time, after obtaining the first model parameters, the target order data and the short-term demand data prediction results can be fused to estimate the second model parameters of the trained initial long-term prediction model; for example, the target order number is aggregated to obtain the short-term real demand data with the same time granularity as the short-term demand data prediction results, and the short-term real demand data and the short-term demand data prediction results are substituted into the loss function of the trained initial long-term prediction model to estimate the second model parameters. Among them, the loss function of the trained initial long-term prediction model can be described by the relational expression (6), as follows:

[0110]

[0111] By converting the short-term real demand data y t Substituting the short-term demand data forecast result Q(t) into the above relational expression (6), and based on the relational expression between a and b, the values ​​of a and b can be estimated, and then the second model parameters can be obtained.

[0112] In an example embodiment of the present disclosure, the target prediction time data can be divided into stages to determine the long-term prediction duration data in the target prediction time data; the long-term prediction duration data is input into the prediction model to obtain the long-term demand data prediction result.

[0113] Among them, long-term prediction time data can refer to the time data with longer stages obtained by dividing the target prediction time period into stages. For example, the target prediction time data can be the time data with a span of 10 years from the current time. The target prediction time data can be divided into short-term prediction time data such as the time data with a span of 3 years from the current time. The long-term prediction time data can be the time data with a span of 7 years based on the short-term prediction time data. Of course, the long-term prediction time data can also be the time data with longer stages obtained by dividing other target prediction time data into stages. This embodiment does not make any special limitation on this.

[0114] After completing the construction of the long-term prediction model, the logistic regression function of the long-term prediction model can be obtained based on the model parameters of the long-term prediction model, and then the long-term prediction duration data can be substituted into the logistic regression function to obtain the long-term demand data prediction results of the city to be predicted.

[0115] Figure 6 The following is a schematic diagram of the prediction result fusion method process according to some embodiments of the present disclosure. Figure 6 As shown, the prediction result fusion method may include the following steps:

[0116] In step S610, short-term time granularity data of the short-term demand data forecast result and long-term time granularity data of the long-term demand data forecast result are determined;

[0117] In step S620, the short-term time granularity data and the long-term time granularity data are uniformly processed to obtain standard time granularity data;

[0118] In step S630, the short-term demand data prediction results and the long-term demand data prediction results are granularly converted based on the standard time granularity data, and the short-term demand data prediction results and the long-term demand data prediction results after granularity conversion are merged to obtain the target demand data prediction results.

[0119] Short-term time granularity data may refer to granularity data corresponding to the time unit of short-term demand data prediction. For example, the time unit of short-term demand data prediction may be month, and the short-term time granularity data may be month granularity data corresponding to the time unit of short-term demand data prediction. The time unit of short-term demand data prediction may also be day, and the short-term time granularity data may also be day granularity data corresponding to the time unit of short-term demand data prediction. Of course, the short-term time granularity data may also be other granularity data corresponding to the time unit of short-term demand data prediction, such as year, and this embodiment does not make any special limitation on this.

[0120] Long-term time granularity data may refer to granularity data corresponding to the time unit of long-term demand data prediction. For example, the time unit of long-term demand data prediction may be a single year, and the long-term time granularity data may be year granularity data corresponding to the time unit of long-term demand data prediction. The time unit of long-term demand data prediction may also be half a year, and the long-term time granularity data may also be half-year granularity data corresponding to the time unit of long-term demand data prediction. Of course, the long-term time granularity data may also be other granularity data corresponding to the time unit of long-term demand data prediction, and this embodiment does not make any special limitation on this.

[0121] Standard time granularity data may refer to time granularity data obtained by uniformly processing the time granularity data of long-term demand data forecast results and short-term demand data forecast results. For example, the long-term time granularity data related to the long-term demand data forecast results and the short-term time granularity data related to the short-term demand data forecast results may be uniformly converted into month granularity data, and the standard time granularity data may be month granularity data. The long-term time granularity data related to the long-term demand data forecast results and the short-term time granularity data related to the short-term demand data forecast results may be uniformly converted into year granularity data. The standard time granularity data may be year granularity data. Of course, the standard time granularity data may also be other time granularity data obtained by uniformly processing the time granularity data of long-term demand data forecast results and short-term demand data forecast results. This embodiment does not make any special limitation to this.

[0122] For example, the short-term time granularity data corresponding to the short-term demand data forecast results may be monthly data, while the long-term time granularity data corresponding to the long-term demand data forecast results may be yearly data. The long-term demand data forecast results may be split into monthly granularity based on the year-on-year idea. For example, based on the proportion of monthly granularity historical data in the previous year, the long-term demand data forecast results may be split month by month to obtain long-term demand data forecast results with monthly granularity. Furthermore, the short-term demand data forecast results with monthly granularity and the long-term demand data forecast results may be spliced ​​to obtain the future demand data of the city to be predicted within the target prediction time period. Of course, the short-term duration data corresponding to the short-term demand data can also be aggregated into year-granularity data. For example, the target prediction time period of the city to be predicted is a future time period with a time span of 10 years from this year on the time axis, the short-term prediction time data is the time span data with a time span of 3 years from this year on the time axis, and the long-term prediction time data is the time span data of the last 7 years of the target prediction time period on the time axis. Then, the short-term demand data prediction results and the long-term demand data prediction results can be spliced ​​to obtain the demand data prediction results for the city to be predicted in the next 10 years.

[0123] It should be noted that, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0124] In addition, in this exemplary embodiment, a demand data prediction device is also provided. Figure 7 As shown, the demand data prediction device 700 includes: a data collection module 710, a short-term demand data prediction result determination module 720, a long-term demand data prediction result determination module 730, and a prediction result fusion module 740. Among them: the data collection module 710 is used to collect target order data and target order related data; the short-term demand data prediction result determination module 720 is used to obtain a pre-built short-term prediction model, and input the target order data into the short-term prediction model to obtain a short-term demand data prediction result; the long-term demand data prediction result determination module 730 is used to fit the target order data, the target order related data, and the short-term demand data prediction result to obtain a long-term prediction model, and call the target prediction time data, and output the long-term demand data prediction result based on the long-term prediction model according to the target prediction time data; the prediction result fusion module 740 is used to fuse the short-term demand data prediction result and the long-term demand data prediction result to obtain the target demand data prediction result.

[0125] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the data collection module 710 includes a data preprocessing unit, and the data preprocessing unit is used to obtain historical order data and historical order associated data; clean the historical order data, and extract the target order data from the cleaned historical order data; calculate the correlation coefficient between the historical order associated data and the target order data, and sort the correlation coefficient to obtain weight data of the historical order associated data; and filter the target order associated data corresponding to the target order data from the historical order associated data according to the weight data.

[0126] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the short-term demand data forecast result determination module 720 includes a short-term demand data forecast result determination unit, and the short-term demand data forecast result determination unit is used to obtain the pre-built first type time series model, the second type time series model, and the third type time series model; input the target order data into the first type time series model, the second type time series model, and the second type time series model respectively to obtain a first short-term forecast result, a second short-term forecast result, and a third short-term forecast result; and integrate the first short-term forecast result, the second short-term forecast result, and the third short-term forecast result to obtain the short-term demand data forecast result.

[0127] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the long-term demand data prediction result determination module 730 includes a long-term prediction model construction unit, and the long-term prediction model construction unit is used to obtain a pre-constructed initial long-term prediction model, and train the initial long-term prediction model based on the target order data, the target order associated data, and the short-term demand data prediction results to obtain the model parameters of the trained initial long-term prediction model; and construct the long-term prediction model based on the model parameters.

[0128] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the long-term demand data prediction result determination module 730 also includes a model parameter determination unit, the model parameter determination unit is used to determine the short-term real demand data corresponding to the target order data, and to filter the key order-related data with slow data changes from the target order-related data; perform linear regression curve fitting on the target order data and the target order-related data to obtain a linear regression model, and input the key order-related data into the linear regression model to obtain the first model parameters; call the loss function of the initial long-term prediction model, and input the short-term demand data prediction result and the short-term real demand data into the loss function to obtain the second model parameters, and use the first model parameters and the second model parameters as the model parameters of the trained initial long-term prediction model.

[0129] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the long-term demand data prediction result determination module 730 also includes a long-term demand data prediction result determination unit, and the long-term demand data prediction result determination unit is used to divide the target prediction time period into stages and determine the long-term prediction duration data in the target prediction time period; input the long-term prediction duration data into the long-term prediction model to obtain the long-term demand data prediction result.

[0130] In some example embodiments of the present disclosure, based on the aforementioned scheme, it is characterized in that the prediction data fusion module 740 includes a prediction result fusion unit, and the prediction result fusion unit is used to determine the short-term time granularity data of the short-term demand data prediction result, and the long-term time granularity data of the long-term demand data prediction result; perform time granularity uniform processing on the short-term time granularity data and the long-term time granularity data to obtain standard time granularity data; perform granularity conversion on the short-term demand data prediction result and the long-term demand data prediction result based on the standard time granularity data, and fuse the short-term demand data prediction result and the long-term demand data prediction result after granularity conversion to obtain the target demand data prediction result.

[0131] The specific details of each module of the above-mentioned demand data prediction device have been described in detail in the corresponding demand data prediction method, so they will not be repeated here.

[0132] It should be noted that although several modules or units of the demand data forecasting device are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be concretized in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.

[0133] In addition, in an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above-mentioned demand data prediction method is also provided.

[0134] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented as systems, methods or program products. Therefore, various aspects of the present disclosure may be specifically implemented in the following forms, namely: complete hardware embodiments, complete software embodiments (including firmware, microcode, etc.), or embodiments combining hardware and software aspects, which may be collectively referred to herein as "circuits", "modules" or "systems".

[0135] Refer to the following Figure 8 800 according to this embodiment of the present disclosure is described. Figure 8 The electronic device 800 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0136] like Figure 8 As shown, the electronic device 800 is in the form of a general computing device. The components of the electronic device 800 may include, but are not limited to: the at least one processing unit 810, the at least one storage unit 820, a bus 830 connecting different system components (including the storage unit 820 and the processing unit 810), and a display unit 840.

[0137] The storage unit stores a program code, which can be executed by the processing unit 810, so that the processing unit 810 performs the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of this specification. For example, the processing unit 810 can perform the following steps: Figure 1In step S110 shown in the figure, target order data and target order-related data are collected; in step S120, a pre-built short-term forecasting model is obtained, and the target order data is input into the short-term forecasting model to obtain a short-term demand data forecasting result; in step S130, a long-term forecasting model is constructed based on the target order data, the target order-related data, and the short-term demand data forecasting result, and the target forecasting time data is called, and the long-term demand data forecasting result is outputted based on the long-term forecasting model according to the target forecasting time data; in step S140, the short-term demand data forecasting result and the long-term demand data forecasting result are merged to obtain a target demand data forecasting result.

[0138] The storage unit 820 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 821 and / or a cache memory unit 822 , and may further include a read-only memory unit (ROM) 823 .

[0139] The storage unit 820 may also include a program / utility 824 having a set (at least one) of program modules 825, such program modules 825 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0140] Bus 830 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0141] The electronic device 800 may also communicate with one or more external devices 870 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 800, and / or communicate with any device that enables the electronic device 800 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 850. Furthermore, the electronic device 800 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 860. As shown, the network adapter 860 communicates with other modules of the electronic device 800 via a bus 830. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 800, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0142] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiment of the present disclosure.

[0143] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible embodiments, various aspects of the present disclosure may also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to enable the terminal device to perform the steps according to various exemplary embodiments of the present disclosure described in the above "Exemplary Method" section of the present specification.

[0144] refer to Fig. 9 As shown, a program product 900 for implementing the above-mentioned demand data forecasting method according to an embodiment of the present disclosure is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0145] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0146] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0147] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0148] Program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0149] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0150] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiment of the present disclosure.

[0151] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

[0152] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A demand data forecasting method, characterized in that: include: Collecting target order data and target order related data; Obtaining a pre-built short-term forecasting model, and inputting the target order data into the short-term forecasting model to obtain a short-term demand data forecast result; Obtaining a pre-built initial long-term forecasting model, determining short-term real demand data corresponding to the target order data, and screening key order-related data with slow data changes from the target order-related data; Performing linear regression curve fitting on the target order data and the target order associated data to obtain a linear regression model, and inputting the key order associated data into the linear regression model to obtain a first model parameter; Calling the loss function of the initial long-term prediction model, and inputting the short-term demand data prediction result and the short-term real demand data into the loss function to obtain second model parameters, and using the first model parameters and the second model parameters as model parameters of the trained initial long-term prediction model; constructing a long-term prediction model based on the model parameters; Calling the target forecast time data, and outputting the long-term demand data forecast result according to the target forecast time data based on the long-term forecast model; The short-term demand data prediction result and the long-term demand data prediction result are integrated to obtain a target demand data prediction result.

2. The demand data forecasting method according to claim 1, characterized in that: The collecting of target order data and target order associated data includes: Obtain historical order data and historical order related data; Cleaning the historical order data, and extracting the target order data from the cleaned historical order data; Calculating the correlation coefficient between the historical order associated data and the target order data, and sorting the correlation coefficients to obtain weight data of the historical order associated data; The target order associated data corresponding to the target order data is obtained by screening the historical order associated data according to the weight data.

3. The demand data forecasting method according to claim 1, characterized in that: The short-term forecasting model includes a first type of time series model, a second type of time series model, and a third type of time series model. The obtaining of the pre-built short-term forecasting model and inputting the target order data into the short-term forecasting model to obtain a short-term demand data forecasting result includes: Obtaining the pre-built first-type time series model, the second-type time series model, and the third-type time series model; Input the target order data into the first type time series model, the second type time series model, and the second type time series model respectively to obtain a first short-term prediction result, a second short-term prediction result, and a third short-term prediction result; The first short-term forecast result, the second short-term forecast result, and the third short-term forecast result are integrated to obtain the short-term demand data forecast result.

4. The demand data forecasting method according to claim 1, characterized in that: The outputting of the long-term demand data prediction result based on the long-term prediction model and according to the target prediction time data comprises: Performing stage division processing on the target prediction time data to determine long-term prediction duration data in the target prediction time data; The long-term prediction duration data is input into the long-term prediction model to obtain the long-term demand data prediction result.

5. The demand data forecasting method according to claim 1, characterized in that: The step of fusing the short-term demand data prediction result and the long-term demand data prediction result to obtain a target demand data prediction result includes: Determining short-term time granularity data of the short-term demand data forecast result and long-term time granularity data of the long-term demand data forecast result; Performing unified time granularity processing on the short-term time granularity data and the long-term time granularity data to obtain standard time granularity data; The short-term demand data prediction result and the long-term demand data prediction result are granularized based on the standard time granularity data, and the short-term demand data prediction result and the long-term demand data prediction result after granularity conversion are merged to obtain the target demand data prediction result.

6. A demand data forecasting device, characterized in that: include: A data collection module, used for collecting target order data and target order related data; Short-term demand data forecast result determination module is used to Obtaining a pre-built short-term forecasting model, and inputting the target order data into the short-term forecasting model to obtain a short-term demand data forecast result; A long-term demand data forecast result determination module is used to obtain a pre-built initial long-term forecast model, determine the short-term real demand data corresponding to the target order data, and filter the key order-related data with slow data changes from the target order-related data; Performing linear regression curve fitting on the target order data and the target order associated data to obtain a linear regression model, and inputting the key order associated data into the linear regression model to obtain a first model parameter; Calling the loss function of the initial long-term prediction model, and inputting the short-term demand data prediction result and the short-term real demand data into the loss function to obtain second model parameters, and using the first model parameters and the second model parameters as model parameters of the trained initial long-term prediction model; constructing a long-term prediction model based on the model parameters; Calling the target forecast time data, and outputting the long-term demand data forecast result according to the target forecast time data based on the long-term forecast model; The prediction result fusion module is used to fuse the short-term demand data prediction result and the long-term demand data prediction result to obtain the target demand data prediction result.

7. An electronic device comprising: processor; as well as A memory having computer-readable instructions stored thereon, wherein the computer-readable instructions, when executed by the processor, implement the demand data forecasting method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the demand data forecasting method according to any one of claims 1 to 5.

Images