Method and device for determining freight rate product demand quantity, and electronic equipment

Abstract

The invention discloses a freight rate product demand quantity determination method and device and electronic equipment, and relates to the technical field of freight rate demand prediction.The determination method comprises the steps that historical synchronous data and current synchronous data are collected, an initial prediction result is determined by adopting a preset prediction model based on the historical synchronous data and the current synchronous data, and a freight rate product demand quantity is determined according to the initial prediction result. And determining a Kalman gain value based on the previous covariance matrix and the previous model parameters, and determining the predicted demand quantity of the target freight rate product in the target time period in the current sales period based on the Kalman gain value, the current synchronous data and the initial prediction result. The technical problem that the demand quantity of the freight rate product cannot be accurately predicted in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of freight demand forecasting, and in particular to a method for determining the demand for freight products, a device thereof, and electronic equipment. Background Art

[0002] With the rapid development of the civil aviation industry, airlines are faced with the problem of meeting the differentiated needs of different passenger groups. To achieve this goal, airlines need to release a reasonable fare product portfolio at the right time to increase market share and increase revenue. Currently, airlines mainly manage route revenue through differentiated fare management and corresponding flight seat control methods. Among them, differentiated fare management includes: demand forecasting and differential pricing, while flight seat control methods include: seat optimization and flight overbooking. However, the accuracy of demand forecasting is crucial to the product management and revenue of airlines, and accurate demand forecasting has always been a problem that needs to be solved in the industry.

[0003] In the related technologies, the methods for civil aviation demand forecasting mainly include: (1) simulating the point-to-point flight structure through the flight node optimization algorithm and calculating the expected marginal seat revenue model; (2) using network optimization algorithms such as the greedy virtual nesting method, shadow price method / permutation virtual nesting method and bidding method to simulate the hub-and-spoke network structure for network demand forecasting.

[0004] However, the above models and algorithms are mainly used to predict the demand for civil aviation fare products with a high degree of product standardization and a relatively simple route layout structure. For fare products with complex route situations, for example, fare products involving routes with a multi-center spoke structure, or fare products involving regional routes and routes between spoke centers, the above models and algorithms cannot accurately predict the demand for fare products.

[0005] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0006] The embodiment of the present invention provides a method and device for determining the demand for a freight rate product, and an electronic device, so as to at least solve the technical problem in the related art that the demand for a freight rate product cannot be accurately predicted.

[0007] According to one aspect of an embodiment of the present invention, a method for determining demand for a freight rate product is provided, comprising: collecting historical contemporaneous data and current contemporaneous data, wherein the historical contemporaneous data refers to historical sales data of the target freight rate product that is consistent with the current sales period of the target freight rate product, and the current contemporaneous data refers to sales data generated by the target freight rate product during the current sales period; based on the historical contemporaneous data and the current contemporaneous data, using a preset prediction model to determine an initial prediction result, wherein the preset prediction model is a prediction model that is iterated when predicting the previous period demand for the target freight rate product, and the preset prediction model includes: previous period model parameters; determining a Kalman gain value based on a previous period covariance matrix and the previous period model parameters; and determining a predicted demand for the target freight rate product in a target time period within the current sales period based on the Kalman gain value, the current contemporaneous data, and the initial prediction result.

[0008] Furthermore, before collecting historical data for the same period and current data for the same period, it also includes: collecting historical data within a historical time period, and annotating the historical data to obtain annotation information, wherein the historical data includes: historical sales data of each freight product, and the annotation information includes: actual sales values ​​in different historical sub-time periods, and the historical sub-time periods are sub-time periods obtained by dividing the historical time period; using the historical data and the annotation information to train an initial prediction model to obtain the preset prediction model, wherein when the preset prediction model is first applied, an initial covariance matrix and initial model parameters are used to predict the demand for the freight product; during each prediction process, the preset prediction model iteratively updates the initial covariance matrix and the initial model parameters to obtain the covariance matrix and model parameters used for the next prediction; the initial covariance matrix is ​​a covariance matrix obtained based on the statistics of the historical data, and the initial model parameters are model parameters obtained after the initial prediction model is trained.

[0009] Furthermore, the step of collecting historical data for the same period and current data for the same period includes: determining the current sales period and the target time period of the target freight rate product, wherein the target time period is any time period within the current sales period that has not been reached; collecting the historical sales data of the target freight rate product, and intercepting data within a period consistent with the current sales period from the historical sales data to obtain the historical data for the same period; when the target freight rate product is on sale, collecting data within a preset time period range that is before the target time period and consistent with the length of the target time period to obtain the current data for the same period, wherein the preset time period range is a time period range within the current sales period.

[0010] Furthermore, the previous period model parameters include: a first previous period coefficient matrix, a second previous period coefficient matrix, and based on the historical data for the same period and the current data for the same period, a preset prediction model is used to determine the step of initial prediction results, including: determining the historical average sales value based on the historical data for the same period, and determining the impact value of the current abnormal event based on the current data for the same period; determining the initial prediction result based on the first previous period coefficient matrix, the second previous period coefficient matrix, the historical average sales value and the impact value, wherein the initial prediction result includes: an initial prediction value.

[0011] Furthermore, the previous period model parameters also include: a third previous period coefficient matrix, a previous period systematic error, and a previous period preset error, wherein the previous period preset error includes at least: a previous period estimation error and a previous period measurement error. Based on the previous period covariance matrix and the previous period model parameters, the step of determining the Kalman gain value includes: obtaining the previous period covariance matrix, wherein the previous period covariance matrix is ​​a covariance matrix obtained when predicting the previous period demand for the target freight rate product; constructing a systematic error covariance matrix of the previous period systematic error, and constructing a preset error covariance matrix of the previous period preset error; constructing an estimated covariance matrix based on the previous period covariance matrix, the first previous period coefficient matrix, and the systematic error covariance matrix; and determining the Kalman gain value based on the third previous period coefficient matrix, the estimated covariance matrix, and the preset error covariance matrix.

[0012] Furthermore, the step of determining the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial forecast result includes: determining the current average sales value based on the current contemporaneous data; determining the predicted demand based on the Kalman gain value, the current average sales value and the initial forecast value included in the initial forecast result.

[0013] Furthermore, after determining the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial forecast result, it also includes: updating the previous period covariance matrix based on the estimated covariance matrix, the third previous period coefficient matrix used by the preset forecasting model and the Kalman gain value to obtain the current period covariance matrix, wherein the current period covariance matrix is ​​used to predict the next period demand for the target freight product in the next time period after the target time period within the current sales period.

[0014] Furthermore, after determining the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial forecast result, it also includes: obtaining the current true value of the target freight product; and when the absolute error value between the current true value and the predicted demand is greater than a preset error threshold, correcting the Kalman gain value used in forecasting the next period demand for the target freight product.

[0015] According to another aspect of an embodiment of the present invention, there is also provided a device for determining demand for a freight rate product, comprising: a collection unit for collecting historical contemporaneous data and current contemporaneous data, wherein the historical contemporaneous data refers to historical sales data of the target freight rate product consistent with the current sales period of the target freight rate product, and the current contemporaneous data refers to sales data generated by the target freight rate product in the current sales period; a first determination unit for determining an initial prediction result based on the historical contemporaneous data and the current contemporaneous data using a preset prediction model, wherein the preset prediction model is a prediction model iterated when predicting the previous period demand for the target freight rate product, and the preset prediction model includes: previous period model parameters; a second determination unit for determining a Kalman gain value based on a previous period covariance matrix and the previous period model parameters; and a third determination unit for determining the predicted demand for the target freight rate product in a target time period in the current sales period based on the Kalman gain value, the current contemporaneous data and the initial prediction result.

[0016] Furthermore, the determination device also includes: a first acquisition module, which is used to collect historical data within a historical time period before collecting historical data of the same period and current data of the same period, and mark the historical data to obtain marking information, wherein the historical data includes: historical sales data of each freight product, and the marking information includes: actual sales values ​​in different historical sub-time periods, and the historical sub-time periods are sub-time periods obtained by dividing the historical time period; a first training module, which is used to train an initial prediction model using the historical data and the marking information to obtain the preset prediction model, wherein the preset prediction model uses an initial covariance matrix and initial model parameters to predict the demand for the freight product when it is first applied; the preset prediction model iteratively updates the initial covariance matrix and the initial model parameters in each prediction process to obtain the covariance matrix and model parameters used for the next prediction; the initial covariance matrix is ​​a covariance matrix obtained based on the statistics of the historical data, and the initial model parameters are model parameters obtained after the training of the initial prediction model is completed.

[0017] Furthermore, the collection unit includes: a first determination module, used to determine the current sales period and the target time period of the target freight rate product, wherein the target time period is any time period within the current sales period that has not been reached; a second collection module, used to collect the historical sales data of the target freight rate product, and intercept data within the period consistent with the current sales period from the historical sales data to obtain the historical contemporaneous data; a third collection module, used to collect data within a preset time period range that is before the target time period and consistent with the length of the target time period when the target freight rate product is sold, to obtain the current contemporaneous data, wherein the preset time period range is a time period range within the current sales period.

[0018] Furthermore, the previous period model parameters include: a first previous period coefficient matrix, a second previous period coefficient matrix, and the first determination unit includes: a second determination module, used to determine the historical average sales value based on the historical data of the same period, and to determine the impact value of the current abnormal event based on the current data of the same period; a third determination module, used to determine the initial prediction result based on the first previous period coefficient matrix, the second previous period coefficient matrix, the historical average sales value and the impact value, wherein the initial prediction result includes: an initial prediction value.

[0019] Furthermore, the previous model parameters also include: a third previous coefficient matrix, a previous systematic error, and a previous preset error, wherein the previous preset error includes at least: a previous estimation error and a previous measurement error, and the second determination unit includes: a first acquisition module, used to acquire the previous covariance matrix, wherein the previous covariance matrix is ​​a covariance matrix obtained when predicting the previous demand for the target freight product; a first construction module, used to construct a systematic error covariance matrix of the previous systematic error, and to construct a preset error covariance matrix of the previous preset error; a second construction module, used to construct an estimated covariance matrix based on the previous covariance matrix, the first previous coefficient matrix and the systematic error covariance matrix; and a fourth determination module, used to determine the Kalman gain value based on the third previous coefficient matrix, the estimated covariance matrix and the preset error covariance matrix.

[0020] Furthermore, the third determination unit includes: a fifth determination module, used to determine the current average sales value based on the current contemporaneous data; a sixth determination module, used to determine the predicted demand based on the Kalman gain value, the current average sales value and the initial prediction value included in the initial prediction result.

[0021] Furthermore, the determination device also includes: a first updating module, which is used to determine the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial prediction result, and then update the previous period covariance matrix based on the estimated covariance matrix, the third previous period coefficient matrix used by the preset prediction model and the Kalman gain value to obtain a current period covariance matrix, wherein the current period covariance matrix is ​​used to predict the next period demand for the target freight product in the next time period after the target time period within the current sales period.

[0022] Furthermore, the determination device also includes: a second acquisition module, which is used to obtain the current true value of the target freight product after determining the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial prediction result; and a first correction module, which is used to correct the Kalman gain value used in predicting the next period demand for the target freight product when the absolute error value between the current true value and the predicted demand is greater than a preset error threshold.

[0023] According to another aspect of an embodiment of the present invention, there is also provided an electronic device, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the above-mentioned methods for determining the demand for freight rate products.

[0024] In the present invention, by collecting historical data for the same period and current data for the same period, based on the historical data for the same period and current data for the same period, a preset prediction model is used to determine the initial prediction result, based on the covariance matrix of the previous period and the model parameters of the previous period, the Kalman gain value is determined, based on the Kalman gain value, the current data for the same period and the initial prediction result, the predicted demand for the target freight product in the target time period within the current sales period is determined, thereby solving the technical problem in the related art that the demand for the freight product cannot be accurately predicted.

[0025] In the present invention, the historical sales data of the target freight product that is consistent with the current sales period of the target freight product (i.e., historical contemporaneous data) and the sales data of the target freight product that has been generated in the current sales period (i.e., current contemporaneous data) can be first collected, and then the historical contemporaneous data and the current contemporaneous data are processed by the prediction model iteratively obtained when predicting the previous period demand of the target freight product to obtain an initial prediction result, and the Kalman gain value can be calculated according to the previous period covariance matrix and the previous period model parameters used by the prediction model, so as to calculate the predicted demand of the target freight product in the target time period in the current sales period according to the Kalman gain value, the current contemporaneous data and the initial prediction result, and by introducing the Kalman filter to continuously iterate the model parameters and comprehensively considering the historical contemporaneous data and the current contemporaneous data to predict the demand of the freight product, the accuracy of predicting the demand of the freight product can be effectively improved, and the technical effect of efficiently and accurately predicting the demand of the freight product is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0027] Figure 1 is a flow chart of an optional method for determining demand for freight rate products according to an embodiment of the present invention;

[0028] Figure 2 is a schematic diagram of an optional freight product demand forecasting device based on a Kalman filter according to an embodiment of the present invention;

[0029] Figure 3 is a schematic diagram of an optional device for determining demand for freight rate products according to an embodiment of the present invention;

[0030] Figure 4 The present invention is a hardware structure block diagram of an electronic device (or mobile device) for determining a freight product demand quantity according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] To facilitate those skilled in the art to understand the present invention, some terms or nouns involved in the embodiments of the present invention are explained below:

[0034] Freight product: It is the concept of a whole product, that is, a freight product is a whole including a freight basis, a set of rules and a price.

[0035] Revenue management: refers to the application of different management methods to maximize the revenue of an airline's product portfolio, including differentiated price management, flight seat control, etc.

[0036] It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) collected and involved in the present invention are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data are in compliance with the relevant laws, regulations and standards of the relevant regions, necessary confidentiality measures are taken, and public order and good customs are not violated, and corresponding operation entrances are provided for users to choose to authorize or refuse. For example, an interface is set between the system and the relevant users or organizations. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain relevant information after receiving the consent information fed back by the aforementioned user or organization.

[0037] In the present invention, by modeling historical sales data, the demand for freight products, such as transportation demand during winter and summer vacations and long holidays, can be analyzed and predicted. However, since it is predicted based on historical spatiotemporal change trend data, there will be estimation errors. Therefore, the model prediction can be corrected in combination with the current data of the same period, which helps to improve the accuracy of the prediction. However, when introducing the current data of the same period, the measurement error caused by environmental noise must also be considered, and the error may vary greatly in different situations. For example, the current sales data during the long vacation is closer to the true value, but if an abnormal event (such as extreme weather) occurs, the confidence level of the next period prediction may be low. Therefore, the present invention constructs a prediction model that can fully consider the two data sources (historical data of the same period and current data of the same period) under different circumstances, and can analyze and predict the next period data with a higher confidence, shorten the prediction lag as much as possible, and provide more accurate demand forecast data, which is of great significance to the development of the civil aviation industry and the setting of airline products.

[0038] The present invention is described in detail below in conjunction with various embodiments.

[0039] Embodiment 1

[0040] According to an embodiment of the present invention, an embodiment of a method for determining the demand for freight rate products is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0041] Figure 1 is a flow chart of an optional method for determining the demand for freight rate products according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0042] Step S101, collecting historical data for the same period and current data for the same period, wherein the historical data for the same period refers to the historical sales data of the target freight rate product that is consistent with the current sales period of the target freight rate product, and the current data for the same period refers to the sales data of the target freight rate product that has been generated by the historical data for the same period in the current sales period.

[0043] In step S101, historical sales data of the target fare product that is consistent with the current sales period of the target fare product can be collected from the sales database (the database can store the sales data that have been generated for all fare products) as historical contemporaneous data (for example, the current sales period of the economy class of a certain flight is October, and the sales data of the economy class of the flight in October last year can be collected), and sales data of the target fare product that has been generated in the historical contemporaneous data during the current sales period can be collected from the sales database as current contemporaneous data (for example, the current sales period of the economy class of a certain flight is October, and the demand on October 10 needs to be predicted, then the sales data before October 10, such as October 9, can be collected).

[0044] In this embodiment, the current data for the same period is sales data including the current abnormal event. For example, if extreme weather occurs on October 9, the current data for the same period collected is sales data affected by the extreme weather.

[0045] In this embodiment, by collecting historical data and current data for the same period to predict the demand for subsequent target freight rate products, the impact of current abnormal events on sales data can be taken into account, thereby facilitating accurate prediction of the demand for target freight rate products in the future.

[0046] Step S102, based on historical data of the same period and current data of the same period, a preset prediction model is used to determine an initial prediction result, wherein the preset prediction model is an iterative prediction model when predicting the previous period demand of the target freight rate product, and the preset prediction model includes: previous period model parameters.

[0047] In step S102, the Kalman filter can be used to continuously iterate and update the prediction model to ensure that the model parameters used in the prediction model have a high degree of confidence. Each time the demand for the freight rate product is predicted, the prediction model iterated during the previous prediction can be used to process the historical data of the same period and the current data of the same period to obtain the initial prediction result.

[0048] In this embodiment, the model parameters of the prediction model include: a first coefficient matrix A, a second coefficient matrix B, a third coefficient matrix H, a systematic error ω k-1 , preset error v k etc., the preset error v k Including: Estimated error e est (i.e., the error in the forecast demand caused by historical data of the same period due to the situation that demand is not met or achieved in other ways, as well as the impact of time changes), measurement error e mea (i.e., errors caused by positive or negative impacts on transportation network demand caused by abnormal events).

[0049] In this embodiment, abnormal events refer to currently identifiable events that will have a significant impact on the demand of the transportation system, such as policy events, such as changes in a certain policy that will affect flight routes; economic events, such as increased tourism promotion in a certain place, concerts, and sports events; natural events, such as extreme weather, natural disasters, etc. These events or event sets will have a positive or negative impact on the current transportation network demand, thereby generating experimental errors in the forecast or market noise, which belongs to measurement errors.

[0050] In this embodiment, the calculated initial prediction result is a prediction result that takes into account the historical sales volume and the impact of the current abnormal event.

[0051] Step S103, determining the Kalman gain value based on the previous period covariance matrix and the previous period model parameters.

[0052] In step S103, the previous period covariance matrix P obtained when predicting the previous period demand of the target freight product can be used. k-1 And the previous model parameters updated when predicting the previous demand of the target freight rate product, calculate the Kalman gain value K k , so that the subsequent Kalman gain value can be used to determine whether the prediction of the target freight rate product is more biased towards the historical data of the same period or the current data of the same period.

[0053] Step S104, based on the Kalman gain value, the current contemporaneous data and the initial prediction result, determine the predicted demand for the target freight rate product in the target time period within the current sales period.

[0054] In step S104, the predicted demand for the target price product in the target time period within the current sales period can be calculated based on the Kalman gain value, the current data of the same period and the initial prediction result. In this embodiment, the target time period can be one day or several consecutive days.

[0055] For example, a demand forecast is conducted for a certain type of fare product on a certain route on December 31, 2023 (e.g., products that meet basic rules such as price range of 600-800, economy class, normal checked luggage of 20kg, sold 28-30 days in advance, and refund and change fees within 200). Based on the calculated Kalman gain value, initial forecast results, and current data collected for the same period (e.g., actual sales data on December 1, 2023), the demand for the fare product on December 31, 2023 can be predicted, so that the airline can control the sale of flight seats according to the predicted demand.

[0056] In summary, by collecting historical data for the same period and current data for the same period, based on the historical data for the same period and current data for the same period, a preset prediction model is used to determine the initial prediction result, based on the covariance matrix of the previous period and the model parameters of the previous period, the Kalman gain value is determined, based on the Kalman gain value, the current data for the same period and the initial prediction result, the predicted demand for the target freight product in the target time period within the current sales period is determined, thereby solving the technical problem that the demand for freight products cannot be accurately predicted in related technologies.

[0057] In order to build a high-precision prediction model, in the method for determining the demand for freight products provided in Example 1 of the present application, before collecting historical data for the same period and current data for the same period, it also includes: collecting historical data within a historical time period, and marking the historical data to obtain marking information, wherein the historical data includes: historical sales data of each freight product, and the marking information includes: actual sales values ​​in different historical sub-time periods, and the historical sub-time periods are sub-time periods obtained by dividing the historical time periods; using the historical data and the marking information to train an initial prediction model to obtain a preset prediction model, wherein the preset prediction model uses an initial covariance matrix and initial model parameters to predict the demand for freight products when it is first applied; the preset prediction model iteratively updates the initial covariance matrix and initial model parameters during each prediction process to obtain the covariance matrix and model parameters used for the next prediction; the initial covariance matrix is ​​a covariance matrix obtained based on historical data statistics, and the initial model parameters are model parameters obtained after the initial prediction model is trained.

[0058] In an embodiment of the present invention, historical data within a historical time period (such as the past three years) can be collected and annotated (for example, the historical time period is divided into multiple historical sub-time periods (such as one day as a historical sub-time period), and the actual sales value corresponding to each historical sub-time period is used as the annotation of the data in the historical sub-time period) to obtain annotation information, and then the historical data and the corresponding annotation information (that is, historical data annotated with actual sales values) can be used to train an initial prediction model until the model converges to obtain a preset prediction model.

[0059] In the embodiment of the present invention, the model parameters of the trained prediction model may be used as initial model parameters, and these model parameters may be continuously iterated and updated during each prediction process.

[0060] In an embodiment of the present invention, when the preset prediction model is applied for the first time, the initial covariance matrix (the initial covariance matrix is ​​a covariance matrix obtained based on historical data statistics, for example, the normal distribution variance of the statistical historical data is σ to obtain the initial covariance matrix) and initial model parameters can be used to predict the demand for freight products, and in each prediction process, the initial covariance matrix and the initial model parameters can be iteratively updated to obtain the covariance matrix and model parameters used for the next prediction. In this way, by continuously iterating the covariance matrix and model parameters, the high confidence and high precision of the model can be guaranteed, thereby ensuring the accuracy of the prediction.

[0061] In order to improve the accuracy of data collection, in the method for determining the demand for freight rate products provided in Example 1 of the present application, the current sales period and the target time period of the target freight rate product are determined, wherein the target time period is any time period within the current sales period that has not been reached; historical sales data of the target freight rate product is collected, and data within a period consistent with the current sales period is intercepted from the historical sales data to obtain historical data for the same period; when the target freight rate product is sold, data within a preset time period range that is before the target time period and consistent with the length of the target time period is collected to obtain current data for the same period, wherein the preset time period range is a time period range within the current sales period.

[0062] In the embodiment of the present invention, it is necessary to predict the demand for the freight rate product based on the historical data for the same period and the current data for the same period.

[0063] In this embodiment, the historical data for the same period includes the actual sales of a certain type of freight rate product for a certain route in a certain time period, which is used to reflect the historical demand for the same period. Therefore, the current sales period of the target freight rate product and the target time period (i.e., any time period that has not been reached in the current sales period) can be determined first, and then the historical sales data of the target freight rate product can be collected, and the data in the period consistent with the current sales period can be intercepted from the historical sales data to obtain the historical data for the same period (for example, last year's data sold in the same period (such as the long holiday period) as the target freight rate product).

[0064] In this embodiment, the current contemporaneous data refers to the current actual freight rate sales data. After the target freight rate product is sold, data within a preset time period range (i.e., a time period range within the current sales period) that is before the target time period and consistent with the length of the target time period can be collected to obtain the current contemporaneous data (i.e., the real sales data generated by a certain time period of the current sales period (which needs to be consistent with the length of the preset target time period)).

[0065] In this embodiment, the freight product may have an estimation error in the demand volume predicted using historical data of the same period due to the fact that the quantity provided during the historical sales did not meet the demand or the demand was met in other ways, as well as the change in time. In addition, the freight product may have a measurement error when using the current data of the same period to predict the demand volume due to various abnormal events during the current sales. Therefore, when using the prediction model to predict the demand volume, it is necessary to introduce estimation error and measurement error to improve the prediction accuracy.

[0066] In the embodiment of the present invention, the previous model parameters include: a first previous coefficient matrix and a second previous coefficient matrix. In order to improve the accuracy of determining the initial prediction result, in the method for determining the demand for freight rate products provided in the first embodiment of the present application, the historical average sales value is determined based on the historical data of the same period, and the impact value of the current abnormal event is determined based on the current data of the same period; the initial prediction result is determined based on the first previous coefficient matrix, the second previous coefficient matrix, the historical average sales value and the impact value, wherein the initial prediction result includes: the initial prediction value.

[0067] In the embodiment of the present invention, the previous model parameters include: a first previous coefficient matrix A, a second previous coefficient matrix B. The a priori estimation formula of the Kalman filter can be used to determine the initial prediction result of the target freight product, specifically: the historical average sales value can be determined based on the historical data of the same period And according to the current data of the same period, determine the impact value u of the current abnormal event k-1 Then, according to the first previous period coefficient matrix A, the second previous period coefficient matrix B, the historical average sales value And the impact value u k-1 , calculate the initial prediction value To get the initial prediction result. Among them, calculate the initial prediction value The prior estimation formula of is:

[0068] In an embodiment of the present invention, historical sales data can be collected and abnormal events that exist in the process of generating historical sales data can be counted to determine the impact value of the historical abnormal event based on the sales value with the abnormal event and the sales value without the abnormal event in the same period, thereby obtaining the impact value of each abnormal event, so as to more quickly determine the impact value of the abnormal event existing in the current data of the same period.

[0069] In an embodiment of the present invention, the previous model parameters also include: a third previous coefficient matrix, a previous systematic error, and a previous preset error, and the previous preset error includes at least: a previous estimation error and a previous measurement error. In order to improve the accuracy of determining the Kalman gain value, in the method for determining the demand for freight products provided in Example 1 of the present application, a previous covariance matrix is ​​obtained, wherein the previous covariance matrix is ​​a covariance matrix obtained when predicting the previous demand for the target freight product; a systematic error covariance matrix of the previous systematic error is constructed, and a preset error covariance matrix of the previous preset error is constructed; based on the previous covariance matrix, the first previous coefficient matrix, and the systematic error covariance matrix, an estimated covariance matrix is ​​constructed; based on the third previous coefficient matrix, the estimated covariance matrix, and the preset error covariance matrix, the Kalman gain value is determined.

[0070] In the embodiment of the present invention, the previous model parameters also include: the third previous coefficient matrix H, the previous systematic error ω k-1 、Previous period preset error v k The previous period preset error includes: the previous period estimated error e est , the measurement error of the previous period e mea Etc. The Kalman filter's prior estimated covariance formula can be used to construct the estimated covariance matrix Specifically: You can first obtain the previous period covariance matrix P k-1 (i.e., the covariance matrix obtained when predicting the previous period demand of the target freight rate product), and then construct the previous period systematic error ω k-1 The systematic error covariance matrix Q (i.e. ω k-1 The covariance matrix of the normal distribution obeyed by the previous period is P k-1 , the first previous period coefficient matrix A and the systematic error covariance matrix Q, construct the estimated covariance matrix Among them, the prior estimated covariance formula is:

[0071]

[0072] In the embodiment of the present invention, the previous period preset error υ can be constructed k The preset error covariance matrix R (i.e., υ k The Kalman gain calculation formula of the Kalman filter is then used to calculate the Kalman gain value, which is specifically: the covariance matrix can be estimated based on the third previous period coefficient matrix H and the And the preset error covariance matrix R, determine the Kalman gain value K k Among them, the Kalman gain calculation formula is:

[0073] In the embodiment of the present invention, the Kalman gain Kk The value range is [0,H -1 ]. From the Kalman gain calculation formula, it can be concluded that if the current measurement process is more accurate, that is, the normal distribution variance of the measurement error is small, the value of R will be small, then K k Tend to H -1 , so that the final prediction will be more inclined to the current real sales value generated by the current data of the same period; if the normal distribution variance of the measurement noise is large, the value of R will be large, then K k tends to 0, so the final forecast will be more inclined to the forecast value calculated based on the historical data of the same period.

[0074] In order to improve the accuracy of determining the predicted demand, in the method for determining the demand for freight products provided in Example 1 of the present application, the current average sales value is determined based on the current data for the same period; the predicted demand is determined based on the Kalman gain value, the current average sales value and the initial prediction value included in the initial prediction result.

[0075] In the embodiment of the present invention, the Kalman filter's posterior estimation can be used to determine the final predicted demand. Specifically, the current average sales value z can be calculated based on the current data of the same period. k (i.e. the current real sales value), and then according to the Kalman gain value K k , current average sales value z k And the initial prediction results contain the initial prediction values Calculate forecast demand Among them, the posterior estimation formula is:

[0076] In the embodiment of the present invention, when calculating the final predicted demand When the Kalman gain K k , to determine whether it is biased towards the initial forecast value calculated based on historical data of the same period or towards the current actual sales value. Therefore, the Kalman gain K k It is critical to improve forecast accuracy and needs to be recalculated every time a demand forecast is made.

[0077] In order to more accurately predict the next period's demand, in the method for determining the demand for freight products provided in Example 1 of the present application, after determining the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial prediction result, the previous period covariance matrix is ​​updated based on the estimated covariance matrix, the third previous period coefficient matrix used by the preset prediction model and the Kalman gain value to obtain the current period covariance matrix, wherein the current period covariance matrix is ​​used to predict the next period's demand for the target freight product in the next time period after the target time period within the current sales period.

[0078] In the embodiment of the present invention, in order to improve the confidence of the prediction, it is necessary to update the covariance matrix P used in the next prediction each time the prediction is made. k , specifically: According to the estimated covariance matrix The third previous period coefficient matrix H and Kalman gain value K used in the preset prediction model k , update the previous period covariance matrix P k-1 , to obtain the current covariance matrix P k (i.e., the covariance matrix used to predict the next period demand of the target freight product in the next time period after the target time period within the current sales period). k-1 The formula is:

[0079]

[0080] In order to avoid a decrease in prediction accuracy caused by abnormal events, in the method for determining the demand for freight products provided in Example 1 of the present application, after determining the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial prediction result, the current true value of the target freight product is obtained; when the absolute error value between the current true value and the predicted demand is greater than a preset error threshold, the Kalman gain value used when predicting the next period demand for the target freight product is corrected.

[0081] In the embodiment of the present invention, as time changes and various factors affect the change of factors, the error between the predicted value and the true value may increase. Therefore, it is necessary to obtain the current true value of the target freight rate product after each prediction. If the absolute error value between the current true value and the predicted demand is greater than a preset error threshold (which can be set according to actual conditions), it is necessary to correct the Kalman gain value used in predicting the next demand for the target freight rate product.

[0082] In this embodiment, when the absolute error value between the current actual value and the predicted demand is greater than the preset error threshold, all sales log data in the sales process are fed back to the correction module, so that the correction personnel analyze the sales log data through the correction module to determine the corrected Kalman gain value.

[0083] In this embodiment, the causes and treatment processes of large errors include: (1) an unidentified important abnormal event occurs, which needs to be included in the event set and learned; (2) as time changes, the influence of the original abnormal event changes, resulting in a change in its dominance in the prediction process, which in turn affects the assessment of the confidence level of historical data and current data, and requires re-analysis and assessment of its error range and influencing factors; (3) the prediction anomaly caused by other sudden events is unsustainable, and this anomaly can be recorded and regarded as noise.

[0084] Figure 2 is a schematic diagram of an optional freight product demand forecasting device based on a Kalman filter according to an embodiment of the present invention, such as Figure 2 As shown, the prediction device includes: a confidence analysis module, a prediction module and a correction module.

[0085] exist Figure 2 In the confidence analysis module, the confidence analysis module can perform confidence analysis based on historical data, current data, and abnormal events. For example, based on machine learning, it can analyze the estimated error e in the current period. est and the measurement error of the current related abnormal events e mea The estimation error and measurement error conform to the normal distribution. Since the error distribution of the current data affected by the abnormal event may involve multiple events, the covariance matrix can be used for analysis.

[0086] exist Figure 2 In the example, the prediction module can perform Kalman filter analysis based on the abnormal events and the analysis results of the confidence analysis module to obtain the predicted value.

[0087] exist Figure 2 In the process, the correction module (including temporary confidence and long-term confidence) can compare the predicted value and the current data. If the difference between the two is large, an early warning feedback is given, and the confidence analysis data used by the confidence analysis module is updated according to the abnormal event.

[0088] In this embodiment, the long-term confidence is the confidence used normally during prediction. The temporary confidence is the confidence used temporarily when a large error occurs in the prediction. The long-term confidence is used again after the factor causing the large error disappears, or the factor is included in the prediction and the long-term confidence is updated again.

[0089] In this embodiment, under the basic theory of Kalman filter, the demand forecast of freight rate products can be converted into: according to the data in two estimation sources / measurement sources (historical data and current data), the historical data estimation results (z1, σ1) and the current actual data (z2, σ2) affected by a series of external factors (i.e., abnormal events) are statistically obtained, wherein z1 represents the estimated value, z2 represents the actual value, σ1 and σ2 represent the variance of the normal distribution, so as to generate a prediction value with relatively small error based on the historical data estimation results (z1, σ1) and the current actual data (z2, σ2).

[0090] According to the historical data of the same period and the error distribution of this period, the state equation for the demand of this period's freight rate products is as follows:

[0091]

[0092] in, represents the result of the calculation using the prediction model (i.e. the initial prediction result), It represents the historical average sales value calculated by historical data of the same period, u k-1 is the impact value of the currently known abnormal event, ω k-1 is the process noise (i.e., systematic error) of the model calculation, υ k represents the measurement noise (i.e., preset error, including estimation error and measurement error) that may be caused by the current measurement result, z k Indicates the current real sales value, It represents the predicted value generated based on the current data, and A, B, and H are the coefficient matrices calculated after linearizing the time sampling in a discrete state.

[0093] The following is a detailed explanation of the demand forecasting process using a specific freight rate product as an example.

[0094] For example, we can make a demand forecast for a certain type of fare product on a certain route on December 31, 2023 (such as fare products that meet the basic rules such as price range of 600-800, economy class, normal checked luggage of 20kg, sold 28-30 days in advance, and refund and change fee within 200). We can first use the confidence analysis module to predict historical data. It should be noted that the confidence of historical data is obtained by previous iterations, that is, using last year's sales data. Preliminary forecast results for this year At the same time, considering the impact of the current abnormal events and their influencing factors on the current demand and their correlation coefficients, Bu k-1 At this time, the confidence analysis module has generated an error ω for the prediction result based on the previous data and the actual situation at that time. k-1 , and generate forecast data for this period based on various coefficients Actual sales data from December 1 to 3, 2023 k After the data is obtained, the confidence level of the data is evaluated under the influence of the current environmental factors, and the reference range of the noise error for the next period prediction is generated. k , and generate the forecast value of the next period (the demand of the same freight rate product 25-27 days in advance) based on the current sales data

[0095] In this embodiment, the prediction module can generate a priori estimation results based on historical data and current data, and generate a covariance matrix based on the confidence level learning results of each abnormal event (i.e., the learning results of the current process noise / measurement noise) of the confidence analysis module to calculate the error covariance of the current prediction result, and further calculate the Kalman gain, so as to generate a more accurate prediction value for the next period of data based on historical data and current data. After this, the prediction module can further update the covariance matrix P of the current error k .

[0096] However, although the prediction module can predict the demand for freight products more accurately, as time changes and various factors affect the factors, a correction module is also needed to track the error between the predicted value and the actual value in real time. When the error is large, timely feedback is provided so that relevant personnel can analyze the cause of the error and update the model parameters or learning results of the confidence analysis module.

[0097] In the embodiment of the present invention, a prediction method based on data fusion of a Kalman filter is proposed, which can comprehensively consider the relative accuracy of historical data and current data under different circumstances, and comprehensively refer to different prediction values ​​and their confidence, and finally generate a more accurate freight product demand prediction result. In addition, by separating the learning process and the prediction process, the data extraction of the prediction reference can be faster, the calculation amount of each prediction is relatively small, and real-time prediction can be achieved.

[0098] In the embodiment of the present invention, multiple influences can be effectively referenced to control the system deviation caused by incomplete data, taking into account both the point-to-point flight structure and the hub-and-spoke network structure. In addition, the correction module is introduced to timely iterate and update the knowledge base of the prediction device, taking into full consideration the fact that there are many factors affecting flights and the influence of factors may change.

[0099] The following is a detailed description in conjunction with another embodiment.

[0100] Embodiment 2

[0101] The device for determining the demand quantity of a freight rate product provided in this embodiment includes a plurality of implementation units, each of which corresponds to each implementation step in the above-mentioned embodiment 1.

[0102] Figure 3 is a schematic diagram of an optional device for determining the demand for freight rate products according to an embodiment of the present invention, such as Figure 3 As shown, the determination device may include: a collection unit 30 , a first determination unit 31 , a second determination unit 32 , and a third determination unit 33 .

[0103] The collection unit 30 is used to collect historical data for the same period and current data for the same period, wherein the historical data for the same period refers to the historical sales data of the target freight rate product that is consistent with the current sales period of the target freight rate product, and the current data for the same period refers to the sales data generated by the target freight rate product in the current sales period;

[0104] The first determination unit 31 is used to determine the initial prediction result based on the historical data of the same period and the current data of the same period by using a preset prediction model, wherein the preset prediction model is a prediction model iterated when predicting the previous period demand of the target freight rate product, and the preset prediction model includes: previous period model parameters;

[0105] A second determination unit 32, configured to determine a Kalman gain value based on a previous period covariance matrix and a previous period model parameter;

[0106] The third determining unit 33 is used to determine the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial prediction result.

[0107] The above-mentioned determination device can collect historical data of the same period and current data of the same period through the collection unit 30, determine the initial prediction result through the first determination unit 31 based on the historical data of the same period and the current data of the same period, adopt a preset prediction model, determine the Kalman gain value based on the previous period covariance matrix and the previous period model parameters through the second determination unit 32, and determine the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current data of the same period and the initial prediction result through the third determination unit 33, thereby solving the technical problem that the demand for freight products cannot be accurately predicted in the related technology.

[0108] Optionally, the determination device also includes: a first acquisition module, which is used to collect historical data within a historical time period before collecting historical contemporaneous data and current contemporaneous data, and mark the historical data to obtain marking information, wherein the historical data includes: historical sales data of each freight product, and the marking information includes: actual sales values ​​in different historical sub-time periods, and the historical sub-time period is a sub-time period obtained by dividing the historical time period; a first training module, which is used to train an initial prediction model using historical data and marking information to obtain a preset prediction model, wherein the preset prediction model uses an initial covariance matrix and initial model parameters to predict the demand for freight products when it is first applied; the preset prediction model iteratively updates the initial covariance matrix and initial model parameters in each prediction process to obtain the covariance matrix and model parameters used for the next prediction; the initial covariance matrix is ​​a covariance matrix obtained based on historical data statistics, and the initial model parameters are model parameters obtained after the initial prediction model training is completed.

[0109] Optionally, the collection unit includes: a first determination module, used to determine the current sales period and target time period of the target freight rate product, wherein the target time period is any time period within the current sales period that has not been reached; a second collection module, used to collect historical sales data of the target freight rate product, and intercept data within a period consistent with the current sales period from the historical sales data to obtain historical contemporaneous data; a third collection module, used to collect data within a preset time period range that is before the target time period and consistent with the length of the target time period when the target freight rate product is sold, to obtain current contemporaneous data, wherein the preset time period range is a time period range within the current sales period.

[0110] Optionally, the previous period model parameters include: a first previous period coefficient matrix, a second previous period coefficient matrix, and the first determination unit includes: a second determination module, used to determine the historical average sales value based on the historical data of the same period, and to determine the impact value of the current abnormal event based on the current data of the same period; a third determination module, used to determine the initial prediction result based on the first previous period coefficient matrix, the second previous period coefficient matrix, the historical average sales value and the impact value, wherein the initial prediction result includes: an initial prediction value.

[0111] Optionally, the previous model parameters also include: a third previous coefficient matrix, a previous systematic error, and a previous preset error, wherein the previous preset error includes at least: a previous estimation error and a previous measurement error, and the second determination unit includes: a first acquisition module for acquiring a previous covariance matrix, wherein the previous covariance matrix is ​​a covariance matrix obtained when predicting the previous demand for the target freight rate product; a first construction module for constructing a systematic error covariance matrix of the previous systematic error, and a preset error covariance matrix of the previous preset error; a second construction module for constructing an estimated covariance matrix based on the previous covariance matrix, the first previous coefficient matrix, and the systematic error covariance matrix; and a fourth determination module for determining the Kalman gain value based on the third previous coefficient matrix, the estimated covariance matrix, and the preset error covariance matrix.

[0112] Optionally, the third determination unit includes: a fifth determination module, used to determine the current average sales value based on current contemporaneous data; a sixth determination module, used to determine the predicted demand based on the Kalman gain value, the current average sales value and the initial prediction value included in the initial prediction result.

[0113] Optionally, the determination device also includes: a first updating module, which is used to determine the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, current contemporaneous data and initial prediction results, and then update the previous period covariance matrix based on the estimated covariance matrix, the third previous period coefficient matrix used by the preset prediction model and the Kalman gain value to obtain the current period covariance matrix, wherein the current period covariance matrix is ​​used to predict the next period demand for the target freight product in the next time period after the target time period within the current sales period.

[0114] Optionally, the determination device also includes: a second acquisition module, which is used to obtain the current true value of the target freight product after determining the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, current contemporaneous data and initial prediction results; a first correction module, which is used to correct the Kalman gain value used in predicting the next period demand for the target freight product when the absolute error value between the current true value and the predicted demand is greater than a preset error threshold.

[0115] The above-mentioned determination device may also include a processor and a memory. The above-mentioned acquisition unit 30, the first determination unit 31, the second determination unit 32, the third determination unit 33, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0116] The processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set to adjust kernel parameters to determine the forecast demand for the target freight product in the target time period during the current sales period based on the Kalman gain value, the current data of the same period and the initial forecast result.

[0117] The above-mentioned memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (fl ash RAM), and the memory includes at least one storage chip.

[0118] The present invention also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that is initialized with the following method steps: collecting historical data for the same period and current data for the same period, based on the historical data for the same period and current data for the same period, using a preset prediction model to determine an initial prediction result, based on the previous period covariance matrix and the previous period model parameters, determining a Kalman gain value, and based on the Kalman gain value, current data for the same period and the initial prediction result, determining the predicted demand for the target freight rate product in the target time period within the current sales period.

[0119] According to another aspect of an embodiment of the present invention, there is also provided an electronic device, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors implement the above-mentioned method for determining the demand for freight rate products.

[0120] Figure 4 1 is a hardware structure block diagram of an electronic device (or mobile device) for determining the demand for freight rate products according to an embodiment of the present invention. Figure 4 As shown, the electronic device may include one or more processors (e.g., Figure 4 The processor 402a, processor 402b, ..., processor 402n, etc., which may include but are not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 404 for storing data. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 4 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 4 More or fewer components as shown, or with Figure 4 Different configurations are shown.

[0121] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0122] The embodiments or examples of the present disclosure are not exhaustive, but are only illustrative of some embodiments or examples, and are not intended to be specific limitations on the scope of protection of the present disclosure. In the absence of contradiction, each step in a certain embodiment or example can be implemented as an independent example, and the steps can be combined arbitrarily. For example, the scheme after removing some steps in a certain embodiment or example can also be implemented as an independent example, and the order of the steps in a certain embodiment or example can be arbitrarily exchanged. In addition, the optional methods or optional examples in a certain embodiment or example can be combined arbitrarily; in addition, the various embodiments or examples can be combined arbitrarily, for example, some or all steps of different embodiments or examples can be combined arbitrarily, and a certain embodiment or example can be combined arbitrarily with the optional methods or optional examples of other embodiments or examples.

[0123] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0124] In the several embodiments provided by the present invention, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0125] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0126] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0127] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0128] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for determining the demand for freight rate products, characterized in that: include: Collect historical data for the same period and current data for the same period, wherein the historical data for the same period refers to the historical sales data of the target freight rate product that is consistent with the current sales period of the target freight rate product, and the current data for the same period refers to the sales data generated by the target freight rate product during the current sales period; Based on the historical data of the same period and the current data of the same period, a preset prediction model is used to determine an initial prediction result, wherein the preset prediction model is an iterative prediction model when predicting the previous period demand of the target freight rate product, and the preset prediction model includes: previous period model parameters; Determining a Kalman gain value based on a previous period covariance matrix and the previous period model parameters; Based on the Kalman gain value, the current contemporaneous data and the initial prediction result, the predicted demand for the target freight rate product in the target time period within the current sales period is determined.

2. The determination method according to claim 1, characterized in that: Before collecting historical data for the same period and current data for the same period, it also includes: Collecting historical data within a historical time period, and marking the historical data to obtain marking information, wherein the historical data includes: historical sales data of each freight rate product, and the marking information includes: real sales values ​​within different historical sub-time periods, and the historical sub-time period is a sub-time period obtained by dividing the historical time period; The historical data and the annotation information are used to train an initial prediction model to obtain the preset prediction model, wherein, when the preset prediction model is first applied, an initial covariance matrix and initial model parameters are used to predict the demand for the freight product; during each prediction process, the preset prediction model iteratively updates the initial covariance matrix and the initial model parameters to obtain the covariance matrix and model parameters used for the next prediction; the initial covariance matrix is ​​a covariance matrix obtained based on the statistics of the historical data, and the initial model parameters are model parameters obtained after the initial prediction model is trained.

3. The determination method according to claim 1, characterized in that: The steps to collect historical and current data for the same period include: Determine the current sales period and the target time period of the target freight rate product, wherein the target time period is any unreached time period within the current sales period; Collecting the historical sales data of the target freight rate product, and extracting data in a period consistent with the current sales period from the historical sales data to obtain the historical data for the same period; When the target freight rate product is sold, data within a preset time period range that is before the target time period and consistent with the length of the target time period is collected to obtain the current data for the same period, wherein the preset time period range is the time period range within the current sales period.

4. The determination method according to claim 1, characterized in that: The previous period model parameters include: a first previous period coefficient matrix and a second previous period coefficient matrix. Based on the historical data of the same period and the current data of the same period, a preset prediction model is used to determine the initial prediction result, including: Based on the historical data for the same period, determine the historical average sales value, and based on the current data for the same period, determine the impact value of the current abnormal event; Based on the first previous period coefficient matrix, the second previous period coefficient matrix, the historical average sales value and the impact value, the initial prediction result is determined, wherein the initial prediction result includes: an initial prediction value.

5. The determination method according to claim 1, characterized in that: The previous model parameters also include: a third previous coefficient matrix, a previous systematic error, and a previous preset error. The previous preset error at least includes: a previous estimation error and a previous measurement error. Based on the previous covariance matrix and the previous model parameters, the step of determining the Kalman gain value includes: Acquire the previous period covariance matrix, wherein the previous period covariance matrix is ​​a covariance matrix obtained when predicting the previous period demand of the target freight rate product; Constructing a systematic error covariance matrix of the previous period systematic error, and constructing a preset error covariance matrix of the previous period preset error; Constructing an estimated covariance matrix based on the previous period covariance matrix, the first previous period coefficient matrix and the systematic error covariance matrix; The Kalman gain value is determined based on the third previous period coefficient matrix, the estimated covariance matrix and the preset error covariance matrix.

6. The determination method according to claim 1, characterized in that: The step of determining the predicted demand for the target freight rate product in a target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial prediction result comprises: Based on the current data for the same period, determining the current average sales value; The predicted demand is determined based on the Kalman gain value, the current average sales value, and the initial prediction value included in the initial prediction result.

7. The determination method according to claim 1, characterized in that: After determining the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial prediction result, the method further includes: Based on the estimated covariance matrix, the third previous period coefficient matrix used by the preset prediction model and the Kalman gain value, the previous period covariance matrix is ​​updated to obtain the current period covariance matrix, wherein the current period covariance matrix is ​​used to predict the next period demand of the target freight product in the next time period after the target time period within the current sales period.

8. The determination method according to claim 1, characterized in that: After determining the predicted demand for the target freight product in the target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial prediction result, the method further includes: Obtaining the current real value of the target freight rate product; When the absolute error between the current actual value and the predicted demand is greater than a preset error threshold, the Kalman gain value used in predicting the demand for the target freight product in the next period is corrected.

9. A device for determining demand for freight rate products, characterized in that: include: A collection unit, used to collect historical data of the same period and current data of the same period, wherein the historical data of the same period refers to the historical sales data of the target freight rate product that is consistent with the current sales period of the target freight rate product, and the current data of the same period refers to the sales data generated by the target freight rate product in the current sales period; A first determining unit is used to determine an initial prediction result based on the historical data of the same period and the current data of the same period by using a preset prediction model, wherein the preset prediction model is a prediction model iterated when predicting the previous period demand of the target freight rate product, and the preset prediction model includes: previous period model parameters; A second determination unit, configured to determine a Kalman gain value based on a previous period covariance matrix and the previous period model parameters; The third determining unit is used to determine the predicted demand for the target freight product in a target time period within the current sales period based on the Kalman gain value, the current contemporaneous data and the initial prediction result.

10. An electronic device, characterized in that: The method comprises one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method for determining the demand for freight rate products as described in any one of claims 1 to 8.

Images