WO2020024718A1

WO2020024718A1 - Method and device for predicting foreign exchange transaction volume

Info

Publication number: WO2020024718A1
Application number: PCT/CN2019/091735
Authority: WO
Inventors: 杨永晟
Original assignee: 阿里巴巴集团控股有限公司
Priority date: 2018-08-03
Filing date: 2019-06-18
Publication date: 2020-02-06
Also published as: CN109359758A; TW202020781A

Abstract

A method and device for predicting a foreign exchange transaction volume. The method comprises: obtaining time series data consisting of foreign exchange transaction business data series (102), wherein each foreign exchange transaction business data series comprises foreign exchange transaction business data of multiple dimensions within the same business day; and then, inputting the time series data to a pre-trained multilayer convolutional neural network to predict a foreign exchange transaction volume in a future time period (104). The mutual influence of the foreign exchange transaction business data between adjacent time nodes is considered during prediction, and the multi-dimensional characteristic attribute of a single time node is utilized, thereby improving the prediction accuracy.

Description

Method and device for predicting foreign exchange transaction volume

Technical field

This specification relates to the field of data processing technology, and particularly to a method and device for predicting foreign exchange transaction volume.

Background technique

In the international currency exchange business, it is necessary to reduce the potential risk of fluctuations in exchange rate exposure and control profit and loss by purchasing the foreign exchange transaction volume of the next foreign exchange purchase settlement cycle in advance. In order to carry out profit and loss control, it is necessary to forecast the foreign exchange transaction volume of each foreign exchange settlement cycle. Therefore, it is necessary to improve the forecasting method of foreign exchange transaction volume.

Summary of the invention

Based on this, this manual provides methods and devices for predicting foreign exchange transaction volume.

According to a first aspect of the embodiments of the present specification, a method for predicting foreign exchange transaction volume is provided. The method includes: acquiring time series data, the time series data including a foreign exchange transaction business data sequence for multiple business days in a historical time period, The foreign exchange transaction business data sequence is arranged in chronological order, and each foreign exchange transaction business data sequence includes multiple dimensions of foreign exchange transaction business data within the same business day; inputting the time series data into a pre-trained multilayer convolutional neural network A model to predict the volume of foreign exchange transactions in the future time period.

Optionally, the method further includes: calculating a prediction accuracy according to the predicted value and the real value of the foreign exchange transaction volume in the future time period; and modeling the model parameters of the multi-layer convolutional neural network model according to the prediction accuracy. Make adjustments.

Optionally, the method further comprises: acquiring historical foreign exchange transaction volumes for several periods in history; calculating an average value and a variance of the foreign exchange transaction volumes in the same period; and anomaly the predicted value of the foreign exchange transaction volume according to the average value and the variance. Detection.

Optionally, the step of abnormally detecting the predicted value of the foreign exchange transaction volume according to the mean and variance includes: if the predicted value of the foreign exchange transaction volume exceeds a preset numerical range, determining the prediction of the foreign exchange transaction volume The value is abnormal; wherein, the value range is: (X1, X2), where X1 = μ-N * δ, X2 = μ + N * δ; where μ is the average value of foreign exchange transactions during the same period, and N is the forecast Set a positive integer, δ is the variance of the foreign exchange transaction volume over the same period.

Optionally, the method further comprises: if the predicted value of the foreign exchange transaction volume is less than X1, modifying the predicted value of the foreign exchange transaction volume to X1; if the predicted value of the foreign exchange transaction volume is greater than X2, modifying the The forecast value of foreign exchange transaction volume was revised to X2.

Optionally, the step of inputting the time series data into a pre-trained multi-layer convolutional neural network model includes: calculating a foreign exchange transaction business feature of a key merchant based on the time series data; and entering a foreign exchange transaction business feature of the key merchant. Enter a pre-trained multilayer convolutional neural network model.

Optionally, the foreign exchange transaction business data includes forward transaction volume, reverse transaction volume, business day date attributes, business day promotion attributes, and / or business day preset marketing amounts for each business day.

According to a second aspect of the embodiments of the present specification, a foreign exchange transaction volume prediction device is provided. The device includes an acquisition module for acquiring time series data, and the time series data includes foreign exchange for multiple business days in a historical time period. Transaction business data sequence, the foreign exchange transaction data sequence is arranged in chronological order, and each foreign exchange transaction data sequence includes multiple dimensions of foreign exchange transaction business data within the same business day; a prediction module is configured to convert the time series data Enter a pre-trained multi-layer convolutional neural network model to predict foreign exchange trading volume in the future time period.

According to a third aspect of the embodiments of the present specification, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method according to any one of the embodiments is implemented.

According to a fourth aspect of the embodiments of the present specification, there is provided a computer device including a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements any implementation when the program is executed. Example method.

Apply the solution of the embodiment of this specification to obtain time series data composed of foreign exchange transaction business data sequences, and each foreign exchange transaction business data sequence includes multiple dimensions of foreign exchange transaction business data within the same business day, and then input the time series data into the pre-training Multi-layer convolutional neural network to predict the foreign exchange transaction volume in the future time period. In the prediction, the mutual influence of foreign exchange transaction business data between adjacent time nodes is considered, and the multi-dimensional characteristic attributes of a single time node are used. To improve prediction accuracy.

It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and should not limit the present specification.

BRIEF DESCRIPTION OF THE DRAWINGS

The drawings herein are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the specification, and together with the description serve to explain the principles of the specification.

FIG. 1 is a flowchart of a method for predicting foreign exchange transaction volume according to an embodiment of the present specification.

FIG. 2 is a schematic diagram of foreign exchange transaction data according to an embodiment of the present specification.

FIG. 3 is a schematic diagram of an abnormality detection process according to an embodiment of the present specification.

FIG. 4 is a block diagram of a foreign exchange transaction volume prediction device according to an embodiment of the present specification.

FIG. 5 is a schematic diagram of a computer device for implementing a method of an embodiment of the present specification according to an embodiment of the present specification.

detailed description

Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with this specification. Rather, they are merely examples of devices and methods consistent with certain aspects of the specification, as detailed in the appended claims.

The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the specification. As used in this specification and the appended claims, the singular forms "a", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

It should be understood that although the terms first, second, third, etc. may be used in this specification to describe various information, the information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this specification, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein can be interpreted as "at" or "when" or "in response to determination".

As shown in FIG. 1, it is a flowchart of a method for predicting foreign exchange transaction volume according to an embodiment of the present specification. The method may include the following steps:

Step 102: Obtain time series data. The time series data includes a foreign exchange transaction data sequence for multiple business days in a historical time period. The foreign exchange transaction data sequence is arranged in chronological order, and each foreign exchange transaction business data sequence includes Data on foreign exchange transactions in multiple dimensions within the same business day;

Step 104: input the time series data into a pre-trained multilayer convolutional neural network to predict the foreign exchange transaction volume in the future time period.

In step 102, time series data is first obtained, that is, a set of data arranged in chronological order. For the business scenario of foreign exchange transactions, you can obtain foreign exchange transaction business data generated on each business day, and arrange these business data according to the time of generation to obtain the required time series data. The business day is a statistical period of business volume, which can be the same as the natural day (0:00 to 23:59), or can be preset by the business system. For example, it can be set to 15 of a natural day: 00 to 14:59 the next day, or set to another time period.

You can obtain business data generated for multiple consecutive business days, and each business day obtains the same data. In order to comprehensively describe the characteristics of business data, multiple different dimensions of business data can be obtained. For example, these dimensions may include forward transaction volume, reverse transaction volume, business day date attributes, business day promotion attributes, and / or business day preset marketing amounts for each business day. In the consumption scenario, the user's purchase and payment behavior and the user's refund behavior bring about the opposite flow of funds. When the funds are settled, from the perspective of the merchant, the user's payment behavior is a forward transaction, and the user's refund behavior is a Reverse transactions, therefore, the forward transaction volume is the total amount of the user's payment to the merchant, and the reverse transaction volume is the total amount of the user's refund from the merchant. The business day date attribute is whether the date corresponding to the business day is at the beginning of the month, the end of the month, the middle of the month, the working day, the non-working day, or the holiday. The business day promotion attribute is whether or not there is a promotion activity and the activity level of the promotion activity on the business day. Among them, the activity level is determined by factors such as the discount strength of the promotion activity, the coverage of merchants, and the expected transaction volume. The preset sales amount for a business day can be a preset sales amount for a single business day, and different business days can be set for different preset sales amounts.

By obtaining time series data composed of foreign exchange transaction data series, and each foreign exchange transaction data series includes multiple dimensions of foreign exchange transaction business data within the same business day, the foreign exchange transaction business between adjacent time nodes is considered in the forecast. The data interact with each other, and the multi-dimensional feature attributes of a single time node are used to improve the prediction accuracy. A schematic diagram of foreign exchange transaction data of an embodiment is shown in FIG. 2. In the figure, T represents the selected predicted time node, and T + N represents the Nth business day after time node T.

In step 104, the time series data obtained in step 102 may be input to a multi-layer convolutional neural network model, and the multi-layer convolutional neural network model is used to predict the foreign exchange transaction volume in the future time period.

Multi-layer convolutional neural network models can be pre-trained. During training, the time series data in the historical time period can be used as input, the real value of the foreign exchange transaction volume in the historical time period can be used as the output, and the model parameters of the multi-layer convolutional neural network model can be solved. The multi-layer convolutional neural network model is used to solve the problem that conventional linear regression methods cannot mine the correlation between adjacent time nodes, and conventional time series processing methods cannot handle high-dimensional data nodes.

The multi-layer convolutional neural network can not only ensure the scalability of the high-dimensional features of each time node data, but also can ensure the sequential evolution between time nodes. At the same time, the neural network can control overfitting and automatically realize the selection of high-dimensional features. In addition, for refund and other services in foreign exchange transactions, the refund window is generally a fixed length. Convolutional neural networks train a relatively short amount of business window data and can capture the rules of short-term business windows.

In one embodiment, after obtaining the predicted value of the foreign exchange transaction volume in the future time period, the prediction accuracy may also be calculated according to the predicted value and the real value of the foreign exchange transaction volume in the future time period; The model parameters of the multi-layer convolutional neural network model are adjusted. In this way, the model parameters of the multi-layer convolutional neural network model can be dynamically adjusted, and the prediction accuracy can be improved.

In one embodiment, as shown in FIG. 3, the method in the embodiment of the present specification further includes:

Step 302: Obtain the historical foreign exchange transaction volume for several periods in history;

Step 304: Calculate the mean and variance of the foreign exchange transaction volume over the same period;

Step 306: Anomaly detection is performed on the predicted value of the foreign exchange transaction volume according to the mean and variance. Through anomaly detection, abnormal predicted values can be filtered out, thereby further improving prediction accuracy.

The period in this embodiment may be a time period including multiple business days with the same business day date attribute. It can be a time period that includes multiple business day date attributes as the business day of the beginning of the month. For example, a cycle includes 3 months, and the three business days include three business day date attributes as the business day of the beginning of the month. It may also be a time period that includes multiple business day date attributes as Sunday business days. For example, a period includes 4 weeks, and these 4 weeks include 4 business day date attributes with Sunday business days. For the same period, the foreign exchange transaction volume is the foreign exchange transaction volume with the same business day and date attributes. For example, it can be two or more business day dates. The date attributes of two or more business days are the foreign exchange transaction volume of the business day at the beginning of the month.

Anomaly detection is performed on the predicted value of the foreign exchange transaction volume according to the mean and variance, and a numerical range of the foreign exchange transaction volume may be set in advance. If the predicted value of the foreign exchange transaction volume exceeds a preset numerical range, determining that the predicted value of the foreign exchange transaction volume is abnormal; wherein the numerical range may be:

(X1, X2),

Among them, X1 = μ-N * δ, X2 = μ + N * δ;

In the formula, μ is the average value of the foreign exchange transaction volume in the same period, N is a preset positive integer, and δ is the variance of the foreign exchange transaction volume in the same period.

Further, if the predicted value of the foreign exchange transaction volume is less than X1, the predicted value of the foreign exchange transaction volume may be corrected to X1; if the predicted value of the foreign exchange transaction volume is greater than X2, the foreign exchange transaction volume may be predicted The value is corrected to X2.

In one embodiment, the step of inputting the time series data into a pre-trained multi-layer convolutional neural network model includes: calculating a foreign exchange transaction business feature of a key merchant according to the time series data; Business feature input is a pre-trained multilayer convolutional neural network model. In this embodiment, the key merchants may be the merchants whose trading volume ranks among the top several in the entire currency's trading volume on that day, or the average value of their trading volume in the historical time period (for example, one month) among the top merchants. A number of merchants can also be the top ones in terms of transaction volume on the same day, and the average number of merchants in the historical period of time also ranks in the top several merchants. Of course, other features or a combination of features can also be used to determine whether a merchant is a key merchant.

The characteristics of foreign exchange transactions of key merchants may include characteristics such as the proportion of key merchant transaction volume, the intensity of key merchant activities, and / or the proportion of key merchants' high-quality goods. Among them, the transaction volume ratio of key merchants is the ratio of the transaction volume of key merchants to the total transaction volume. The intensity of key merchant activities is used to characterize which level of promotion activities held by key merchants are on the trading platform's promotional activities. The level of activities can be divided according to the expected sales of key merchants. The proportion of key merchandise's high-quality goods can be calculated based on the ratio of the sales of the merchandise rated above a certain rating value to the total sales of the key merchant.

The influence of key merchants on the forecast is relatively large. Therefore, the introduction of foreign exchange transaction characteristics of key merchants as the basis for foreign exchange transaction volume prediction can improve the accuracy of the forecast.

The various technical features in the above embodiments can be arbitrarily combined, as long as there is no conflict or contradiction in the combination of features, but it is limited in space and has not been described one by one. Therefore, the various technical features in the above embodiments can be arbitrarily combined. It also belongs to the scope disclosed in this specification.

As shown in FIG. 4, an embodiment of the present specification further provides a device for predicting foreign exchange transaction volume. The device may include:

The obtaining module 402 is configured to obtain time series data, where the time series data includes a foreign exchange transaction data sequence of multiple business days in a historical time period, the foreign exchange transaction data sequence is arranged in chronological order, and each foreign exchange transaction service The data sequence includes multiple dimensions of foreign exchange transaction business data within the same business day;

A prediction module 404 is configured to input the time series data into a pre-trained multi-layer convolutional neural network model to predict a foreign exchange transaction volume in a future time period.

For details of the implementation process of the functions and functions of each module in the above device, refer to the implementation process of corresponding steps in the foregoing method, and details are not described herein again.

As for the device embodiment, since it basically corresponds to the method embodiment, the relevant part may refer to the description of the method embodiment. The device embodiments described above are only schematic, and the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, which may be located in One place, or can be distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution in this specification. Those of ordinary skill in the art can understand and implement without creative efforts.

The embodiments of the apparatus of the present specification can be applied to a computer device, such as a server or a terminal device. The device embodiments may be implemented by software, or by hardware or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, it is formed by reading the corresponding computer program instructions in the non-volatile memory into the memory and running the processor through the file processing processor. In terms of hardware, as shown in FIG. 5, it is a hardware structure diagram of the computer equipment where the device of this specification is located, except for the processor 502, the memory 504, the network interface 506, and the non-volatile memory 508 shown in FIG. 5. In addition, the server or electronic device where the device is located in the embodiment may generally include other hardware according to the actual function of the computer device, and details are not described herein again.

Correspondingly, the embodiment of the present specification also provides a computer storage medium. The storage medium stores a program, and when the program is executed by a processor, the method in any of the foregoing embodiments is implemented.

Accordingly, an embodiment of the present specification further provides a computer device including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, any one of the foregoing embodiments is implemented. Method.

The embodiments of the present specification may take the form of a computer program product implemented on one or more storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-usable storage media includes permanent and non-permanent, removable and non-removable media, and information can be stored by any method or technology. Information may be computer-readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, read-only disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, Magnetic tape cartridges, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media may be used to store information that can be accessed by computing devices.

Those skilled in the art will readily think of other embodiments of the present disclosure after considering the specification and practicing the specification disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include the common general knowledge or conventional technical means in the technical field not disclosed by this disclosure. . It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It should be understood that the present disclosure is not limited to the precise structure that has been described above and illustrated in the drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the disclosure is limited only by the following claims.

The above are merely preferred embodiments of the present disclosure, and are not intended to limit the present disclosure. Any modification, equivalent replacement, or improvement made within the spirit and principles of the present disclosure shall be included in the present disclosure. Within the scope of protection.

Claims

A method for predicting foreign exchange transaction volume, the method includes:

Obtaining time series data, the time series data includes a foreign exchange transaction data series of multiple business days in a historical time period, the foreign exchange transaction business data series are arranged in chronological order, and each foreign exchange transaction business data series includes the same business day Multi-dimensional foreign exchange transaction business data;

The time series data is input into a pre-trained multi-layer convolutional neural network model to predict the foreign exchange transaction volume in the future time period.
The method according to claim 1, further comprising:

Calculating prediction accuracy according to the predicted value and real value of the foreign exchange transaction volume in the future time period;

Adjusting model parameters of the multi-layer convolutional neural network model according to the prediction accuracy.
The method according to claim 1, further comprising:

Obtain the historical foreign exchange transaction volume for several periods in history;

Calculate the mean and variance of the foreign exchange transaction volume over the same period;

Anomaly detection is performed on the predicted value of the foreign exchange transaction volume according to the mean and variance.
The method according to claim 3, the step of performing anomaly detection on the predicted value of the foreign exchange transaction volume based on the mean and variance comprises:

If the predicted value of the foreign exchange transaction volume exceeds a preset numerical range, determining that the predicted value of the foreign exchange transaction volume is abnormal;

The value range is:

(X1, X2),

Among them, X1 = μ-N * δ, X2 = μ + N * δ;

In the formula, μ is the average value of the foreign exchange transaction volume in the same period, N is a preset positive integer, and δ is the variance of the foreign exchange transaction volume in the same period.
The method according to claim 4, further comprising:

If the predicted value of the foreign exchange transaction volume is less than X1, correct the predicted value of the foreign exchange transaction volume to X1;

If the predicted value of the foreign exchange transaction volume is greater than X2, modify the predicted value of the foreign exchange transaction volume to X2.
The method according to claim 1, wherein the step of inputting the time series data into a pre-trained multilayer convolutional neural network model comprises:

Calculating foreign exchange transaction characteristics of key merchants according to the time series data;

The foreign exchange transaction business features of the key merchants are input into a pre-trained multilayer convolutional neural network model.
The method according to any one of claims 1 to 6, wherein the foreign exchange transaction business data includes forward transaction volume, reverse transaction volume, business day date attribute, business day promotion attribute, and / or business day forecast for each business day. Set marketing amount.
A foreign exchange transaction volume prediction device, the device includes:

An obtaining module, configured to obtain time series data, where the time series data includes a foreign exchange transaction data sequence of multiple business days in a historical time period, the foreign exchange transaction data sequence is arranged in chronological order, and each foreign exchange transaction data The sequence includes multiple dimensions of foreign exchange transaction business data within the same business day;

A prediction module is configured to input the time series data into a pre-trained multi-layer convolutional neural network model to predict a foreign exchange transaction volume in a future time period.
A computer-readable storage medium having stored thereon a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 7.
A computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.