CN115660728B

CN115660728B - Air ticket sales order prediction method and device, electronic equipment and storage medium

Info

Publication number: CN115660728B
Application number: CN202211241700.0A
Authority: CN
Inventors: 石路路; 赵英廷; 孟平; 史超
Original assignee: Nanjing Yibo Software Technology Co ltd
Current assignee: Nanjing Yibo Software Technology Co ltd
Priority date: 2022-10-11
Filing date: 2022-10-11
Publication date: 2023-10-20
Anticipated expiration: 2042-10-11
Also published as: CN115660728A

Abstract

The application relates to an air ticket sales order prediction method, an air ticket sales order prediction device, electronic equipment and a storage medium. Wherein the method comprises the following steps: acquiring air ticket order history data; obtaining target prediction data according to the historical target data and at least one characteristic; the historical target data comprise first order quantity information of the air ticket sales order in a first time period in the air ticket order historical data; at least one feature is constructed according to the travel information of the ticket sales order in the ticket order history data; the target forecast data includes second order quantity information of the air ticket sales order within a specified time period in at least one airline share return agreement; determining whether to recommend a protocol flight to the user according to the target prediction data; wherein the agreement flights include at least one flight corresponding to an airline return agreement. The method and the device can improve the accuracy of the prediction result, are beneficial to reasonably completing the airline-department commission returning protocol, improve the completion rate of the airline-department commission returning protocol and improve the income brought by the airline-department commission returning protocol.

Description

Air ticket sales order prediction method and device, electronic equipment and storage medium

Technical Field

The present application relates to the field of computer technologies, and in particular, to a method and apparatus for predicting an air ticket sales order, an electronic device, and a storage medium.

Background

The business travel management company (Travel Management Companies, TMC) can help the enterprise to comprehensively execute monitoring on the whole planning of business travel activities under the assistance of the professional business travel management service team, optimize business travel management flow and policy and integrally purchase resources, so that business travel cost is reduced and employee travel efficiency is improved on the premise of not affecting business development and travel experience. The TMC provides a travel reservation platform for enterprise staff to reserve traffic and accommodation in travel. The traffic comprises air tickets, train tickets and vehicles, wherein the air ticket resources mainly originate from airlines/suppliers, and TMC companies sign a series of contracts such as flight refreshing, agreement price, commission and the like with the airlines/suppliers according to the air ticket travel amount of service enterprises in the past year.

The department returns the agreement to agree that: under the conditions of a specified time range, a specified round trip city, a specified flight and the like, when the total fare or the air range quantity meets a specified quantity, a commission of the specified condition is given; conversely, if the specified amount is not satisfied, the return is 0. Thus, how to ensure that more return revenue is obtained from the airline operators is an urgent business problem to be solved.

Disclosure of Invention

In view of the above, an air ticket sales order prediction method, an air ticket sales order prediction device, an electronic device and a storage medium are provided.

In a first aspect, an embodiment of the present application provides a method for predicting an air ticket sales order, the method comprising: acquiring air ticket order history data; obtaining target prediction data according to the historical target data and at least one characteristic; wherein the historical target data comprises first order quantity information of the air ticket sales order in a first time period in the air ticket order historical data; the at least one feature is constructed according to the travel information of the ticket sales order in the ticket order history data; the target forecast data comprises second order quantity information of the air ticket sales order in a specified time period in at least one air department return agreement; determining whether to recommend a protocol flight to a user according to the target prediction data; wherein the agreement flights include flights corresponding to the at least one airline employment agreement.

Based on the technical scheme, the target prediction data of the air ticket sales order is obtained according to the historical target data of the air ticket sales order and at least one feature, and the at least one feature is constructed according to the travel information of the air ticket sales order in the air ticket order historical data, so that the accuracy of the prediction result is improved; and determining whether to recommend flights specified in the airline share return agreement to the user according to the target prediction data of the air ticket sales order, so that the airline share return agreement can be reasonably completed, the completion rate of the airline share return agreement can be improved, and the income brought by the airline share return agreement can be improved.

In a first possible implementation manner of the first aspect according to the first aspect, the method further includes: under the condition that the recommendation of the protocol flight to the user is determined, obtaining a recommendation index of at least one protocol flight according to the user portrait and the protocol flight portrait; wherein, the user portrait is constructed according to the historical order information of the user; the protocol flight portrait is constructed according to the information of the protocol flight; and recommending the agreement flight to the user according to the recommendation index.

Based on the technical scheme, under the condition that the protocol flight is determined to be recommended to the user, the recommendation index of the protocol flight is calculated according to the user portrait and the protocol flight portrait, and the protocol flight is recommended to the user according to the recommendation index, so that the order taking rate of the user for the protocol flight ticket is improved, the impulse can be carried out on the air department return commission protocol which does not finish the target, the completion rate of the air department return commission protocol is improved, and the income brought by the air department return commission protocol is improved.

In a second possible implementation manner of the first aspect according to the first aspect or the first possible implementation manner of the first aspect, the obtaining the target prediction data according to the historical target data and the at least one feature includes: obtaining a historical time sequence according to the historical target data; inputting the historical time sequence and the at least one feature into a preset model to obtain the target prediction data; the preset model is obtained by training based on an STL time sequence decomposition algorithm.

Based on the technical scheme, the historical time sequence and the characteristics are input into a preset model obtained by training based on the STL algorithm together to obtain target prediction data, and the extraction method of the periodic term sequence and the trend term sequence in the STL algorithm can be optimized, so that the traditional STL algorithm is improved, the problem of dependence on local data is solved, a more accurate time sequence decomposition result can be obtained, the accuracy of the prediction result can be improved, and more accurate target prediction data can be obtained.

In a third possible implementation manner of the first aspect according to the second possible implementation manner of the first aspect, the preset model includes a first sub-model, a second sub-model, and a third sub-model; the first sub-model, the second sub-model and the third sub-model are obtained by training based on a regression algorithm; the step of inputting the historical time sequence and the at least one feature into a preset model to obtain the target prediction data comprises the following steps: obtaining an input sequence according to the historical time sequence and the at least one feature; obtaining a trend item sequence in the k-1 iteration, and obtaining a first intermediate sequence according to the input sequence and the trend item sequence in the k-1 iteration; wherein the initial value of k is 1, and the initial value of the trend item sequence is 0; inputting the first intermediate sequence into the first sub-model to obtain a second intermediate sequence; obtaining a third intermediate sequence according to the second intermediate sequence; inputting the third intermediate sequence into the second sub-model to obtain a fourth intermediate sequence; according to the input sequence, the second intermediate sequence and the fourth intermediate sequence, a periodic item sequence and a fifth intermediate sequence in the kth iteration are obtained; inputting the fifth intermediate sequence into the third sub-model to obtain a trend item sequence in the kth iteration; judging whether the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, stopping iteration under the condition that the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, determining the periodic item sequence in the last iteration as the periodic item sequence corresponding to the input sequence, and determining the trend item sequence in the last iteration as the trend item sequence corresponding to the input sequence; otherwise, adding 1 to the value of k, and returning to execute the step of acquiring the trend item sequence in the k-1 iteration; and obtaining the target prediction data according to the periodic item sequence corresponding to the input sequence and the trend item sequence corresponding to the input sequence.

Based on the technical scheme, the historical time sequence and the characteristics are input into a preset model obtained by training based on the STL algorithm together to obtain target prediction data, and the extraction method of the periodic term sequence and the trend term sequence is optimized, so that the traditional STL algorithm is improved, the problem of dependence on local data is solved, the first sub-model, the second sub-model and the third sub-model obtained by calculation based on the regression algorithm are utilized to accurately decompose and divide and treat the sequence information, more accurate time sequence decomposition is realized, and therefore, the accuracy of a prediction result can be improved, and more accurate target prediction data is obtained.

In a fourth possible implementation manner of the first aspect according to the third possible implementation manner of the first aspect, the obtaining, according to the input sequence, the second intermediate sequence, and the fourth intermediate sequence, a periodic term sequence and a fifth intermediate sequence in a kth iteration includes: subtracting the fourth intermediate sequence from the second intermediate sequence to obtain a periodic item sequence in the kth iteration; subtracting the periodic term sequence in the kth iteration from the input sequence to obtain the fifth intermediate sequence.

Based on the technical scheme, the historical time sequence and the characteristics are combined to form the input of the trend item sequence and the periodic item sequence in the time sequence, so that the extraction method of the periodic item sequence and the trend item sequence is optimized, the more accurate periodic item sequence and trend item sequence can be obtained, the accuracy of a prediction result can be improved, and more accurate target prediction data can be obtained.

In a fifth possible implementation form of the first aspect as such or any of the various possible implementation forms thereof, the target forecast data further comprises third order quantity information of the ticket sales order over a second time period corresponding to the at least one airline employment back agreement; and the third order quantity information of the air ticket sales order in the second time period corresponding to the at least one air department commission agreement is used for evaluating the air department commission agreement established in the second time period.

Based on the technical scheme, the order quantity information of the voyage returning protocol for a period of time in the future is predicted, so that references can be provided for preparing the voyage returning protocol of the next stage, and more reasonable voyage returning protocol can be prepared, thereby improving the completion rate of the voyage returning protocol and improving the income brought by the voyage returning protocol.

In a sixth possible implementation manner of the first aspect according to the first aspect or the various possible implementation manners of the first aspect, the first order quantity information and the second order quantity information include one or more of a total fare, a leg quantity, and a commission amount.

Based on the technical scheme, the total fare, the leg quantity and the return amount of the air ticket sales order specified by the air ticket return agreement are met in the air ticket order history data, the total fare, the leg quantity and the return amount of the air ticket return agreement in a period of time in the future are predicted, and the predicted target completion rate of the air ticket return agreement can be calculated, so that whether the air ticket return agreement of an incomplete target needs to be imputed is determined, reasonable completion of the air ticket return agreement is facilitated, the completion rate of the air ticket return agreement is improved, and the income brought by the air ticket return agreement is improved.

In a seventh possible implementation manner of the first aspect according to the first aspect or the various possible implementation manners of the first aspect, the trip information includes one or more of date information, holiday information, city information, and business trip application information.

Based on the technical scheme, holiday characteristics can be constructed according to the date information and holiday information of the ticket sales order in the ticket order history data, business trip application number characteristics are constructed according to city information and business trip application information, and the holiday characteristics and business trip application number characteristics are used for predicting target prediction data, so that accuracy of a prediction result is improved.

In a second aspect, an embodiment of the present application provides an air ticket sales order forecasting apparatus, the apparatus comprising: the acquisition module is used for acquiring the history data of the air ticket order; the prediction module is used for obtaining target prediction data according to the historical target data and at least one characteristic; wherein the historical target data comprises first order quantity information of the air ticket sales order in a first time period in the air ticket order historical data; the at least one feature is constructed according to the travel information of the ticket sales order in the ticket order history data; the target forecast data comprises second order quantity information of the air ticket sales order in a specified time period in at least one air department return agreement; the recommending module is used for determining whether to recommend the protocol flight to the user according to the target prediction data; wherein the agreement flights include flights corresponding to the at least one airline employment agreement.

In a first possible implementation manner of the second aspect according to the second aspect, the recommendation module is further configured to: under the condition that the recommendation of the protocol flight to the user is determined, obtaining a recommendation index of at least one protocol flight according to the user portrait and the protocol flight portrait; wherein, the user portrait is constructed according to the historical order information of the user; the protocol flight portrait is constructed according to the information of the protocol flight; and recommending the agreement flight to the user according to the recommendation index.

In a second possible implementation manner of the second aspect or the first possible implementation manner of the second aspect, the prediction module is further configured to: obtaining a historical time sequence according to the historical target data; inputting the historical time sequence and the at least one feature into a preset model to obtain the target prediction data; the preset model is obtained by training based on an STL time sequence decomposition algorithm.

In a third possible implementation manner of the second aspect according to the second possible implementation manner of the second aspect, the preset model includes a first sub-model, a second sub-model, and a third sub-model; the first sub-model, the second sub-model and the third sub-model are obtained by training based on a regression algorithm; the prediction module is further configured to: obtaining an input sequence according to the historical time sequence and the at least one feature; obtaining a trend item sequence in the k-1 iteration, and obtaining a first intermediate sequence according to the input sequence and the trend item sequence in the k-1 iteration; wherein the initial value of k is 1, and the initial value of the trend item sequence is 0; inputting the first intermediate sequence into the first sub-model to obtain a second intermediate sequence; obtaining a third intermediate sequence according to the second intermediate sequence; inputting the third intermediate sequence into the second sub-model to obtain a fourth intermediate sequence; according to the input sequence, the second intermediate sequence and the fourth intermediate sequence, a periodic item sequence and a fifth intermediate sequence in the kth iteration are obtained; inputting the fifth intermediate sequence into the third sub-model to obtain a trend item sequence in the kth iteration; judging whether the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, stopping iteration under the condition that the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, determining the periodic item sequence in the last iteration as the periodic item sequence corresponding to the input sequence, and determining the trend item sequence in the last iteration as the trend item sequence corresponding to the input sequence; otherwise, adding 1 to the value of k, and returning to execute the step of acquiring the trend item sequence in the k-1 iteration; and obtaining the target prediction data according to the periodic item sequence corresponding to the input sequence and the trend item sequence corresponding to the input sequence.

In a fourth possible implementation manner of the second aspect according to the third possible implementation manner of the second aspect, the prediction module is further configured to: subtracting the fourth intermediate sequence from the second intermediate sequence to obtain a periodic item sequence in the kth iteration; subtracting the periodic term sequence in the kth iteration from the input sequence to obtain the fifth intermediate sequence.

In a fifth possible implementation manner of the second aspect according to the second aspect or the various possible implementation manners of the second aspect, the target prediction data further includes third order quantity information of the air ticket sales order within a second time period corresponding to the at least one air ticket return commission protocol; and the third order quantity information of the air ticket sales order in the second time period corresponding to the at least one air department commission agreement is used for evaluating the air department commission agreement established in the second time period.

In a sixth possible implementation manner of the second aspect according to the second aspect or the various possible implementation manners of the second aspect, the first order quantity information and the second order quantity information include one or more of a total fare, a leg quantity, and a return commission amount.

In a seventh possible implementation manner of the second aspect according to the second aspect or the various possible implementation manners of the second aspect, the trip information includes one or more of date information, holiday information, city information, and business trip application information.

In a third aspect, embodiments of the present application provide an electronic device, where the terminal device may perform the above method for predicting an air ticket sales order according to the first aspect or one or more of the implementations of the first aspect.

In a fourth aspect, embodiments of the present application provide a computer program product comprising computer readable code, or a non-transitory computer readable storage medium carrying computer readable code, which when run in an electronic device, a processor in the electronic device performs the ticket sales order forecasting method of the first aspect or one or more of the implementations of the first aspect.

In a fifth aspect, embodiments of the present application provide a computer readable storage medium having stored thereon computer program instructions which when executed by a processor implement the ticket sales order forecasting method of the first aspect or one or more of the various implementations of the first aspect.

Other features and aspects of the present application will become apparent from the following detailed description of exemplary embodiments, which proceeds with reference to the accompanying drawings.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate exemplary embodiments, features and aspects of the application and together with the description, serve to explain the principles of the application.

Fig. 1 shows a flow chart of an STL algorithm according to an embodiment of the present application.

FIG. 2 illustrates a flow chart of a method of air ticket sales order forecasting according to an embodiment of the present application.

FIG. 3 illustrates a schematic diagram of a user representation in accordance with an implementation of the application.

FIG. 4 illustrates a schematic diagram of a user profile and a protocol flight profile in accordance with an implementation of the application.

FIG. 5 illustrates a flow chart of a method of air ticket sales order forecasting according to an embodiment of the present application.

Fig. 6 shows a flowchart of constructing a preset model according to an embodiment of the present application.

FIG. 7 shows a schematic diagram of an airline ticket sales order prediction system according to an embodiment of the present application.

Fig. 8 (a) -8 (c) show schematic diagrams of an air ticket sales order forecasting system according to an embodiment of the present application.

Fig. 9 shows a schematic diagram of a prediction curve of the air ticket sales order prediction method and the total fare of the air ticket return agreement according to the STL algorithm according to an embodiment of the present application.

FIG. 10 illustrates a target lift rate schematic of an airline ticket sales order prediction system after impulsion of three airline share return agreements in accordance with an embodiment of the present application.

Fig. 11 shows a block diagram of an air ticket sales order forecasting apparatus according to an embodiment of the present application.

Fig. 12 is a schematic structural view of an electronic device according to an embodiment of the present application.

Detailed Description

Various exemplary embodiments, features and aspects of the application will be described in detail below with reference to the drawings. In the drawings, like reference numbers indicate identical or functionally similar elements. Although various aspects of the embodiments are illustrated in the accompanying drawings, the drawings are not necessarily drawn to scale unless specifically indicated.

Reference in the specification to "one embodiment" or "some embodiments" or the like means that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the application. Thus, appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," and the like in the specification are not necessarily all referring to the same embodiment, but mean "one or more but not all embodiments" unless expressly specified otherwise. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless expressly specified otherwise.

In the present application, "at least one" means one or more, and "a plurality" means two or more. "and/or", describes an association relationship of an association object, and indicates that there may be three relationships, for example, a and/or B, and may indicate: including the case where a alone exists, both a and B together, and B alone, where a, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (one) of a, b, or c may represent: a, b, c, a-b, a-c, b-c, or a-b-c, wherein a, b, c may be single or plural.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

The word "exemplary" is used herein to mean "serving as an example, embodiment, or illustration. Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. In addition, numerous specific details are set forth in the following description in order to provide a better illustration of the application. It will be understood by those skilled in the art that the present application may be practiced without some of these specific details.

The air ticket revenue management has ideal implementation effect on international aviators, and usually 3% -6% of revenue improvement can be generated after the international aviators implement the revenue management, and the main reasons that the international revenue management system can fall to the ground are as follows: (1) The international route is a typical hub-type route network, and has few flights, few co-flying companies and relatively stable transport capacity structure; (2) Foreign passengers are usually reserved before flights take off for 2-3 weeks, so that foreign market demands (middle stages) are predicted accurately, and the prediction accuracy means that price war can be accurately played. TMC company acts as a voyage agent and should have its revenue management system, but no system or product related to voyage return agreement is found from the current disclosure, and the upper layer of revenue management data is not applied.

In the related art, a Seasonal trend algorithm (seasal-Trend decomposition procedure based on Loess, STL) based on local weighted regression may be used to predict the total fare and/or leg volume of the air ticket sales order over a specified period of time in the airline commission return agreement, thereby providing a reference for how to reasonably complete the airline commission return agreement. The STL algorithm is a better-effect and robust algorithm in time sequence decomposition, and can decompose the time sequence into a trend term sequence, a period term sequence and a residual term sequence. The STL algorithm has the following advantages: (1) The period term can change along with time, and the change rate can be customized by a user; (2) trend items may be user-defined; (3) Insensitive to outliers, but may make local residuals large.

The STL algorithm performs local polynomial regression on the periodic and trend terms by local weighted regression (Locally weighted regression, loess). The Loess algorithm is a robust regression algorithm, is a common method for smoothing a two-dimensional scatter diagram, and combines the simplicity of traditional linear regression with the flexibility of nonlinear regression. When a certain response variable is to be estimated, a data subset is firstly taken from the vicinity of the predicted variable, then linear regression or quadratic regression is carried out on the subset, a weighted least square method is adopted in regression, namely the closer to the estimated point, the larger the weight is, the value of the corresponding variable is estimated by utilizing the obtained local regression model, and the whole fitting curve is obtained by carrying out point-by-point operation by the method. The Robust local weighted regression algorithm (robustLoess) is implemented as follows:

(1) Selecting proper window number, for each observation point x _i (i=1, 2, …, n) as much as x _i The window width is selected for the center.

(2) The weights of all points within the window are defined, the weights being determined by a weight function.

(3) Each with a weight w by using least square method _k (x _i ) Observation point (x) _i ,y _i ) Calculate regression coefficient alpha (x _i ) Estimate of (2)At this time, y is obtained _i At x _i Fitting value at

(4) Let B (z) be the defined 4 th-order weight function:

let the residual of the fitting valueS is |e _i Median of I, define +.>

(5) For each i, in (x _i ,y _i ) Use xi at the point _k w _k (x _i ) Instead of the original weight w _k (x _i ) Calculating a d-order polynomial fit by using a least square method, and calculating a new one

In general, when fitting an observation point by using the Loess algorithm, the polynomial order, the weight function, the iteration number and the window width are very important, wherein the polynomial order, the weight function and the iteration number can be given in advance.

The local weighted regression process and the robustness process of the robust local weighted regression algorithm are implemented in the inner loop and the outer loop of the STL algorithm, respectively. Fig. 1 shows a flow chart of an STL algorithm according to an embodiment of the present application. As shown in fig. 1, the inner loop step of the STL algorithm includes:

s100, giving an initial value.

Trend term T _v The initial value of (1) is given as 0, i.e

S101, removing trend items.

From the original time sequence Y _v Subtracting the trend term sequence in the k-1 iterationWherein k=1, 2, …, inner, inner is the number of inner layer cycles, original time series Y _v The length of (2) is denoted as N.

S102, smoothing the periodic subsequence.

The sample points at the same position in each period in the original time sequence form a sub-sequence (i.e. a period sub-sequence), and the period sub-sequence can be calculated to obtain n _p And each. Smoothing each periodic sub-sequence by using Loess algorithm, wherein the smoothing parameter of Loess is denoted as parameter _c And extending a time point forwards and backwards respectively, and combining the smoothing results to obtain a temporary periodic component in the kth iterationWherein v= -n _p +1,…,1,2,…N,…,N+n _p Temporary period component->Length of (2) is N+2n _p 。

S103, low-flux filtering of the periodic subsequence.

For the temporary periodic component obtained in step S102Sequentially with length n _p ，n _p 3, thenSmoothing by using Loess algorithm, wherein the smoothing parameter of Loess is recorded as parameter _l Obtaining the low flux of the periodic subsequence in the kth iteration>Wherein v=1, 2, … N, < >>Is N in length.

S104, removing the smooth period subsequence trend.

From temporary periodic componentsSubtracting the low flux +.>Obtaining the periodic item sequence in the kth iteration +.>I.e. < ->

S105, removing the period item.

From the original time sequence Y _v Subtracting the periodic term sequence obtained in step S104 fromObtaining the intermediate sequence->I.e. < ->

S106, trend smoothing.

Using the Loess algorithm for the intermediate sequence obtained in step S105Smoothing, and recording the smoothing parameter of Loess as parameter _t Obtaining trend item sequence +.>

S107, judging whether convergence is achieved.

Determining a periodic term sequence in a kth iterationAnd trend item sequence->If it is, stopping the iteration and adding the periodic sequence of items in the kth iteration +.>As the original time series Y _v Corresponding periodic item sequence S _v The trend item sequence in the kth iteration is +.>As the original time series Y _v Corresponding trend item sequence T _v From the original time sequence Y _v Subtracting its corresponding periodic item sequence S _v And trend item sequence T _v The original time sequence Y can be obtained _v Corresponding residual term sequence R _v I.e. R _v ＝Y _v -S _v -T _v The method comprises the steps of carrying out a first treatment on the surface of the If not, returning to the step S101, and continuing iteration.

The outer loop of the STL algorithm is mainly used to adjust the weights, and if there are outliers in the time series, the residual term will be larger. Defining a weight function w as:

h＝6*median(|R _v |)

w＝(1-(|R _v |/h)^2)^2

In the inner loop of each iteration, when the Loess regression is performed in step S102 and step S106, the neighborhood weight needs to be updated to w to reduce the influence of the outlier on the regression.

In order to make the STL algorithm have enough robustness, the inner layer cycle and the outer layer cycle are designed, and when the number of times of the inner layer cycle is enough, the trend term sequence and the period term sequence are converged when the inner layer cycle is ended; if there is no significant outlier in the time series, the outer loop number may be set to 0.

In the STL decomposition algorithm, the Loess algorithm adopted in the steps S102, S103 and S106 only performs regression prediction by means of the sequence itself and the peripheral values, resulting in inaccurate prediction, thereby reasonably completing the navigation commission return protocol.

In order to reasonably complete the airline department commission return agreement, the embodiment of the application provides an air ticket sales order prediction method.

FIG. 2 illustrates a flow chart of a method of air ticket sales order forecasting according to an embodiment of the present application. As shown in fig. 2, the method includes:

s201, acquiring air ticket order history data.

Illustratively, the ticket order history data may include sales volume, total fare, leg volume, trip information, commission amount, etc. of ticket sales orders over a period of time; for example, data may be included for sales of ticket sales orders, total fare, leg amounts, trip information, return amounts, etc. pushed two years forward from the current time.

S202, obtaining target prediction data according to historical target data and at least one feature; wherein the historical target data comprises first order quantity information of the air ticket sales order in a first time period in the air ticket order historical data; the at least one feature is constructed according to the travel information of the ticket sales order in the ticket order history data; the target forecast data includes second order quantity information for the ticket sales order over a specified time period in at least one airline share-back agreement.

For example, the first order quantity information may include information reflecting an order quantity such as a total fare and/or a leg quantity, and the history target data may include a total fare and/or a leg quantity of the ticket sales order in a first time period corresponding to at least one airline rebate agreement in the ticket order history data, that is, the total fare and/or the leg quantity of the ticket sales order satisfying the specified airline, the specified departure-arrival city, and the specified leg in the airline rebate agreement in the first time period. For example, the first order quantity information may include a return commission amount, and the historical target data may include a return commission amount obtained for the ticket sales order in a first time period corresponding to the at least one airline return commission agreement in the ticket order history data, i.e., a return commission amount obtained for the ticket sales order satisfying a specified airline, a specified airline segment, a specified flight number, a specified cabin combination in the airline return commission agreement in the first time period. Illustratively, the first time period may be a period of time from the current time forward, for example, may be 2 years forward from the current time, such as 2022, 6, 27 days, then the first time period may be 2020, 6, 27 days to 2022, 6, 27 days. For example, the historical target data may be data that is statistically by day.

For example, a start-stop time period may be specified in the airline share-return commission protocol, and a target to be completed within this start-stop time period, and if the current time is not within the start-stop time period specified in the airline share-return commission protocol, the specified time period corresponding to the target prediction data may be the start-stop time period specified in the airline share-return commission protocol; for example, the current time is 2021, 12, 1, and the beginning and ending time specified by the airline commission return agreement is 2022, 1, and 2022, 12, 31, and the specified time period may be 2022, 1, and 2022, 12, 31; if the current time is within a start-stop time period specified in the airline share-back commission protocol, the specified time period corresponding to the target prediction data may be a termination time specified in the airline share-back commission protocol from the current time; for example, the current time is 2022, 8 and 1, and the beginning and ending time specified by the department commission agreement is 2022, 1 and 1, 2022, 12 and 31, and the specified time period may be 2022, 8, 1, 2022, 12, 31.

For example, the second order quantity information may include a total fare and/or leg quantity, and the target forecast data may include a total fare and/or leg quantity for the air ticket sales order for the specified leg, the specified departure-arrival city, and the specified leg for the specified time period in the at least one leg return agreement. For example, the second order quantity information may include a return commission amount, and the target forecast data may include a return commission amount available for the ticket sales order for the specified leg, the specified flight number, and the specified combination of slots in the at least one leg return commission agreement.

Illustratively, in step S202, the at least one feature may be constructed according to trip information of the ticket sales order corresponding to the first order quantity information in the ticket order history data, where the trip information may include one or more of date information, holiday information, city information, and business trip application information.

As one example, features may be constructed from date information and holiday information for ticket sales orders in ticket order history data. For example, holiday features may be constructed based on information in the ticket order history data, such as whether the departure date of the ticket sales order is a holiday, whether the departure date is one, two, three days before and after the holiday, and so on. Therefore, holiday characteristics are constructed according to the date information and holiday information of the ticket sales order in the ticket order history data, and the holiday characteristics are used for predicting target prediction data, so that more accurate prediction results can be obtained.

As one example, features may be constructed from city information and business trip application information for ticket sales orders in ticket order history data. For example, the number of business applications feature may be constructed from the number of business applications in the round-trip city corresponding to the ticket sales order in the ticket order history data. As an example, the departure cities corresponding to the air ticket sales orders in the last two years may be ranked from high to low according to the number of orders taking each city as a departure place, and a departure city list is established; the arriving cities corresponding to the air ticket sales orders in the past two years can be ordered from high to low according to the order quantity of each city as a destination, and an arriving city list is established; the round trip cities corresponding to the air ticket sales orders in the past two years can be ordered from high to low according to the number of orders taking the two cities as departure places to arrival places, and a departure-arrival city list is established; the business trip application number feature may be constructed from at least one of data such as the business trip application number with the city a as the departure place, the business trip application number with the city B as the destination, and the business trip application number with the city C as the departure place and the city D as the destination; wherein, the city A can be the first 11 cities in the departure city list; the B city may be the city that reaches the first 11 bits in the city list; the C city and D city may be the departure-arrival city of the first 15 in the departure-arrival city list. Therefore, since the airline department commission return protocol is related to the city and the airline, the business trip application number characteristics are constructed according to the round trip city corresponding to the ticket sales order and the business trip application number of the round trip city in the ticket order history data, and the business trip application number characteristics are used for predicting target prediction data, so that the accuracy of a prediction result is improved.

S203, determining whether to recommend an agreement flight to a user according to the target prediction data; wherein the agreement flights include flights corresponding to the at least one airline employment agreement.

Illustratively, the goal prediction data may include a total fare for the ticket sales order for a specified time period in at least one airline commission return agreement, further, a predicted goal completion rate may be calculated based on the predicted total fare (the goal completion rate = the total fare for the ticket sales order for the specified time period predicted/the total fare specified for the specified time period in the corresponding airline commission return agreement), a predicted return amount may be calculated based on the predicted total fare and the specification of the corresponding airline commission return agreement, and thus whether to recommend an agreement flight to the user may be determined based on the predicted goal completion rate and/or the predicted return amount of the airline commission return agreement. Illustratively, if the predicted target completion rate reaches a preset threshold, determining to recommend an agreement flight to the user; otherwise, not recommending flights to the user. Preferably, the interval of the preset threshold value may be [0.9,0.95]. Illustratively, if the predicted return amount reaches the desired return amount, determining to recommend an agreement flight to the user; otherwise, not recommending flights to the user; the ideal return amount may be set according to business needs.

As an example, a certain air department return commission agreement specifies a start-stop time period of 2022, 1 st, and 2022 nd 12 nd 31 st, and specifies a total fare interval of 1000 ten thousand yuan to 9999 ten thousand yuan for which a return commission can be obtained, that is, the total fare meeting the air ticket sales order specified by the air department return commission agreement in the time period of 2022, 1 st, and 2022 nd 12 th 31 st needs to reach 1000 ten thousand yuan for obtaining the return commission; if the current time is 2022, 1 st, the specified time period may be 2022, 1 st, 2022 nd, 12 nd and 31 st, the preset threshold may be 0.92, and if the ratio obtained by dividing the total fare price of the air ticket sales order conforming to the air ticket return commission agreement by 1000 ten thousand yuan exceeds 0.92 in the predicted period of 2022, 1 st, 2022, 12 nd and 31 st, the minimum target specified by the air ticket return commission agreement is likely to be achieved, and a flight specified in the air ticket return commission agreement (namely, an agreement flight) can be recommended to the user when the user inquires about the flight, so that the completion rate of the air ticket return commission agreement is improved, and the benefit brought by the air ticket return commission agreement is improved.

In this way, according to the historical target data and at least one feature of the air ticket sales order, target prediction data of the air ticket sales order is obtained, and accuracy of the prediction result is improved; and determining whether to recommend flights specified in the airline share return agreement to the user according to the target prediction data of the air ticket sales order, so that the airline share return agreement can be reasonably completed, the completion rate of the airline share return agreement can be improved, and the income brought by the airline share return agreement can be improved.

In one possible implementation, in the event that a recommendation for a protocol flight to a user is determined, deriving a recommendation index for at least one protocol flight based on the user profile and the protocol flight profile; wherein, the user portrait is constructed according to the historical order information of the user; the protocol flight portrait is constructed according to the information of the protocol flight; and recommending the agreement flight to the user according to the recommendation index.

It should be noted that, under the condition that the user is determined to recommend the protocol flight, if the user does not have the protocol flight in the flight list conforming to the user searching condition when searching for the flight, the recommendation flow is ended, and the user is not recommended for the flight; if there are agreement flights in the flight list, the following recommended flow is entered.

Illustratively, the user portrait can be constructed according to the information of flight departure time period, arrival time period, air route, machine type, price index and the like in the historical order information of the user; where the price index may represent how close the flight's ticket price is to the lowest price in the searched results when the current user searches, for example, price index= (economy class no discount price for the flight-the flight's ticket price when the current user searches)/(economy class no discount price for the flight-the lowest price for the ticket searched by the current user).

FIG. 3 illustrates a schematic diagram of a user representation in accordance with an implementation of the application. As shown in FIG. 3, the user representation may include information such as air-route preference, departure time preference, model preference, price index preference, etc., and the preference coefficients of the various preferences may be calculated by dividing the frequency of orders of the category to which the user's historical orders belong by the total number of orders; for example, the total number of historical orders of a user is n, wherein the number of orders of an A-voyage is k, the number of orders of a B-voyage is s, the number of orders of a C-voyage is n-k-s, the preference coefficient of the A-voyage of the user is k/n, the preference coefficient of the B-voyage is s/n, the preference coefficient of the C-voyage is (n-k-s)/n, and the preference coefficients of other voyages are all 0. Preference coefficients for departure time preference, model preference, price index preference, etc. may also be calculated in a similar manner. Illustratively, if there is no historical order information for the user, a user representation may be constructed from the historical order information for the user's co-department co-workers, i.e., the user representation for the user may be used as the user representation for the user's co-department co-workers; if the historical order information of the co-workers of the user is not available, the user portrait can be constructed according to the historical order information of the company where the user is located for purchasing the air ticket, and the company portrait of the company where the user is located can be used as the user portrait of the user.

For example, the agreement flight portrayal may be constructed from information on whether the agreement flight is a direct flight, a departure time period, an arrival time period, a flight share, a price index, a model, and the like. Illustratively, the preference characteristics of the protocol flights may be constructed according to the preference types contained in the user profile information, and the preference characteristics of the protocol flights may be in one-to-one correspondence with the preference types of the users; and the characteristic value corresponding to each preference feature of the protocol flight is 0 or 1 according to the preference coefficient corresponding to each preference type in the user portrait information and the protocol flight information, wherein the number of the characteristic values corresponding to each preference feature of the protocol flight is the same as the number of the preference coefficients corresponding to each preference type in the user portrait information. For example, the user portrait information comprises a voyage preference and a model preference, wherein the voyage preference comprises three preference coefficients of an A voyage preference coefficient, a B voyage preference coefficient and a C voyage preference coefficient, and the model preference comprises three preference coefficients of a large-scale preference coefficient, a medium-scale preference coefficient and a small-scale preference coefficient; the airline company to which the protocol flight belongs is an A airline company, and the model of the protocol flight is small; the preference characteristics of the protocol flight constructed according to the user portrait information comprise a voyage preference characteristic and a model preference characteristic, wherein the voyage preference characteristic comprises an A voyage, a B voyage and a C voyage (respectively corresponding to an A voyage preference coefficient, a B voyage preference coefficient and a C voyage preference coefficient in the user portrait information), and the protocol flight belongs to the A voyage, so that the value of the A voyage is 1, and the values of the B voyage and the C voyage are 0; the model preference characteristics comprise large, medium and small (respectively corresponding to the large preference coefficient, the medium preference coefficient and the small preference coefficient in the image information of the user), and the small value can be 1 and the large and medium values can be 0 because the model of the protocol flight is small. According to the preference coefficient corresponding to each preference type in the user portrait information and the feature value corresponding to each preference feature of the protocol flight, calculating the recommendation index corresponding to each preference feature of the protocol flight according to the following formula:

Wherein c _i Recommendation index corresponding to ith preference feature of agreement flight, S _i Representing the number of feature values, a, corresponding to the ith preference feature of a protocol flight _i,j Representing the type of preference of the user (and agreement navigationThe preference type corresponding to the ith preference feature of the class), x _i,j And the j feature value corresponding to the i-th preference feature of the protocol flight is represented. For example, the 1 st preference feature of the protocol flight is a model preference feature (i.e., i=1 in the above formula), the model of the protocol flight is large, and the model preferences of the user include 3 preference coefficients of large preference coefficient, medium preference coefficient, and small preference coefficient, wherein the large preference coefficient is 0.3 (i.e., a in the above formula _1,1 =0.3), the medium preference coefficient is 0.4 (i.e., a in the above formula _1,2 =0.4) and the small preference coefficient is 0.3 (i.e., a in the above formula _1,3 =0.3), the number of feature values corresponding to the model preference feature of the protocol flight is 3 (i.e. S in the above formula ₁ =3), the model preference feature has a large value of 1 (i.e., x in the above formula _1,1 =1), the medium and small values are 0 (i.e., x in the above formula _1,2 ＝0，x _1,3 =0); the model recommendation index c of the protocol flight can be calculated according to the formula ₁ ＝0.3×1+0.4×0+0.3×1＝0.3。

After the recommendation index corresponding to each preference feature of the agreement flight is calculated, the recommendation indexes corresponding to the preference features can be added to obtain the recommendation index of the agreement flight, and the calculation formula is as follows:

where N represents the number of preference characteristics of the agreement flight (i.e., the number of preference types of the user). After calculating the recommendation index of each protocol flight, recommending the protocol flight with the highest recommendation index to the user; and a recommendation list of the protocol flights can be output to the user, and all the protocol flights in the recommendation list can be arranged from high to low according to the recommendation index. For example, when the user searches for flights, the flights meeting the user search conditions include flights a, b and c, the recommendation index of the flight a is calculated to be 2.6, the recommendation index of the flight b is calculated to be 1.5, the recommendation index of the flight c is calculated to be 2.1, the flight a with the highest recommendation index can be recommended to the user, and the flights can be ranked from high to low according to the recommendation index, and a recommendation list of the flights is output to the user according to the ranks of the flights a, c and b.

FIG. 4 illustrates a schematic diagram of a user profile and a protocol flight profile in accordance with an implementation of the application. FIG. 4 (a) shows a schematic diagram of a user representation including 4 preference types of airline preferences, departure time preferences, model preferences, price index preferences, as shown in FIG. 4 (a), according to an implementation of the present application. FIG. 4 (b) is a schematic diagram of an agreement flight portrait according to an embodiment of the present application, where the agreement flight belongs to an A airline, the departure time is 11:00, the model is large, and the price index is 0.5, as shown in FIG. 4 (b). Based on the preference type and the information of the protocol flight contained in the user portrait information, the airline preference feature, departure time preference feature, model preference feature and price index preference feature of the protocol flight can be constructed (i.e. the number n=4 of preference features of the protocol flight in the above formula). If the protocol flight belongs to the A-department, the A-department value is 1, the B-department value and the C-department value are 0 in the preference characteristic of the navigation department, and the navigation department recommendation index C of the protocol flight can be calculated according to the preference coefficient corresponding to the navigation department preference of the user ₁ =0.3×1+0.2×0+0.5×0=0.3. The departure time of the protocol flight is 11:00, the departure time period [8:00,12:00 ] in the departure time preference feature is 1, the departure time period [0:00,4:00 ], [4:00,8:00 ], [12:00,16:00 ], [16:00, 20:00), [20:00, 24:00) is 0, and the departure time recommendation index c of the protocol flight can be calculated according to the preference coefficient corresponding to the departure time preference of the user ₂ =0.1×0+0.1×0+0.2×1+0.3×0+0.2×0+0.1×0=0.2. The model of the protocol flight is large, the large value is 1, the medium and small values are 0 in the model preference characteristics, and the model recommendation index c of the protocol flight can be calculated according to the preference coefficient corresponding to the model preference of the user ₃ =0.4×1+0.3×0+0.3×0=0.4. The price index of the agreement flight is 0.5, then in the price index preference feature, the price index interval (0.4,0.6]Takes a value of 1, and the price index interval (0,0.2)]、(0.2,0.4]、(0.6,0.8]、(0.8,1]The value is 0, and the price index recommendation index c of the agreement flight can be calculated according to the preference coefficient corresponding to the price index preference of the user ₄ =0.2×0+0.2×0+0.1×1+0.3×0+0.2×0=0.1. The calculated navigation recommendation index c of the flight of the protocol ₁ Departure time recommendation index c ₂ Model recommendation index c ₃ Price index recommendation index c ₄ Adding to obtain the recommendation index C=c of the flights of the protocol ₁ +c ₂ +c ₃ +c ₄ ＝0.3+0.2+0.4+0.1＝1。

Under the condition that the protocol flight is recommended to the user, the recommendation index of the protocol flight is calculated according to the user portrait and the protocol flight portrait, and the protocol flight is recommended to the user according to the recommendation index, so that the order taking rate of the user on the protocol flight ticket is improved, the impulse of the airline department commission returning protocol which does not finish the target can be increased, the completion rate of the airline department commission returning protocol is improved, and the income brought by the airline department commission returning protocol is increased.

In one possible implementation, the target forecast data in step S202 further includes third order quantity information for the ticket sales order over a second time period corresponding to the at least one airline employment agreement; and the third order quantity information of the air ticket sales order in the second time period corresponding to the at least one air department commission agreement is used for evaluating the air department commission agreement established in the second time period.

Illustratively, the second time period may be a time period after a flight control return protocol expiration time. For example, the third order quantity information may include a total fare and/or leg quantity, and the target forecast data may include a total fare and/or leg quantity for the air ticket sales order for the specified leg, the specified departure-arrival city, and the specified leg for the second time period corresponding to the at least one leg return agreement.

As an example, a department return commission agreement specifies a start-stop time of 2022, 1 st 1 nd day to 2022 nd 12 nd 31 nd, a second time period may be 2023, 1 st 1 nd to 2023 nd 12 nd 31 nd, the target prediction data is a total fare and/or a quantity of air-ticket segments according to the department return commission agreement specified by the department return commission agreement in a time period of 2023, 1 st 1 nd to 2023 nd 12 nd 31 nd, the target prediction data may be output to a business person, and a reference may be provided for the department return commission agreement in the time period of 2023, 1 st 1 nd to 2023 nd 12 nd 31 nd may be made for the business person, so that the business person may reasonably make the department return commission agreement of the next stage.

In this way, by predicting order quantity information of the airline hostwood return commission agreement for a period of time in the future, reference can be provided for formulating the airline hostwood return commission agreement of the next stage, and more reasonable airline hostwood return commission agreements can be formulated, so that the completion rate of the airline hostwood return commission agreements is improved, and the income brought by the airline hostwood return commission agreements is improved.

The following describes exemplary possible implementations of the target prediction data obtained in step S202 according to the historical target data and at least one feature.

Fig. 5 is a flowchart of a method for predicting an air ticket sales order according to an embodiment of the present application, and in step S102, obtaining target prediction data according to historical target data and at least one feature may include the following steps:

S2021, obtaining a historical time sequence according to the historical target data.

For example, historical target data that is counted by day (or may be counted by other time units) may be arranged in time order to obtain a historical time series. Table 1 shows a set of historical target data according to an embodiment of the application. As shown in table 1, the current time is 2022, 6 and 27 days, and the historical target data is the total fare and the return amount of the ticket sales order corresponding to the same airline return agreement pushed forward from the current time for 2 years, that is, the total fare of the ticket sales order satisfying the airline return agreement, the specified departure-arrival city, the specified leg, and the specified airline, the specified leg, the specified flight number, and the ticket sales order satisfying the combination of the specified cabin position in the airline return agreement within the period of 27 days 2020, 6 and 27 days 2022. For example, for a missing value in the historical target data, the missing value can be replaced by using the historical data of the same period at the same time on the missing value based on a method of interpolation of the historical value; for example, if the total fare on month 1 of 2021 is missing, the total fare on month 1 of 2019 may be used instead. The historical target data can be counted by day and arranged according to the date, and a historical time sequence can be obtained.

TABLE 1

S2022, inputting the historical time sequence and the at least one feature into a preset model to obtain the target prediction data; the preset model is obtained by training based on an STL time sequence decomposition algorithm, so that the STL algorithm is improved.

In a possible implementation manner, the preset model comprises a first sub-model, a second sub-model and a third sub-model; the first sub-model, the second sub-model and the third sub-model are obtained by training based on a regression algorithm; the inputting the historical time sequence and the at least one feature into a preset model to obtain the target prediction data may include:

(1) And obtaining an input sequence according to the historical time sequence and the at least one characteristic.

For example, the feature values corresponding to at least one feature may be arranged in a time sequence, and the input sequence may be formed together with the historical time sequence.

For example, for an input sequence, the input data at the same position per cycle may be grouped into one sub-sequence (i.e., cycle sub-sequence). Table 2 shows a set of input data according to an embodiment of the application. As shown in Table 2, the input data is the total face price and the corresponding feature values of n features of the air ticket sales order according to the air ticket return commission agreement in the period from 27 days of 2020 6 to 27 days of 2022 Wherein n.gtoreq.1, feature 1 may be, for example, the holiday feature in step S202 in FIG. 2, and feature 2 may be, for example, the business trip application number feature in step S202 in FIG. 2; setting the number of days of one period to 7 days, wherein the period value of the current time can be 1, namely the period value of 2022.06.27 is 1; for the previous period of the current time, namely 2022.06.20-2022.06.26, the period value is sequentially and time-sequentially 1,2 and … 7, namely 2022.06.20 is 1 and 2022.06.26 is 7; according to the method, the cycle value corresponding to each date of 2020.06.27-2022.06.27 can be obtained; all data corresponding to the date with the period value of i (i.e. the same position i) can form a period sub-sequence c _i Wherein i=1, 2, … 7; for example, the period values 2020.06.29, 2020.07.06, …, 2022.06.20, 2022.06.27, which are all 1, are the total fare value corresponding to the date and the feature value corresponding to each feature, may form a period sub-sequence c ₁ The method comprises the steps of carrying out a first treatment on the surface of the The input data may be arranged in a time sequence to obtain an input sequence.

TABLE 2

(2) Obtaining a trend item sequence in the k-1 iteration, and obtaining a first intermediate sequence according to the input sequence and the trend item sequence in the k-1 iteration; wherein the initial value of k is 1, and the initial value of the trend term sequence is 0.

Illustratively, the trend term sequence in the k-1 th iteration may be subtracted from the input sequence to yield a first intermediate sequence.

(3) And inputting the first intermediate sequence into the first sub-model to obtain a second intermediate sequence.

For example, the data of the same position in each period of the first intermediate sequence may be formed into a period sub-sequence, so as to obtain at least one period sub-sequence corresponding to the first intermediate sequence; each cycle subsequence can be input into the first sub-model for regression, and extended forward and backward for one cycle, and the results of regression of each cycle subsequence are combined to obtain the second intermediate sequence.

As an example, taking the input data in Table 2 as an example, the total fare corresponding to the date and the feature value corresponding to each feature with the period values of 2020.06.29, 2020.07.06, …, 2022.06.20, 2022.06.27 being 1 can form a period sub-sequence c ₁ C, adding ₁ The first sub-model is input for regression, and after one cycle of forward and backward expansion, the obtained prediction data includes 2020.06.22 (one cycle of forward expansion) and 2022.07.04 (one cycle of backward expansion) prediction data.

Illustratively, the first sub-model is a regression model obtained by training; in the training process, a part of data in an input sequence can be used as a training set, and the rest part of data is used as a test set; for example, the first 80% of the data may be used as a training set and the last 20% of the data may be used as a test set; at least one periodic subsequence corresponding to the training set can be used as a training sample, and the preset regression model is trained, so that a first sub-model is obtained.

(4) Based on the second intermediate sequence, a third intermediate sequence is obtained.

For example, the second intermediate sequence may be subjected to a low-pass filtered running average in turn, e.g. a length n in turn _p ，n _p 3, to obtain a third intermediate sequence, wherein n _p The number of the periodic subsequences corresponding to the second intermediate sequence. Sequentially carrying out the second intermediate sequence with the length of n _p ，n _p The process of calculating the running average of 3 is as follows:

wherein N represents the length of the periodic subsequence corresponding to the second intermediate sequence, C represents the second intermediate sequence, and ma1, ma2 and ma3 are respectively the second intermediate sequences and are sequentially processed with the length of N _p ，n _p And 3, the sequence ma3 is a third intermediate sequence.

(5) And inputting the third intermediate sequence into the second sub-model to obtain a fourth intermediate sequence.

For example, the third intermediate sequence may be input into the second sub-model for regression to obtain the fourth intermediate sequence.

The second sub-model is a regression model obtained through training, and in the training process, a part of data in the third intermediate sequence can be used as a training set, and the rest of data is used as a test set, so that a preset regression model is trained, and the second sub-model is obtained; for example, the first 80% of the data may be used as a training set and the last 20% of the data may be used as a test set.

(6) And obtaining a periodic term sequence and a fifth intermediate sequence in the kth iteration according to the input sequence, the second intermediate sequence and the fourth intermediate sequence.

Illustratively, the fourth intermediate sequence may be subtracted from the second intermediate sequence to obtain a periodic sequence of terms in the kth iteration; the fifth intermediate sequence may be obtained by subtracting the sequence of periodic terms in the kth iteration from the input sequence.

(7) And inputting the fifth intermediate sequence into the third sub-model to obtain a trend item sequence in the kth iteration.

Illustratively, the fifth intermediate sequence may be input to a third sub-model for regression, resulting in a sequence of trend terms in the kth iteration.

The third sub-model is a regression model obtained through training, and in the training process, a part of data in the fifth intermediate sequence can be used as a training set, and the rest of data is used as a test set, so that a preset regression model is trained, and the third sub-model is obtained; for example, the first 80% of the data may be used as a training set and the last 20% of the data may be used as a test set.

(8) Judging whether the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, stopping iteration under the condition that the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, determining the periodic item sequence in the last iteration as the periodic item sequence corresponding to the input sequence, and determining the trend item sequence in the last iteration as the trend item sequence corresponding to the input sequence; otherwise, the value of k is added by 1, and the step of acquiring the trend item sequence in the k-1 iteration is carried out in a return mode.

Illustratively, the input sequence may be subtracted from the periodic term sequence in the kth iteration and subtracted from the trend term sequence in the kth iteration to obtain the residual term sequence in the kth iteration. Illustratively, it may also be determined whether to stop the iteration based on whether the sequence of residual terms in the kth iteration converges.

(9) And obtaining the target prediction data according to the periodic item sequence corresponding to the input sequence and the trend item sequence corresponding to the input sequence.

For example, the input sequence may be subtracted from the periodic term sequence corresponding to the input sequence and then subtracted from the trend term sequence corresponding to the input sequence to obtain the residual term sequence corresponding to the input sequence, thereby completing the time sequence decomposition of the input sequence; further, the second order amount information of the air ticket sales order in the designated time period in the air ticket return agreement may be predicted based on the time sequence decomposition result of the input sequence, that is, the target prediction data may be obtained by prediction, for example, the second order amount information of the air ticket sales order in the designated time period in the air ticket return agreement may be predicted using the period item sequence, the trend item sequence, and the residual item sequence.

The first sub-model, the second sub-model and the third sub-model may be XGBoost models, which are integrated machine learning models, can be used for regression problems, do not need feature selection, and have very strong robustness and generalization. For example, during the model training process, for each iteration, the first sub-model, the second sub-model, and the third sub-model in steps (3), (5), and (7) above may share a set of super-parametrics, and the optimal super-parametrics is normalized to the minimum root mean square error (RMSE, root Mean Square Error) across the test set. Illustratively, the hyperparameters of the XGBoost model may include: max_depth: the maximum depth of the tree is mainly used for avoiding overfitting, and when the value is large, the model can learn more specific and local samples; min_child_weight: the minimum leaf node sample weight sum is also used for avoiding overfitting, and when the value is large, the model can be prevented from learning local special samples; subsamples: controlling the random sampling proportion of each tree; gamma: the minimum loss function drop value required for node splitting may be specified.

In the model training process, the first sub-model, the second sub-model and the third sub-model which are output in each iteration process can be stored in a prediction file, and after model training is finished, all the first sub-model, the second sub-model and the third sub-model stored in the prediction file can be compared; for all the first sub-models stored in the prediction file, selecting the first sub-model with the minimum RMSE on the test set as a trained first sub-model; for all second sub-models stored in the prediction file, selecting the second sub-model with the minimum RMSE on the test set as a trained second sub-model; for all third sub-models stored in the prediction file, selecting the third sub-model with the minimum RMSE on the test set as a trained third sub-model; according to the trained first sub-model, the trained second sub-model and the trained third sub-model, a trained preset model can be obtained, and therefore target prediction data can be predicted through the preset model.

Thus, unlike the existing method of directly inputting the historical time sequence into the traditional STL algorithm model for prediction, the traditional STL algorithm only depends on a single time sequence, and is difficult to extract trend items and period items with better effects; in the embodiment of the application, the historical time sequence and the characteristics are input into the preset model together to obtain the target prediction data, so that the extraction method of the periodic term sequence and the trend term sequence in the STL algorithm can be optimized, and the STL algorithm improvement is realized; in addition, unlike the existing method which does not consider long-period data, the method only relies on local data of a value to be predicted to construct a prediction model; the embodiment of the application solves the problem of dependence on local data, and can obtain more accurate time sequence decomposition results, thereby improving the accuracy of the prediction results and obtaining more accurate target prediction data.

Fig. 6 shows a flowchart of constructing a preset model according to an embodiment of the present application. As shown in fig. 6, a historical time series may be obtained according to the historical target data, and the process of constructing the historical time series may refer to step S2021 in fig. 5; the feature engineering may include two parts of the missing value processing and the feature construction, and the missing value processing may refer to step S2021 in fig. 5; holiday features and business trip application number features can be constructed according to the travel information of the ticket sales order in the ticket order history data, and the feature construction process can refer to step 202 in fig. 2; since most employees typically submit business trip applications within 2 weeks prior to business trip, the business trip application number features may be loaded into the build task of the pre-set model only within 2 weeks prior to the end of the airline employment agreement; the model construction can comprise data set division and algorithm type selection, wherein 80% of data in input data can be selected as a training set for model training, and the other 20% of data are used as a test set for model verification; the algorithm can select an STL algorithm, the historical time sequence and at least one characteristic can be input into an STL algorithm model for training, and the training process can refer to the steps (1) - (9); in the training process of the preset model, the first sub-model, the second sub-model and the third sub-model which are output in each iteration process of the STL algorithm can be stored in a prediction file, and when the periodic term sequence and the trend term sequence which are output by the STL algorithm model are converged, the training is stopped; after the training of the preset model is finished, the first sub-model, the second sub-model and the third sub-model stored in the prediction file can be compared; for all the first sub-models stored in the prediction file, selecting the first sub-model with the minimum RMSE on the test set as a trained first sub-model; for all second sub-models stored in the prediction file, selecting the second sub-model with the minimum RMSE on the test set as a trained second sub-model; for all third sub-models stored in the prediction file, selecting the third sub-model with the minimum RMSE on the test set as a trained third sub-model; according to the trained first sub-model, the trained second sub-model and the trained third sub-model, a trained preset model can be obtained, and therefore target prediction data can be predicted through the preset model.

Compared with the existing prediction method based on STL algorithm only requiring input sequence, sequence local data and weight, the prediction method provided by the embodiment of the application is improved to a full-quantity regression prediction method, besides the input sequence, external features such as holiday features, business trip application number features and the like are also input, the two features together form the input of a trend item sequence and a period item sequence in a time sequence, the extraction method of the period item sequence and the trend item sequence is optimized, the problem of dependence on local data is solved, and more accurate time sequence decomposition is realized by the method of accurate decomposition and divide-and-conquer of sequence information, so that the accuracy of a prediction result can be improved, and more accurate target prediction data is obtained.

The air ticket sales order prediction method provided by the embodiment of the application can be applied to an air ticket sales order prediction system. FIG. 7 shows a schematic diagram of an airline ticket sales order prediction system according to an embodiment of the present application. As shown in fig. 7, for a certain airline department commission returning agreement, the system predicts the total fare of the ticket sales order in a specified period of time according to the total fare and/or the amount of flight segments (i.e. agreement metering settlement data) of the ticket sales order meeting the requirement of the airline department commission returning agreement in the historical target data, so as to calculate the target completion rate (i.e. predicted target completion rate) of each airline department commission returning agreement at the end of the agreement period, and predicts the commission returning amount of each airline department commission returning agreement according to the commission returning data (i.e. agreement prize settlement data) meeting the requirement of the airline department commission returning agreement in the historical target data, and the prediction process can refer to steps S2021-S2022 in fig. 5; the system outputs the recommendation degree of each airline department commission agreement according to the predicted target completion rate of each airline department commission agreement, and the higher the predicted target completion rate of each airline department commission agreement is, the higher the recommendation degree of the corresponding agreement is; an impulse switch can be set for each airline department commission agreement, and service personnel can comprehensively evaluate the current progress of each airline department commission agreement and predict target completion rate to determine whether to open the impulse switch of each airline department commission agreement; for example, if the predicted target completion rate for a department return agreement is greater than 90%, an impulse switch for the department return agreement may be opened to impulse the department return agreement; illustratively, the impulse switches of multiple airline share-back protocols may be opened simultaneously; under the condition that an impulse switch of a return commission protocol of a certain airline is determined to be opened, if a user logs in a foreground to inquire an air ticket and a protocol flight exists in a flight list conforming to the search condition of the user, a protocol flight portrait can be constructed according to the information of the protocol flight, a user portrait can be constructed according to the historical order information of the user, and the method for constructing the protocol flight portrait and the user portrait can refer to the description corresponding to the above-mentioned figures 3 and 4; calculating recommendation indexes of all the protocol flights by combining the protocol flight figures and the user figures, wherein the calculation method of the recommendation indexes can refer to the description corresponding to the figure 4; recommending the protocol flight with the highest recommendation index to the user, or arranging the protocol flights from high to low according to the recommendation index of the protocol flight, outputting a recommendation list to the user flight, recommending the user to order through a front page, and completing the protocol target; if the user subscribes to the agreement flight, an impulse is completed for the airline return agreement; this process is known as revenue recommendation. When it is determined that the agreement flight is recommended to the user, if no agreement flight exists in the flight list, the recommendation flow is ended, and the user is not recommended for the flight. The system predicts the total fare interval of the air ticket sales order in a period of time in the future of the airline department return agreement according to the agreement metering settlement data, outputs the prediction result to business personnel, and provides reference for reasonably evaluating the target of the next stage, and the process can be called as profit negotiable. The air ticket sales order prediction system of the embodiment of the application can embody the air ticket order data of each air ticket commission agreement in the air ticket profit management system, utilizes the air ticket order history data to complete two parts of contents of profit recommendation and profit negotiable, and provides a complete solution for business personnel to solve the target impulse and target formulation of the air ticket commission agreement.

Fig. 8 (a) - (c) show schematic diagrams of an air ticket sales order forecasting system according to an embodiment of the present application. As shown in fig. 8 (a), the air ticket sales order prediction system in the embodiment of the present application is divided into two parts of revenue recommendation and revenue counseling, including a protocol prediction module, an impulse switch and a recommendation module. The protocol prediction module may construct a time sequence prediction model according to the historical target data of each air route commission agreement, and predict the order quantity information of the air ticket sales order in the specified time period, and the prediction process may refer to steps S2021-S2022 in fig. 5; and calculates a protocol recommendation index for each airline share return protocol. The impulse switch is controlled by a service person, and the service person can determine whether to impulse the current protocol according to order quantity information and/or other prediction data obtained based on the order quantity information, for example, can determine whether to start the impulse switch of a certain airline department commission returning protocol according to the current completion rate, the prediction target completion rate, the prediction commission returning amount and the like of the airline department commission returning protocol; if the user is started, entering a recommendation module, otherwise, not entering the recommendation module; in addition, order quantity information and/or other predictive data may also provide scientific advice guidance for business personnel to evaluate total fare intervals for a specified period of time. After starting an impulse switch of a department return protocol, if a user inquires about a certain airline flight, if the user has a protocol flight, calculating a recommendation index of the protocol flight according to a user portrait formed by historical order information of the user and a protocol flight portrait formed by information of the protocol flight, wherein the method for constructing the protocol flight portrait and the user portrait and calculating the recommendation index of the protocol flight can refer to the description corresponding to the above-mentioned figures 3 and 4; the flight with the highest recommendation index can be recommended to the user; if the user subscribes to the protocol flight, a protocol impulse is completed. FIG. 8 (b) shows a schematic diagram of a protocol prediction module of an airline ticket sales order prediction system according to an embodiment of the present application. As shown in fig. 8 (b), the protocol prediction module trains the prediction model according to the historical target data in the ticket order historical data; wherein, the prediction model may be trained based on the STL algorithm with external features introduced, and the training process may refer to step S2022 in fig. 5; after obtaining the trained prediction model, the protocol prediction module can predict the total fare and the return amount of the air ticket sales order of each air ticket return agreement in a specified time period, and according to the predicted total fare, the predicted target completion rate of each air ticket return agreement can be calculated, so as to obtain the agreement recommendation index of each air ticket return agreement. FIG. 8 (c) shows a schematic diagram of a recommendation module of an airline ticket sales order prediction system according to an embodiment of the present application. As shown in fig. 8 (c), in the case of determining to open the impulse switch of a certain flight commission returning protocol, if there is a corresponding protocol flight of the flight commission returning protocol in the flight list meeting the user searching condition when the user searches for the air ticket, the recommendation module may construct a protocol flight portrait according to the information of the protocol flight, and the method for constructing the protocol flight portrait and the user portrait according to the historical order information of the user may refer to the descriptions corresponding to fig. 3 and 4; calculating recommendation indexes of all protocol flights by combining the protocol flight figures and the user figures so as to recommend flights to users, wherein the calculation method of the recommendation indexes of the protocol flights can refer to the corresponding description of the figure 4; if no protocol flight exists in the flight list meeting the user search condition, ending the recommendation flow and not recommending the flight. The air ticket sales order prediction system in the embodiment of the application can predict the target completion rate and the return amount of each airline department return agreement at the end of the agreement period, business personnel can evaluate whether to impulse the airline department return agreement of the incomplete target according to the prediction result, and calculate the recommendation index of the agreement flight by combining the agreement flight figure and the user figure, thereby recommending the agreement flight to the user to fulfill the agreement target; the method can also predict the total fare of the air ticket sales order meeting the requirements of the air ticket sales order history data meeting the requirements of the air ticket return commission protocol, and output the prediction result to business personnel to provide reference for reasonably evaluating the targets of the next stage; meanwhile, the air ticket sales order prediction system in the embodiment of the application trains the prediction model based on the STL algorithm introducing the external features, namely trains the prediction model through the improved STL algorithm, thereby improving the prediction accuracy of the model.

The performance of the air ticket sales order forecasting method and the air ticket sales order forecasting system provided by the application are respectively described below by taking the forecasting of total fare and the impulse of the air ticket return commission agreement as examples.

Fig. 9 shows a schematic diagram of a prediction curve of the air ticket sales order prediction method and the total fare of the air ticket return agreement according to the STL algorithm according to an embodiment of the present application. Wherein fig. 9 (a) shows a schematic diagram of a prediction curve for predicting a total fare of an air ticket sales order prediction method according to an embodiment of the present application, and fig. 9 (b) shows a schematic diagram of a prediction curve for predicting a total fare of an air ticket return agreement according to an STL algorithm. Table 3 shows the RMSE and aggregate data error (gap) of the air ticket sales order prediction method and the prediction of the total fare of the airline commission return agreement according to the STL algorithm, according to an embodiment of the application. As can be seen from FIG. 9 and Table 3, the air ticket sales order prediction method provided by the embodiment of the application has higher prediction accuracy and smaller prediction error compared with the conventional STL algorithm.

TABLE 3 Table 3

	RMSE	gap
			The application relates to an air ticket sales order prediction method	80856	0.057
STL algorithm	176763	0.095

As an example, a certain airline share commission agreement period is 2021-08-01 to 2021-08-31, and the total fare interval agreed with the airline is 1000 ten thousand yuan-9999 ten thousand yuan; that is, in the agreement period, the TMC company can complete the target only if the TMC company can complete the total fare of 1000 ten thousand yuan. At the end of the agreement period, the actual total fare of the air ticket sales order is 1249 ten thousand yuan, and the completion rate is calculated to be 124.9%. Table 4 shows a total fare prediction for the air ticket sales order prediction method according to an embodiment of the present application, starting the first day (i.e., 2021-08-01) for the airline commission return agreement, predicting to the end of the agreement period (i.e., 2021-08-31).

TABLE 4 Table 4

Date of day	Predicting total faresPrice (Yuan)	Real total fare (Yuan)	Whether or not to reach the standard (minimum target)	Error of
					2021.08.01	16834470	12490452	Is that	0.347787
2021.08.02	16917320	12490452	Is that	0.35442
					2021.08.03	26767780	12490452	Is that	1.14306
2021.08.04	25984690	12490452	Is that	1.080364
					2021.08.05	25320630	12490452	Is that	1.027199
2021.08.06	24613880	12490452	Is that	0.970616
					2021.08.07	24050160	12490452	Is that	0.925484
2021.08.08	16163380	12490452	Is that	0.294059
					2021.08.09	18723150	12490452	Is that	0.498997
2021.08.10	15911930	12490452	Is that	0.273927
					2021.08.11	15659310	12490452	Is that	0.253702
2021.08.12	14061100	12490452	Is that	0.125748
					2021.08.13	14167490	12490452	Is that	0.134266
2021.08.14	13653070	12490452	Is that	0.093081
					2021.08.15	13737790	12490452	Is that	0.099863
2021.08.16	13626490	12490452	Is that	0.090953
					2021.08.17	13484090	12490452	Is that	0.079552
2021.08.18	13323810	12490452	Is that	0.066717
					2021.08.19	12892080	12490452	Is that	0.032155
2021.08.20	12862440	12490452	Is that	0.029782
					2021.08.21	12783910	12490452	Is that	0.023495
2021.08.22	12635940	12490452	Is that	0.011648
					2021.08.23	12508690	12490452	Is that	0.00146
2021.08.24	11517810	12490452	Is that	0.077871
					2021.08.25	11489770	12490452	Is that	0.080116
2021.08.26	11303190	12490452	Is that	0.095053
					2021.08.27	12203360	12490452	Is that	0.022985
2021.08.28	11885860	12490452	Is that	0.048404
					2021.08.29	12257470	12490452	Is that	0.018653
2021.08.30	12420140	12490452	Is that	0.005629
					2021.08.31	12538770	12490452	Is that	0.003868
Average of	12557154	12490452	Is that	0.046365

It can be found from the prediction data of table 4 that the first day from the beginning of the protocol, i.e., the prediction can complete the 1000 ten thousand yuan goal, and the average error of the predicted value and the true value is only 4%. According to the air ticket sales order prediction method provided by the embodiment of the application, the external characteristics are introduced, the historical time sequence and the characteristics are input into the preset model together to obtain the target prediction data, so that the extraction method of the periodic term sequence and the trend term sequence in the STL algorithm can be optimized, the problem of dependence on local data is solved, and the accuracy of the prediction result is improved.

FIG. 10 illustrates a target lift rate schematic of an airline ticket sales order prediction system after impulsion of three airline share return agreements in accordance with an embodiment of the present application. As shown in fig. 10, after the platform is online on the line of the air ticket sales order prediction system provided by the embodiment of the application on the 1 st of 2021 in 12 th year, the platform is stopped on the 31 rd of 2022 in 3 rd year, the return commission agreements of three air ticket are imputed by the income recommending subsystem of the air ticket sales order prediction system, and the single target lifting rates under the recommendation of 1.00%, 1.12% and 0.76% are respectively realized by agreeing on the flight recommendation. The target impulse for the completed 346 ten thousand yuan ticket sales order is accumulated. The air ticket sales order prediction method provided by the embodiment of the application can provide air ticket income management for TMC companies, and business personnel can determine whether to impulse the air ticket return commission protocol of an incomplete target according to the prediction data of the air ticket sales order prediction system, thereby improving the completion rate of the air ticket return commission protocol and improving the income brought by the air ticket return commission protocol.

Based on the same inventive concept of the above method embodiments, the embodiments of the present application also provide an air ticket sales order predicting device, which may be used to execute the technical solutions described in the above method embodiments. For example, the steps of the methods shown in fig. 2, 5, or 6 described above may be performed.

FIG. 11 shows a block diagram of an airline ticket sales order forecasting apparatus according to an embodiment of the present application, as shown in FIG. 11, including: an acquisition module 1101, configured to acquire ticket order history data; a prediction module 1102, configured to obtain target prediction data according to the historical target data and at least one feature; wherein the historical target data comprises first order quantity information of the air ticket sales order in a first time period in the air ticket order historical data; the at least one feature is constructed according to the travel information of the ticket sales order in the ticket order history data; the target forecast data comprises second order quantity information of the air ticket sales order in a specified time period in at least one air department return agreement; a recommending module 1103, configured to determine whether to recommend a protocol flight to a user according to the target prediction data; wherein the agreement flights include flights corresponding to the at least one airline employment agreement.

According to the embodiment of the application, the target prediction data of the air ticket sales order is obtained according to the historical target data of the air ticket sales order and at least one feature, and the at least one feature is constructed according to the travel information of the air ticket sales order in the air ticket order historical data, so that the accuracy of the prediction result is improved; and determining whether to recommend flights specified in the airline share return agreement to the user according to the target prediction data of the air ticket sales order, so that the airline share return agreement can be reasonably completed, the completion rate of the airline share return agreement can be improved, and the income brought by the airline share return agreement can be improved.

In a possible implementation manner, the recommendation module 1103 is further configured to: under the condition that the recommendation of the protocol flight to the user is determined, obtaining a recommendation index of at least one protocol flight according to the user portrait and the protocol flight portrait; wherein, the user portrait is constructed according to the historical order information of the user; the protocol flight portrait is constructed according to the information of the protocol flight; and recommending the agreement flight to the user according to the recommendation index.

In one possible implementation, the prediction module 1102 is further configured to: obtaining a historical time sequence according to the historical target data; inputting the historical time sequence and the at least one feature into a preset model to obtain the target prediction data; the preset model is obtained by training based on an STL time sequence decomposition algorithm.

In a possible implementation manner, the preset model comprises a first sub-model, a second sub-model and a third sub-model; the first sub-model, the second sub-model and the third sub-model are obtained by training based on a regression algorithm; the prediction module is further configured to: obtaining an input sequence according to the historical time sequence and the at least one feature; obtaining a trend item sequence in the k-1 iteration, and obtaining a first intermediate sequence according to the input sequence and the trend item sequence in the k-1 iteration; wherein the initial value of k is 1, and the initial value of the trend item sequence is 0; inputting the first intermediate sequence into the first sub-model to obtain a second intermediate sequence; obtaining a third intermediate sequence according to the second intermediate sequence; inputting the third intermediate sequence into the second sub-model to obtain a fourth intermediate sequence; according to the input sequence, the second intermediate sequence and the fourth intermediate sequence, a periodic item sequence and a fifth intermediate sequence in the kth iteration are obtained; inputting the fifth intermediate sequence into the third sub-model to obtain a trend item sequence in the kth iteration; judging whether the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, stopping iteration under the condition that the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, determining the periodic item sequence in the last iteration as the periodic item sequence corresponding to the input sequence, and determining the trend item sequence in the last iteration as the trend item sequence corresponding to the input sequence; otherwise, adding 1 to the value of k, and returning to execute the step of acquiring the trend item sequence in the k-1 iteration; and obtaining the target prediction data according to the periodic item sequence corresponding to the input sequence and the trend item sequence corresponding to the input sequence.

In one possible implementation, the prediction module 1102 is further configured to: subtracting the fourth intermediate sequence from the second intermediate sequence to obtain a periodic item sequence in the kth iteration; subtracting the periodic term sequence in the kth iteration from the input sequence to obtain the fifth intermediate sequence.

In one possible implementation, the target forecast data further includes third order quantity information for the ticket sales order over a second time period corresponding to the at least one airline share-back agreement; and the third order quantity information of the air ticket sales order in the second time period corresponding to the at least one air department commission agreement is used for evaluating the air department commission agreement established in the second time period.

In one possible implementation, the first order quantity information and the second order quantity information include one or more of a total fare, a leg quantity, and a return amount.

In one possible implementation, the trip information includes one or more of date information, holiday information, city information, and business trip application information.

The technical effects and specific descriptions of the ticket sales order predicting device and the various possible implementation manners thereof shown in fig. 11 may be referred to the ticket sales order predicting method, and are not repeated herein.

It should be understood that the division of the modules in the above air ticket sales order predicting device is only a division of a logic function, and may be fully or partially integrated into a physical entity or may be physically separated. Furthermore, modules in the apparatus may be implemented in the form of processor-invoked software; the device comprises, for example, a processor, which is connected to a memory, in which instructions are stored, the processor calling the instructions stored in the memory to implement any of the above methods or to implement the functions of the modules of the device, wherein the processor is, for example, a general-purpose processor, such as a central processing unit (Central Processing Unit, CPU) or microprocessor, and the memory is internal or external to the device. Alternatively, the modules in the apparatus may be implemented in the form of hardware circuitry, some or all of which may be implemented by the design of hardware circuitry, which may be understood as one or more processors; for example, in one implementation, the hardware circuit is an application-specific integrated circuit (ASIC), and the functions of some or all of the above modules are implemented by the design of the logic relationships of elements within the circuit; for another example, in another implementation, the hardware circuit may be implemented by a programmable logic device (programmable logic device, PLD), for example, a field programmable gate array (Field Programmable Gate Array, FPGA), which may include a large number of logic gates, and the connection relationship between the logic gates is configured by a configuration file, so as to implement the functions of some or all of the above modules. All modules of the above device may be realized in the form of processor calling software, or in the form of hardware circuits, or in part in the form of processor calling software, and in the rest in the form of hardware circuits.

In an embodiment of the present application, the processor is a circuit with signal processing capability, in one implementation, the processor may be a circuit with instruction reading and running capability, such as a CPU, microprocessor, graphics processor (graphics processing unit, GPU), digital signal processor (digital signal processor, DSP), neural-network processor (neural-network processing unit, NPU), tensor processor (tensor processing unit, TPU), etc.; in another implementation, the processor may perform a function through a logical relationship of hardware circuitry that is fixed or reconfigurable, e.g., a hardware circuit implemented by the processor as an ASIC or PLD, such as an FPGA. In the reconfigurable hardware circuit, the processor loads the configuration document, and the process of implementing the configuration of the hardware circuit can be understood as a process of loading instructions by the processor to implement the functions of some or all of the above modules.

It will be seen that each module in the above apparatus may be one or more processors (or processing circuits) configured to implement the methods of the above embodiments, for example: CPU, GPU, NPU, TPU, microprocessor, DSP, ASIC, FPGA, or a combination of at least two of these processor forms. In addition, all or part of the modules in the above apparatus may be integrated together or may be implemented independently, which is not limited.

The embodiment of the application also provides electronic equipment, which comprises: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to implement the method of the above embodiments when executing the instructions. Illustratively, the steps of the methods shown in fig. 2, 5, or 6 described above may be performed.

Fig. 12 is a schematic structural view of an electronic device according to an embodiment of the present application, and as shown in fig. 12, the electronic device may include: at least one processor 1701, communication lines 1702, memory 1703, and at least one communication interface 704.

The processor 1701 may be a general purpose central processing unit, microprocessor, application specific integrated circuit, or one or more integrated circuits for controlling the execution of the program of the present application; the processor 1701 may also include a heterogeneous computing architecture of a plurality of general purpose processors, e.g., may be a combination of at least two of a CPU, GPU, microprocessor, DSP, ASIC, FPGA; as one example, the processor 1701 may be a cpu+gpu or cpu+asic or cpu+fpga.

Communication line 1702 may include a pathway to transfer information between the aforementioned components.

The communication interface 1704, using any transceiver or like device, is used to communicate with other devices or communication networks, such as ethernet, RAN, wireless local area network (wireless local area networks, WLAN), etc.

The memory 1703 may be, but is not limited to, a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (random access memory, RAM) or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (electrically erasable programmable read-only memory, EEPROM), a compact disc read-only memory (compact disc read-only memory) or other optical disk storage, a compact disc storage (including compact disc, laser disc, optical disc, digital versatile disc, blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. The memory may be self-contained and coupled to the processor via communication line 1702. The memory may also be integrated with the processor. The memory provided by embodiments of the present application may generally have non-volatility. The memory 1703 is used for storing computer-executable instructions for performing aspects of the present application, and is controlled by the processor 1701 for execution. The processor 1701 is configured to execute computer-executable instructions stored in the memory 1703, thereby implementing the method provided in the above-described embodiment of the present application; illustratively, the steps of the methods shown in fig. 2, 5, or 6 described above may be implemented.

Alternatively, the computer-executable instructions in the embodiments of the present application may be referred to as application program codes, which are not particularly limited in the embodiments of the present application.

Illustratively, the processor 1701 may include one or more CPUs, e.g., CPU0 in fig. 12; the processor 1701 may also include any one of a CPU, and GPU, ASIC, FPGA, for example, CPU0+ GPU0 or CPU0+ asic0 or CPU0+ FPGA0 in fig. 12.

By way of example, the electronic device may include multiple processors, such as processor 1701 and processor 1707 in fig. 12. Each of these processors may be a single-core (single-CPU) processor, a multi-core (multi-CPU) processor, or a heterogeneous computing architecture including a plurality of general-purpose processors. A processor herein may refer to one or more devices, circuits, and/or processing cores for processing data (e.g., computer program instructions).

In a particular implementation, the electronic device may also include an output device 1705 and an input device 1706, as one embodiment. The output device 1705 communicates with the processor 1701 and can display information in a variety of ways. For example, the output device 1705 may be a liquid crystal display (liquid crystal display, LCD), a light emitting diode (light emitting diode, LED) display device, a Cathode Ray Tube (CRT) display device, or a projector (projector) or the like, and may be, for example, a vehicle-mounted HUD, AR-HUD, display or the like. The input device 1706 is in communication with the processor 1701 and may receive input from a user in a variety of ways. For example, the input device 1706 may be a mouse, keyboard, touch screen device, or sensing device, among others.

An embodiment of the present application provides a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method of the above-described embodiment. Illustratively, the steps of the methods shown in fig. 2, 5, or 6 described above may be implemented.

Embodiments of the present application provide a computer program product, for example, which may include computer readable code, or a non-volatile computer readable storage medium bearing computer readable code; the computer program product, when run on a computer, causes the computer to perform the method in the above-described embodiments. Illustratively, the steps of the methods shown in fig. 2, 5, or 6 may be described above.

The computer readable storage medium may be a tangible device that can hold and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: portable computer disks, hard disks, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static Random Access Memory (SRAM), portable compact disk read-only memory (CD-ROM), digital Versatile Disks (DVD), memory sticks, floppy disks, mechanical coding devices, punch cards or in-groove structures such as punch cards or grooves having instructions stored thereon, and any suitable combination of the foregoing. Computer-readable storage media, as used herein, are not to be construed as transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., optical pulses through fiber optic cables), or electrical signals transmitted through wires.

The computer readable program instructions described herein may be downloaded from a computer readable storage medium to a respective computing/processing device or to an external computer or external storage device over a network, such as the internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmissions, wireless transmissions, routers, firewalls, switches, gateway computers and/or edge servers. The network interface card or network interface in each computing/processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium in the respective computing/processing device.

Computer program instructions for carrying out operations of the present application may be assembly instructions, instruction Set Architecture (ISA) instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, c++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions may be executed entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider). In some embodiments, aspects of the present application are implemented by personalizing electronic circuitry, such as programmable logic circuitry, field Programmable Gate Arrays (FPGAs), or Programmable Logic Arrays (PLAs), with state information for computer readable program instructions, which can execute the computer readable program instructions.

Various aspects of the present application are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer-readable program instructions.

These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, programmable data processing apparatus, and/or other devices to function in a particular manner, such that the computer readable medium having the instructions stored therein includes an article of manufacture including instructions which implement the function/act specified in the flowchart and/or block diagram block or blocks.

The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer, other programmable apparatus or other devices implement the functions/acts specified in the flowchart and/or block diagram block or blocks.

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The foregoing description of embodiments of the application has been presented for purposes of illustration and description, and is not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the various embodiments described. The terminology used herein was chosen in order to best explain the principles of the embodiments, the practical application, or the technical improvement of the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method of predicting an air ticket sales order, the method comprising:

acquiring air ticket order history data;

obtaining target prediction data according to the historical target data and at least one characteristic; wherein the historical target data comprises first order quantity information of the air ticket sales order in a first time period in the air ticket order historical data; the at least one feature is constructed according to the travel information of the ticket sales order in the ticket order history data; the target forecast data comprises second order quantity information of the air ticket sales order in a specified time period in at least one air department return agreement;

Determining whether to recommend a protocol flight to a user according to the target prediction data; wherein the agreement flights include flights corresponding to the at least one airline employment agreement.

2. The method according to claim 1, wherein the method further comprises:

under the condition that the recommendation of the protocol flight to the user is determined, obtaining a recommendation index of at least one protocol flight according to the user portrait and the protocol flight portrait; wherein, the user portrait is constructed according to the historical order information of the user; the protocol flight portrait is constructed according to the information of the protocol flight;

and recommending the agreement flight to the user according to the recommendation index.

3. The method according to claim 1 or 2, wherein the deriving target prediction data from historical target data and at least one feature comprises:

obtaining a historical time sequence according to the historical target data;

inputting the historical time sequence and the at least one feature into a preset model to obtain the target prediction data;

the preset model is obtained by training based on an STL time sequence decomposition algorithm.

4. A method according to claim 3, wherein the pre-set model comprises a first sub-model, a second sub-model, a third sub-model; the first sub-model, the second sub-model and the third sub-model are obtained by training based on a regression algorithm;

The step of inputting the historical time sequence and the at least one feature into a preset model to obtain the target prediction data comprises the following steps:

obtaining an input sequence according to the historical time sequence and the at least one feature;

obtaining a trend item sequence in the k-1 iteration, and obtaining a first intermediate sequence according to the input sequence and the trend item sequence in the k-1 iteration; wherein the initial value of k is 1, and the initial value of the trend item sequence is 0;

inputting the first intermediate sequence into the first sub-model to obtain a second intermediate sequence;

obtaining a third intermediate sequence according to the second intermediate sequence;

inputting the third intermediate sequence into the second sub-model to obtain a fourth intermediate sequence;

according to the input sequence, the second intermediate sequence and the fourth intermediate sequence, a periodic item sequence and a fifth intermediate sequence in the kth iteration are obtained;

inputting the fifth intermediate sequence into the third sub-model to obtain a trend item sequence in the kth iteration;

judging whether the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, stopping iteration under the condition that the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, determining the periodic item sequence in the last iteration as the periodic item sequence corresponding to the input sequence, and determining the trend item sequence in the last iteration as the trend item sequence corresponding to the input sequence; otherwise, adding 1 to the value of k, and returning to execute the step of acquiring the trend item sequence in the k-1 iteration;

And obtaining the target prediction data according to the periodic item sequence corresponding to the input sequence and the trend item sequence corresponding to the input sequence.

5. The method of claim 4, wherein the deriving the periodic sequence of terms and the fifth intermediate sequence in the kth iteration from the input sequence, the second intermediate sequence, and the fourth intermediate sequence comprises:

subtracting the fourth intermediate sequence from the second intermediate sequence to obtain a periodic item sequence in the kth iteration;

subtracting the periodic term sequence in the kth iteration from the input sequence to obtain the fifth intermediate sequence.

6. The method of any of claims 1-5, wherein the target forecast data further includes third order quantity information for the ticket sales order over a second time period corresponding to the at least one airline return agreement;

and the third order quantity information of the air ticket sales order in the second time period corresponding to the at least one air department commission agreement is used for evaluating the air department commission agreement established in the second time period.

7. The method of any of claims 1-6, wherein the first order quantity information and the second order quantity information include one or more of a total fare, a leg quantity, a return amount.

8. The method of any of claims 1-7, wherein the trip information includes one or more of date information, holiday information, city information, business trip application information.

9. An air ticket sales order forecasting device, the device comprising: the acquisition module is used for acquiring the history data of the air ticket order; the prediction module is used for obtaining target prediction data according to the historical target data and at least one characteristic; wherein the historical target data comprises first order quantity information of the air ticket sales order in a first time period in the air ticket order historical data; the at least one feature is constructed according to the travel information of the ticket sales order in the ticket order history data; the target forecast data comprises second order quantity information of the air ticket sales order in a specified time period in at least one air department return agreement; the recommending module is used for determining whether to recommend the protocol flight to the user according to the target prediction data; wherein the agreement flights include flights corresponding to the at least one airline employment agreement.

10. The apparatus of claim 9, wherein the recommendation module is further configured to: under the condition that the recommendation of the protocol flight to the user is determined, obtaining a recommendation index of at least one protocol flight according to the user portrait and the protocol flight portrait; wherein, the user portrait is constructed according to the historical order information of the user; the protocol flight portrait is constructed according to the information of the protocol flight; and recommending the agreement flight to the user according to the recommendation index.

11. The apparatus of claim 9 or 10, wherein the prediction module is further configured to: obtaining a historical time sequence according to the historical target data; inputting the historical time sequence and the at least one feature into a preset model to obtain the target prediction data; the preset model is obtained by training based on an STL time sequence decomposition algorithm.

12. The apparatus of claim 11, wherein the pre-set model comprises a first sub-model, a second sub-model, a third sub-model; the first sub-model, the second sub-model and the third sub-model are obtained by training based on a regression algorithm; the prediction module is further configured to: obtaining an input sequence according to the historical time sequence and the at least one feature; obtaining a trend item sequence in the k-1 iteration, and obtaining a first intermediate sequence according to the input sequence and the trend item sequence in the k-1 iteration; wherein the initial value of k is 1, and the initial value of the trend item sequence is 0; inputting the first intermediate sequence into the first sub-model to obtain a second intermediate sequence; obtaining a third intermediate sequence according to the second intermediate sequence; inputting the third intermediate sequence into the second sub-model to obtain a fourth intermediate sequence; according to the input sequence, the second intermediate sequence and the fourth intermediate sequence, a periodic item sequence and a fifth intermediate sequence in the kth iteration are obtained; inputting the fifth intermediate sequence into the third sub-model to obtain a trend item sequence in the kth iteration; judging whether the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, stopping iteration under the condition that the periodic item sequence in the kth iteration and the trend item sequence in the kth iteration are converged, determining the periodic item sequence in the last iteration as the periodic item sequence corresponding to the input sequence, and determining the trend item sequence in the last iteration as the trend item sequence corresponding to the input sequence; otherwise, adding 1 to the value of k, and returning to execute the step of acquiring the trend item sequence in the k-1 iteration; and obtaining the target prediction data according to the periodic item sequence corresponding to the input sequence and the trend item sequence corresponding to the input sequence.

13. The apparatus of claim 12, wherein the prediction module is further configured to: subtracting the fourth intermediate sequence from the second intermediate sequence to obtain a periodic item sequence in the kth iteration; subtracting the periodic term sequence in the kth iteration from the input sequence to obtain the fifth intermediate sequence.

14. The apparatus of any of claims 9-13, wherein the target forecast data further includes third order quantity information for the ticket sales order over a second time period corresponding to the at least one airline employment agreement; and the third order quantity information of the air ticket sales order in the second time period corresponding to the at least one air department commission agreement is used for evaluating the air department commission agreement established in the second time period.

15. The apparatus of any of claims 9-14, wherein the first order quantity information and the second order quantity information include one or more of a total fare, a leg quantity, a return amount.

16. The apparatus of any one of claims 9-15, wherein the trip information includes one or more of date information, holiday information, city information, business trip application information.

17. An electronic device, comprising:

a processor;

a memory for storing processor-executable instructions;

wherein the processor is configured to implement the method of any of claims 1-8 when executing the instructions.

18. A computer readable storage medium having stored thereon computer program instructions, which when executed by a processor, implement the method of any of claims 1-8.