CN115049137A

CN115049137A - Prediction method and device of transaction yield, storage medium and electronic equipment

Info

Publication number: CN115049137A
Application number: CN202210719471.2A
Authority: CN
Inventors: 皇甫晓洁; 张倩妮; 高睿; 周魁
Original assignee: Industrial and Commercial Bank of China Ltd ICBC
Current assignee: Industrial and Commercial Bank of China Ltd ICBC
Priority date: 2022-06-23
Filing date: 2022-06-23
Publication date: 2022-09-13

Abstract

The invention discloses a method and a device for predicting transaction yield, a storage medium and electronic equipment. Wherein, the method comprises the following steps: acquiring historical transaction data of a target product, and determining historical trend data according to historical characteristic data and a target period in the historical transaction data; dividing historical trend data into a plurality of initial training samples through a sliding window, wherein the initial characteristic samples comprise at least one characteristic factor, and the initial prediction samples comprise at least one prediction factor; determining a model training sample by combining the characteristic factor data corresponding to the characteristic factor and the prediction factor data corresponding to the prediction factor; and inputting the real-time transaction characteristic data into the transaction prediction model to obtain a transaction yield prediction result of the target product in the next time period. The invention solves the technical problem that accurate and objective market information is difficult to analyze and obtain by adopting a manual decision or semi-automatic mode in the related technology.

Description

Prediction method and device of transaction yield, storage medium and electronic equipment

Technical Field

The invention relates to the field of financial science and technology, in particular to a method and a device for predicting transaction yield, a storage medium and electronic equipment.

Background

At present, the automatic transaction timeliness of transaction products such as funds, precious metals and the like reaches ten milliseconds to hundred milliseconds, and the automatic market transaction gradually replaces manual transaction. The marketing strategy is generally to hang a marketing order, aiming at providing market liquidity and profitability. However, with the development of precious metal business, a simple marketing strategy cannot meet business requirements, and a precious metal profit strategy for acquiring information from various channels, analyzing market trend and guiding a hang-and-withdraw operation is generated.

The profitability strategy in the related art is generally to hang a token sheet with the goal of consuming market liquidity and profitability. The precious metal profit strategy needs to analyze market conditions changing in real time, pre-judge market conditions in a future period of time, and grasp profit opportunities to carry out hang-and-withdraw-list operations. The forecasting of future market quotations needs to consider a plurality of factors such as price, quantity, change rate and the like of the market quotations at the same time, but most of the existing manual decision-making and semi-automatic execution modes are difficult to make timely and objective prejudgment and operation.

In view of the above problems, no effective solution has been proposed.

Disclosure of Invention

The embodiment of the invention provides a method and a device for predicting transaction yield, a storage medium and electronic equipment, which are used for at least solving the technical problem that accurate and objective market information is difficult to analyze and obtain by adopting a manual decision or semi-automatic mode in the related technology.

According to an aspect of an embodiment of the present invention, there is provided a method for predicting a transaction yield, including: acquiring historical transaction data of a target product, and determining historical trend data according to historical characteristic data and a target period in the historical transaction data; dividing the historical trend data into a plurality of initial training samples through a sliding window, wherein each initial training sample at least comprises: an initial feature sample and an initial prediction sample, the initial feature sample comprising at least one feature factor, the initial prediction sample comprising at least one predictor; determining a model training sample by combining the characteristic factor data corresponding to the characteristic factor and the prediction factor data corresponding to the prediction factor, wherein the model training sample is used for training to obtain a transaction prediction model; and inputting the real-time transaction characteristic data into the transaction prediction model to obtain a transaction yield prediction result of the target product in the next time period.

Optionally, the step of dividing the historical trend data into a plurality of initial training samples through a sliding window includes: acquiring a preset sliding segmentation time period, wherein the sliding segmentation time period is used for determining a tendency segmentation sample; determining a plurality of trend segmentation samples in a history interval in the history trend data as a sliding window; sliding the sliding window in the historical trend data to determine a plurality of sample intervals to be analyzed; dividing each sample interval to be analyzed into a characteristic interval and a prediction interval; obtaining an interval sample in the characteristic interval to obtain the initial characteristic sample, and obtaining an interval sample in the prediction interval to obtain the initial prediction sample; and determining the plurality of initial training samples according to the initial characteristic samples and the initial prediction samples in each sample interval to be analyzed.

Optionally, the step of determining a model training sample by combining feature factor data corresponding to the feature factor and predictor data corresponding to the predictor includes: acquiring a feature calculation strategy corresponding to each feature factor, and calculating feature factor data corresponding to the feature factors respectively by adopting the feature calculation strategies to obtain a feature data set; acquiring a factor calculation strategy corresponding to each prediction factor, and calculating prediction factor data corresponding to the prediction factors respectively by adopting the factor calculation strategies to obtain a prediction data set; calculating the yield data of each unit interval in the model training sample by adopting a preset yield calculation strategy, and determining the prediction yield, wherein the prediction interval in the model training sample comprises a plurality of unit intervals; and determining the model training sample by combining the characteristic data set, the prediction data set and the yield data.

Optionally, the step of calculating the yield data of each unit interval in the model training sample by using a preset yield calculation strategy to determine the predicted yield includes: calculating sample data of each unit interval in the prediction interval, and determining a plurality of yield rate data of the prediction interval; calculating a yield difference value between each yield data and the last yield data in the prediction interval through a time sequence label strategy; accumulating all the yield difference values in the prediction interval to determine a yield area; and comparing the area of the yield rate with a preset area threshold value to determine the predicted yield rate.

Optionally, before determining the model training sample by combining the feature factor data corresponding to the feature factor and the predictor data corresponding to the predictor, the method further includes: calculating the characteristic change rate of each characteristic data through a preset characteristic calculation strategy; calculating the data correlation between every two feature data through a preset correlation calculation strategy; and deleting the characteristic data with the characteristic change rate smaller than a preset change rate threshold value, and/or deleting the characteristic data with the data correlation smaller than a correlation threshold value.

Optionally, the step of obtaining historical transaction data of the target product, and determining historical trend data according to historical feature data in the historical transaction data and the target period includes: carrying out hierarchical quantitative processing on the historical transaction data to determine first index data; and analyzing the first index data according to the target period, and determining the historical trend data.

Optionally, before inputting the real-time transaction characteristic data into the transaction prediction model to obtain a prediction result of the transaction yield of the target product in the next time period, the method further includes: establishing an incidence relation between the transaction prediction model and a quantitative transaction platform, wherein the quantitative transaction platform provides real-time transaction data for the transaction prediction model; and extracting real-time transaction characteristic data in the real-time transaction data.

Optionally, the target product is a precious metal trading product.

According to another aspect of the embodiments of the present invention, there is also provided a device for predicting a transaction yield, including: the first determining unit is used for acquiring historical transaction data of a target product and determining historical trend data according to historical characteristic data and a target period in the historical transaction data; a first processing unit, configured to divide the historical trend data into a plurality of initial training samples through a sliding window, where each of the initial training samples at least includes: an initial feature sample and an initial prediction sample, the initial feature sample comprising at least one feature factor, the initial prediction sample comprising at least one predictor; a second determining unit, configured to determine a model training sample by combining feature factor data corresponding to the feature factor and predictor data corresponding to the predictor, where the model training sample is used for training to obtain a transaction prediction model; and the second processing unit is used for inputting the real-time transaction characteristic data into the transaction prediction model so as to obtain a transaction yield prediction result of the target product in the next time period.

Optionally, the first processing unit comprises: the device comprises a first acquisition subunit, a second acquisition subunit and a third acquisition subunit, wherein the first acquisition subunit is used for acquiring a preset sliding segmentation time period, and the sliding segmentation time period is used for determining tendency segmentation samples; the first determining subunit is used for determining a plurality of trend segmentation samples in a history interval in the history trend data as a sliding window; the second determining subunit is used for sliding the sliding window in the historical trend data to determine a plurality of sample intervals to be analyzed; the dividing subunit is used for dividing each sample interval to be analyzed into a characteristic interval and a prediction interval; the first processing subunit is configured to obtain an interval sample in the feature interval to obtain the initial feature sample, and obtain an interval sample in the prediction interval to obtain the initial prediction sample; and the third determining subunit is configured to determine the multiple initial training samples according to the initial feature sample and the initial prediction sample in each sample interval to be analyzed.

Optionally, the second determining unit includes: the second processing subunit is used for acquiring a feature calculation strategy corresponding to each feature factor, and calculating feature factor data corresponding to the feature factors respectively by adopting the feature calculation strategies to obtain a feature data set; the second obtaining subunit is configured to obtain a factor calculation policy corresponding to each of the prediction factors, and calculate prediction factor data corresponding to the prediction factors by using the factor calculation policies, to obtain a prediction data set; the fourth determining subunit is configured to calculate, by using a preset profitability calculation strategy, profitability data of each unit interval in the model training sample, and determine a predicted profitability, where a prediction interval in the model training sample includes a plurality of unit intervals; and the fifth determining subunit is used for determining the model training sample by combining the characteristic data set, the prediction data set and the yield data.

Optionally, the fourth determining subunit includes: the first determining module is used for calculating sample data of each unit interval in the prediction interval and determining a plurality of yield rate data of the prediction interval; the first calculation module is used for calculating a yield difference value between each yield data and the last yield data in the prediction interval through a time sequence label strategy; a second determining module, configured to accumulate all the yield difference values in the prediction interval to determine a yield area; and the third determining module is used for comparing the area of the yield rate with a preset area threshold value and determining the predicted yield rate.

Optionally, the prediction device of the transaction yield further comprises: the first calculation unit is used for calculating the characteristic change rate of each characteristic data through a preset characteristic calculation strategy before determining a model training sample by combining the characteristic factor data corresponding to the characteristic factors and the prediction factor data corresponding to the prediction factors; the second calculation unit is used for calculating the data correlation between every two feature data through a preset correlation calculation strategy; and the deleting unit is used for deleting the characteristic data of which the characteristic change rate is smaller than a preset change rate threshold value and/or deleting the characteristic data of which the data correlation is smaller than a correlation threshold value.

Optionally, the first determination unit includes: the sixth determining subunit is used for performing hierarchical quantization processing on the historical transaction data to determine first index data; and the seventh determining subunit is configured to analyze the first index data according to the target period, and determine the historical trend data.

Optionally, the prediction device of the transaction yield further comprises: the relationship establishing unit is used for establishing an incidence relationship between the transaction prediction model and a quantitative transaction platform before inputting the real-time transaction characteristic data into the transaction prediction model to obtain a transaction yield prediction result of the target product in the next time period, wherein the quantitative transaction platform provides real-time transaction data for the transaction prediction model; and the extraction unit is used for extracting the real-time transaction characteristic data in the real-time transaction data.

Optionally, the target product is a precious metal trading product.

In the invention, historical transaction data of a target product is obtained, historical trend data is determined according to historical characteristic data and a target period in the historical transaction data, and then the historical trend data is divided into a plurality of initial training samples through a sliding window, wherein each initial training sample at least comprises: the method comprises the steps of obtaining an initial characteristic sample and an initial prediction sample, wherein the initial characteristic sample comprises at least one characteristic factor, the initial prediction sample comprises at least one prediction factor, then determining a model training sample by combining characteristic factor data corresponding to the characteristic factor and prediction factor data corresponding to the prediction factor, the model training sample is used for training to obtain a transaction prediction model, and finally inputting real-time transaction characteristic data into the transaction prediction model to obtain a transaction yield prediction result of the target product in the next time period. In the invention, an initial training sample is determined through a sliding window, a model training sample is obtained by calculating characteristic factor data and prediction factor data in the initial training sample, and finally, the transaction yield is predicted according to a transaction prediction model obtained by the model training sample, so that the purpose of more accurately predicting the income market quotation is achieved, and the technical problem that accurate and objective market quotation information is difficult to analyze and obtain by adopting a manual decision or semi-automatic mode in the related technology is solved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:

FIG. 1 is a flow diagram of an alternative prediction of transaction profitability in accordance with an embodiment of the present invention;

FIG. 2 is a flow diagram of an alternative predicted transaction rate of return according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of an alternative decision process for xgboost in accordance with an embodiment of the present invention;

FIG. 4 is a schematic diagram of an alternative transaction rate of return prediction arrangement in accordance with an embodiment of the present invention;

fig. 5 is a block diagram of a hardware architecture of an electronic device (or mobile device) for predicting transaction profitability according to an embodiment of the present invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

In order to facilitate the understanding of the present invention for those skilled in the art, some terms or terms related to the embodiments of the present application are described below:

a marking sheet: refers to an order that is less expensive than the market price.

A token sheet: refers to an order whose price intersects with market conditions. For example, the price of the sales order in the market is 1.0, and the purchase order with the price more than or equal to 1.0 is the ticket order.

XGboost: the method is a machine learning function library which is focused on a gradient lifting algorithm.

It should be noted that relevant information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for presentation, analyzed data, etc.) referred to in the present disclosure are information and data that are authorized by the user or sufficiently authorized by various parties. For example, an interface is provided between the system and the relevant user or organization, before obtaining the relevant information, an obtaining request needs to be sent to the user or organization through the interface, and after receiving the consent information fed back by the user or organization, the relevant information is obtained.

The method can be applied to various software products, systems and client (including but not limited to mobile clients and PC) systems of financial institutions, and the method for predicting the transaction yield can predict various transaction platforms (including but not limited to precious metals, funds, securities, financing and the like) to realize the prediction of the transaction quotation of the target time period.

According to the invention, the precious metal hierarchical quantitative market data is analyzed from multiple dimensions and multiple levels, the key data indexes are subjected to characteristic processing, the data are labeled by a time sequence labeling algorithm, the time sequence market data can be predicted, a reliable behavior basis is provided for a trading profit strategy, and the profit capacity is increased. Better providing financial product analysis to the user.

The invention will now be illustrated with reference to the following examples.

Example one

In accordance with an embodiment of the present invention, there is provided an alternative method embodiment of predicting transaction profitability, it being noted that the steps illustrated in the flowchart of the figures may be performed in a computer system such as a set of computer-executable instructions and that although a logical ordering is illustrated in the flowchart, in some cases the steps illustrated or described may be performed in an order different than that illustrated herein.

FIG. 1 is a flow chart of an alternative method of trading profitability in accordance with an embodiment of the present invention, as shown in FIG. 1, comprising the steps of:

step S101, acquiring historical transaction data of a target product, and determining historical trend data according to historical characteristic data and a target period in the historical transaction data;

step S102, dividing the historical trend data into a plurality of initial training samples through a sliding window, wherein each initial training sample at least comprises: the method comprises the steps of obtaining an initial characteristic sample and an initial prediction sample, wherein the initial characteristic sample comprises at least one characteristic factor, and the initial prediction sample comprises at least one prediction factor;

step S103, determining a model training sample by combining the characteristic factor data corresponding to the characteristic factor and the prediction factor data corresponding to the prediction factor, wherein the model training sample is used for training to obtain a transaction prediction model;

and step S104, inputting the real-time transaction characteristic data into a transaction prediction model to obtain a transaction yield prediction result of the target product in the next time period.

Through the steps, historical transaction data of a target product are obtained, historical trend data are determined according to historical feature data and a target period in the historical transaction data, and then the historical trend data are divided into a plurality of initial training samples through a sliding window, wherein each initial training sample at least comprises: the method comprises the steps of obtaining an initial characteristic sample and an initial prediction sample, wherein the initial characteristic sample comprises at least one characteristic factor, the initial prediction sample comprises at least one prediction factor, then, combining characteristic factor data corresponding to the characteristic factor and prediction factor data corresponding to the prediction factor to determine a model training sample, the model training sample is used for training to obtain a transaction prediction model, and finally, real-time transaction characteristic data is input into the transaction prediction model to obtain a transaction yield prediction result of the target product in the next time period. In the embodiment, an initial training sample is determined through a sliding window, a model training sample is obtained by calculating characteristic factor data and prediction factor data in the initial training sample, and finally, a transaction prediction model obtained according to the model training sample is used for predicting the transaction profitability, so that the purpose of more accurately predicting the profitability is achieved, and the technical problem that accurate and objective market information is difficult to analyze and obtain by adopting a manual decision or semi-automatic mode in the related technology is solved.

The following further describes embodiments of the present invention in conjunction with the above-described implementation steps.

Step S101, obtaining historical transaction data of a target product, and determining historical trend data according to historical characteristic data and a target period in the historical transaction data.

It should be noted that, the present embodiment does not limit the specific target product, and the target product may be a bulk commodity (such as petroleum, cotton, etc.), or may also be a precious metal, or a fund, and the precious metal trading product is schematically illustrated in the present embodiment.

Optionally, the step of obtaining historical transaction data of the target product, and determining historical trend data according to historical feature data and a target period in the historical transaction data includes: carrying out hierarchical quantitative processing on historical transaction data to determine first index data; and analyzing the first index data according to the target period, and determining historical trend data.

The historical transaction data that can be adopted in this embodiment is a transaction quotation list of the target product in a certain historical time period, and the database can be called by the quantitative analysis platform, for example, the quotation list of the target product is quantitatively analyzed by calling the Spark (which is an API provided by the Spark for a Python developer) to access the target database, and the acquisition of the existing host lake entering data is realized by the autonomous programming. The table may store product market information for each minimum fluctuation range, including but not limited to: market time (accurate to millisecond), channel, contract, latest bargaining price, high and low opening price, market position, latest bargaining market quantity, price, direction and other information.

The method comprises the steps of determining first index data by carrying out hierarchical quantitative analysis on a transaction quotation list, analyzing the first index data according to a target period of a product, and determining historical trend data of the product in a certain historical time period.

Step S102, dividing the historical trend data into a plurality of initial training samples through a sliding window, wherein each initial training sample at least comprises: the method comprises the steps of obtaining initial characteristic samples and initial prediction samples, wherein the initial characteristic samples comprise at least one characteristic factor, and the initial prediction samples comprise at least one prediction factor.

In an alternative embodiment, the step of dividing the historical trend data into a plurality of initial training samples through a sliding window includes: acquiring a preset sliding segmentation time period, wherein the sliding segmentation time period is used for determining a tendency segmentation sample; determining a plurality of trend segmentation samples in a history interval in the history trend data as a sliding window; sliding the sliding window in the historical trend data to determine a plurality of sample intervals to be analyzed; dividing each sample interval to be analyzed into a characteristic interval and a prediction interval; obtaining an interval sample in a characteristic interval to obtain an initial characteristic sample, and obtaining an interval sample in a prediction interval to obtain an initial prediction sample; and determining a plurality of initial training samples according to the initial characteristic samples and the initial prediction samples in each sample interval to be analyzed.

Optionally, the preset sliding partition time period is set by each user or a system management user, which is not specifically limited in this embodiment.

In the embodiment of the present invention, the training samples may be extracted by setting a sliding window, where the extracting manner of the sliding window is set by itself, for example, a market trend within a specified time period (e.g., 1 hour, one day, one week, etc.) is taken as a k-line trend (simply referred to as k), an interval of n k is determined to be referred to as a sliding window, n k within an interval from i to n + i is taken as a sample, an interval from i to n + i is taken as a window, the window is slid by s (where s may set a size value by itself) for market, and a next sample interval is taken as an interval from i + s to n + i + s. In a sample interval, the sample interval is divided into a characteristic interval and a prediction interval, a sample in the characteristic interval is an initial characteristic sample, a sample in the prediction interval is an initial prediction sample, and a plurality of initial training samples are determined through the initial characteristic sample and the initial prediction sample in each sample interval to be analyzed.

An optional embodiment, before determining the model training sample by combining the feature factor data corresponding to the feature factor and the predictor data corresponding to the predictor, further includes: calculating the characteristic change rate of each characteristic data through a preset characteristic calculation strategy; calculating the data correlation between every two characteristic data through a preset correlation calculation strategy; and deleting the characteristic data with the characteristic change rate smaller than a preset change rate threshold value, and/or deleting the characteristic data with the data correlation smaller than a correlation threshold value.

In the calculated characteristic interval, some characteristics in different characteristic factor data are not changed greatly, so that change parameters can be obtained through a characteristic calculation strategy, and deletion, adjustment, updating and white filling processing are carried out; but there is stronger correlation between some characteristics, can only keep one; the relevance of some characteristics and the market trend target is not strong, and the characteristics can be deleted to improve the data processing efficiency.

And S103, determining a model training sample by combining the characteristic factor data corresponding to the characteristic factor and the prediction factor data corresponding to the prediction factor, wherein the model training sample is used for training to obtain a transaction prediction model.

Optionally, the step of determining the model training sample by combining the feature factor data corresponding to the feature factor and the predictor data corresponding to the predictor includes: acquiring a characteristic calculation strategy corresponding to each characteristic factor, and calculating characteristic factor data corresponding to the characteristic factors respectively by adopting the characteristic calculation strategy to obtain a characteristic data set; acquiring a factor calculation strategy corresponding to each prediction factor, and calculating prediction factor data corresponding to the prediction factors respectively by adopting the factor calculation strategy to obtain a prediction data set; calculating the yield data of each unit interval in the model training sample by adopting a preset yield calculation strategy, and determining the prediction yield, wherein the prediction interval in the model training sample comprises a plurality of unit intervals; and determining a model training sample by combining the characteristic data set, the prediction data set and the yield data.

The characteristic factor may refer to the associated characteristic of the product, and may include, but is not limited to: the bargain price, the bargain amount, the position taken amount, the weighted optimal buying price, the optimal buying amount, the weighted optimal selling price, the optimal selling amount, the bargain price change, the bargain amount change, the position taken amount change and the like, and the calculation strategy corresponding to each characteristic factor can be various calculation formulas, such as: the calculation formula of the bargaining price is as follows: the final bargain price of k and the bargain amount are calculated by the following formula: sum of the volume of all k within the characteristic interval.

And acquiring a factor calculation strategy corresponding to each prediction factor, and calculating prediction factor data corresponding to the prediction factors respectively by adopting the factor calculation strategy to obtain a prediction data set.

And calculating the yield data of each unit interval in the model training sample by adopting a preset yield calculation strategy, and determining the predicted yield through mathematical operation according to the yield data of each unit interval.

The manner in which the predicted profitability is determined is described in detail below.

Optionally, the step of calculating the yield data of each unit interval in the model training sample by using a preset yield calculation strategy to determine the predicted yield includes: calculating sample data of each unit interval in the prediction interval, and determining a plurality of yield data of the prediction interval; calculating a yield difference value between each yield data and the last yield data in the prediction interval through a time sequence label strategy; accumulating all yield difference values in the prediction interval to determine a yield area; and comparing the area of the yield rate with a preset area threshold value to determine the predicted yield rate.

It should be noted that the calculation method of the label greatly affects the model prediction effect. Dividing the sample into a characteristic interval and a prediction interval, calculating interval characteristics for training in the characteristic interval, and marking a label in the prediction interval.

In this embodiment, a time-series label algorithm is proposed, in which the area of the yield is sum (yi-ye) for each yield yi in the prediction interval, with the last yield ye in the prediction interval as a reference. The calculated area distribution of the profitability can be classified into falling, leveling and rising 3 categories according to the comparison result by comparing the area of the profitability with a preset area threshold.

Wherein, the profitability y is (calculation section ending time closing price-calculation section starting time closing price)/calculation section starting time closing price, wherein, the calculation section ending time closing price is the closing price of the ending time of one unit section in the prediction section; and calculating the closing price at the starting time of the interval, namely the closing price at the ending time of the unit interval in the predicted interval.

An optional embodiment, before inputting the real-time transaction characteristic data into the transaction prediction model to obtain a prediction result of the transaction yield of the target product in the next time period, the method further comprises: establishing an incidence relation between a transaction prediction model and a quantitative transaction platform, wherein the quantitative transaction platform provides real-time transaction data for the transaction prediction model; and extracting real-time transaction characteristic data in the real-time transaction data.

And deploying the trained transaction prediction model on a quantitative transaction platform of a transaction system, issuing the trained transaction prediction model as a sub-strategy, and calling other strategies.

It should be noted that, when the transaction prediction model is deployed, the real-time market of the target product can be connected upstream, characteristic processing is performed before market data enters the model, the model output is market trend in a future period of time, and the output can be used as a behavior reference basis of a profit strategy of the target product.

The transaction prediction model can be obtained by training a model through a model training sample, for example; and taking the characteristic information of the characteristic interval of the sample as input, taking the predicted yield data as a label, and performing supervised learning by adopting an xgboost method to obtain a transaction rate prediction model of the target product.

Optionally, the step of inputting the real-time transaction characteristic data into the transaction prediction model to obtain a prediction result of the transaction yield of the target product in the next time period includes: integrating a plurality of discrete decision tree classifiers into a target classifier by adopting a pre-configured distributed gradient enhancement algorithm; inputting the real-time transaction characteristic data into a target classifier for supervised learning; and accumulating the real-time transaction characteristic data of each decision tree after supervised learning to obtain a transaction yield prediction result.

The pre-configured distributed gradient enhancement algorithm can be an xgboost algorithm, the algorithm can integrate a plurality of weak decision tree classifiers together to form a strong classifier, a tree is continuously grown through continuous tree addition and feature splitting, namely a new function is continuously learned, so that a predicted error is fitted, and supervised learning can be performed by adopting a support vector machine (svm) and a logistic regression algorithm.

In this embodiment, the processed market characteristic information is used as the input of the strong classifier, and the scores corresponding to each decision tree are accumulated to obtain the predicted probability value of the order state through continuous supervised learning.

The embodiment can input the real-time transaction characteristic data extracted from the transaction data in the transaction platform into the transaction prediction model to obtain the prediction result of the transaction yield of the target product in the next time period, namely, the transaction trend condition in the next time period can be obtained.

According to the method, market data of the target product are analyzed from multiple dimensions and multiple levels, characteristic processing is carried out on key data indexes, the data are labeled by a time sequence labeling algorithm, time sequence market data can be predicted, a reliable behavior basis is provided for a transaction profit strategy, profit capacity is improved, and accordingly financial product analysis is better provided for users.

The invention will now be described with reference to another embodiment.

Example two

The embodiment provides another method for predicting the profitability of a financial product, which schematically illustrates a detailed technical scheme by using a precious metal as a target product, and schematically illustrates the embodiment by combining an xgboost algorithm and a precious metal profitability prediction model, wherein fig. 2 is a flow chart of another optional method for predicting the transaction profitability according to the embodiment of the invention, and the main flow of the embodiment is as shown in fig. 2:

(1) data acquisition: firstly, collecting history data of noble metal market quotations;

(2) and (5) factor processing. Processing original data into factors, and performing characteristic engineering processing on the factors to form a training sample;

(3) label processing, namely endowing a training label to the sample according to the market information rising and falling condition, and providing a time sequence label algorithm to mark the sample;

(4) and (4) feature screening. Some characteristics are not changed greatly and can be deleted; some characteristics have stronger correlation, and only one characteristic can be reserved; some characteristics are not strongly correlated with the market trend target and can be deleted.

(5) And obtaining a test sample and a training sample after characteristic screening. The training sample is used for training the xgboost model, and the testing sample is used for testing whether the trained model is converged.

(6) In the training process by adopting the training sample, the xgboost model is used for training, which comprises the following steps: optimizing parameters, obtaining a model, and then training a final sample set by adopting an xgboost method to obtain a noble metal market situation classification model;

(7) and predicting the market trend. And finally, deploying the model in a real-time trading platform, and inputting the real-time market of the precious metal into the model to obtain the predicted market trend so as to provide reliable behavior basis for the precious metal profit strategy.

The respective steps will be described in detail below.

1) And (6) data acquisition.

The data adopted by the invention is a precious metal market statement list, a large database can be accessed by calling spark through the pyspark by using a quantitative analysis platform of financial transaction, the quantitative analysis of the precious metal market statement API is realized by autonomous programming, and the acquisition of the existing host lake entering data is realized. The table stores the precious metal quotation of each tick (the minimum fluctuation point or fluctuation interval of the quotation), including the information of the quotation time (accurate to millisecond), channel, contract, latest bargain price, high and low price, full market position, latest bargain quotation amount, price, direction and the like. The embodiment predicts the market trend of a short time in the future according to the market change of a period of time in the past. Therefore, the hierarchical quantitative quotation of the precious metal can be taken as basic data.

2) And (5) factor processing.

First, the market is processed into K-line data, in this embodiment, K minutes, for example, one K-line interval data per minute. The specific processing logic of the index in each period is as shown in table 1:

TABLE 1

In order to increase the number of samples, the market conditions overlap when calculating k-line data.

Then, according to the processed k line, the characteristic factor is processed. The concept of sliding window is introduced here, n k in the interval from the i-th to the n + i is taken as a sample and also becomes a window, the window is slid by s quotations, and the next sample interval is the interval from the i + s to the n + i + s. In a sample interval, dividing a sample into a characteristic interval and a prediction interval, calculating interval characteristics for training in the characteristic interval, and labeling according to a time sequence labeling algorithm in the prediction interval. The interval feature calculation method is shown in table 2:

TABLE 2

3) And (5) processing the label.

The calculation mode of the label can greatly influence the prediction effect of the model. Dividing the sample into a characteristic interval and a prediction interval, calculating interval characteristics for training in the characteristic interval, and marking a label in the prediction interval. The embodiment provides a time sequence label algorithm. That is, the yield area is sum (yi-ye) for each yield yi in the prediction section with reference to the last yield ye in the prediction section. The calculated area distribution of the profitability can be classified into falling, leveling and rising 3 categories according to the comparison result by comparing the area of the profitability with a preset threshold. The profit rate y is (the calculation section ending time closing price-the calculation section starting time closing price)/the calculation section starting time closing price.

4) And (4) feature screening.

The calculated interval characteristics (corresponding to the characteristic factor data) have small change per se and can be deleted; some characteristics have stronger correlation, and only one characteristic can be reserved; some characteristics are not strongly correlated with the market trend target and can be deleted.

5) And (5) constructing a model.

And taking the characteristic information of the sample as input, taking the market fluctuation situation as a label, and performing supervised learning by adopting an xgboost method to obtain a precious metal market trend prediction model. The xgboost algorithm integrates a plurality of weak decision tree classifiers together to form a strong classifier, and a tree is continuously grown through continuous tree addition and characteristic splitting, namely a new function is continuously learned, so that the predicted error is fitted.

Fig. 3 is a schematic diagram of an alternative decision process of xgboost according to an embodiment of the present invention, and as shown in fig. 3, the decision process of xgboost is schematically shown.

As shown in fig. 3, in the figure, blp represents the optimal purchase price, qty _ sum represents the volume of delivery, position represents the capacity of taken positions, std5 represents the standard deviation of the first five latest prices, close _ std5 represents the standard deviation of the first 5 closing prices, close _ std10 represents the standard deviation of the first 10 closing prices, volume _ std5 represents the standard deviation of the first 5 volume of delivery, alv _ weighted _ mean represents the weighted average optimal sale price, timetamp _ first represents the start time, a2to10v _ last represents the sum of the suboptimal sale quantities of all k in the feature interval, blv _ weighted _ mean represents the weighted optimal purchase quantity, alv _ weighted represents the weighted optimal sale quantity, and position _ diff10 represents the standard deviation of the first 10 head sizes.

(6) And (6) deploying the model.

And deploying the trained transaction prediction model on a quantitative transaction platform for transaction, issuing the trained transaction prediction model as a sub-strategy and calling other strategies. When the model is deployed, the upstream can be connected with the real-time quotation of the precious metal and the characteristic processing is carried out before the quotation enters the model. The output of the model is the market trend or the predicted transaction yield in a future period of time, and the output can be used as a behavior reference basis of the precious metal profit strategy.

According to the embodiment of the invention, the hierarchical quantitative market data of the precious metal is analyzed from multiple dimensions and multiple levels, the key data indexes are characterized, the data are labeled by a time sequence labeling algorithm, and an xgboost machine learning algorithm is used for constructing a market trend prediction model. The model can predict time sequence market data, provide reliable behavior basis for precious metal profit strategies and increase profit capacity. In addition, the yield prediction model of the time-series tag algorithm provided in the embodiment is extensible from the aspect of architecture and feature information, and can be extended to other financial transaction fields to better provide financial product analysis for users.

The invention is described below in connection with an alternative embodiment.

EXAMPLE III

The embodiment provides an alternative prediction device of transaction yield, and each implementation unit included in the prediction device corresponds to each implementation step in the embodiments.

Fig. 4 is a schematic diagram of an alternative transaction yield prediction apparatus according to an embodiment of the present invention, as shown in fig. 4, including: a first determination unit 41, a first processing unit 42, a second determination unit 43, a second processing unit 44.

The first determining unit 41 is configured to acquire historical transaction data of a target product, and determine historical trend data according to historical feature data in the historical transaction data and a target period;

a first processing unit 42, configured to divide the historical trend data into a plurality of initial training samples through a sliding window, where each of the initial training samples at least includes: the method comprises the steps of obtaining an initial characteristic sample and an initial prediction sample, wherein the initial characteristic sample comprises at least one characteristic factor, and the initial prediction sample comprises at least one prediction factor;

a second determining unit 43, configured to determine a model training sample by combining feature factor data corresponding to the feature factor and prediction factor data corresponding to the prediction factor, where the model training sample is used for training to obtain a transaction prediction model;

and the second processing unit 44 is configured to input the real-time transaction characteristic data into the transaction prediction model to obtain a prediction result of the transaction yield of the target product in the next time period.

The above prediction apparatus for transaction profitability may obtain historical transaction data of a target product through the first determining unit 41, determine historical trend data according to historical feature data and a target period in the historical transaction data, and divide the historical trend data into a plurality of initial training samples through the sliding window by the first processing unit 42, where each of the initial training samples at least includes: the method comprises the steps of obtaining an initial characteristic sample and an initial prediction sample, wherein the initial characteristic sample comprises at least one characteristic factor, the initial prediction sample comprises at least one prediction factor, a second determining unit 43 is used for combining characteristic factor data corresponding to the characteristic factor and prediction factor data corresponding to the prediction factor to determine a model training sample, the model training sample is used for training to obtain a transaction prediction model, and finally, a second processing unit 44 is used for inputting real-time transaction characteristic data into the transaction prediction model to obtain a transaction yield prediction result of a target product in the next time period. In the embodiment, an initial training sample is determined through a sliding window, a model training sample is obtained by calculating characteristic factor data and prediction factor data in the initial training sample, and finally, a transaction prediction model obtained according to the model training sample is used for predicting the transaction profitability, so that the purpose of improving the accuracy of revenue market prediction is achieved, and the technical problem that accurate and objective market information is difficult to analyze and obtain in the related art by adopting a manual decision or semi-automatic mode is solved.

Optionally, the first processing unit includes: the device comprises a first acquisition subunit, a second acquisition subunit and a third acquisition subunit, wherein the first acquisition subunit is used for acquiring a preset sliding segmentation time period, and the sliding segmentation time period is used for determining tendency segmentation samples; the first determining subunit is used for determining a plurality of trend segmentation samples in a history interval in the history trend data as a sliding window; the second determining subunit is used for sliding the sliding window in the historical trend data to determine a plurality of sample intervals to be analyzed; the dividing subunit is used for dividing each sample interval to be analyzed into a characteristic interval and a prediction interval; the first processing subunit is used for acquiring an interval sample in the characteristic interval to obtain an initial characteristic sample, and acquiring an interval sample in the prediction interval to obtain an initial prediction sample; and the third determining subunit is used for determining a plurality of initial training samples according to the initial characteristic samples and the initial prediction samples in each sample interval to be analyzed.

Optionally, the second determining unit includes: the second processing subunit is used for acquiring a feature calculation strategy corresponding to each feature factor, and calculating feature factor data corresponding to the feature factors respectively by adopting the feature calculation strategy to obtain a feature data set; the second obtaining subunit is used for obtaining factor calculation strategies corresponding to each prediction factor, and adopting the factor calculation strategies to calculate prediction factor data corresponding to the prediction factors respectively to obtain a prediction data set; the fourth determining subunit is configured to calculate, by using a preset profitability calculation strategy, profitability data of each unit interval in the model training sample, and determine a predicted profitability, where a prediction interval in the model training sample includes a plurality of unit intervals; and the fifth determining subunit is used for determining the model training sample by combining the characteristic data set, the prediction data set and the yield data.

Optionally, the fourth determining subunit includes: the first determining module is used for calculating sample data of each unit interval in the prediction interval and determining a plurality of yield rate data of the prediction interval; the first calculation module is used for calculating a yield difference value between each yield data and the last yield data in the prediction interval through a time sequence label strategy; the second determining module is used for accumulating all the yield difference values in the prediction interval and determining the yield area; and the third determining module is used for comparing the area of the yield rate with a preset area threshold value and determining the prediction yield rate.

Optionally, the device for predicting the transaction return rate further includes: the first calculation unit is used for calculating the characteristic change rate of each characteristic data through a preset characteristic calculation strategy before determining the model training sample by combining the characteristic factor data corresponding to the characteristic factors and the prediction factor data corresponding to the prediction factors; the second calculation unit is used for calculating the data correlation between every two characteristic data through a preset correlation calculation strategy; and the deleting unit is used for deleting the characteristic data of which the characteristic change rate is smaller than a preset change rate threshold value and/or deleting the characteristic data of which the data correlation is smaller than a correlation threshold value.

Optionally, the first determining unit includes: the sixth determining subunit is used for performing hierarchical quantitative processing on the historical transaction data to determine first index data; and the seventh determining subunit is used for analyzing the first index data according to the target period and determining the historical trend data.

Optionally, the device for predicting the transaction return rate further includes: the relationship establishing unit is used for establishing an incidence relationship between the transaction prediction model and a quantitative transaction platform before inputting the real-time transaction characteristic data into the transaction prediction model to obtain a transaction yield prediction result of the target product in the next time period, wherein the quantitative transaction platform provides real-time transaction data for the transaction prediction model; and the extraction unit is used for extracting the real-time transaction characteristic data in the real-time transaction data.

Optionally, the target product is a precious metal trading product.

The aforementioned prediction device of transaction yield may further include a processor and a memory, where the aforementioned first determining unit 41, the first processing unit 42, the second determining unit 43, the second processing unit 44, and the like are stored in the memory as program units, and the processor executes the aforementioned program units stored in the memory to implement corresponding functions.

The processor comprises a kernel, and the kernel calls a corresponding program unit from the memory. The kernel can be set to be one or more, and real-time transaction characteristic data is input into the transaction prediction model by adjusting kernel parameters so as to obtain a transaction yield prediction result of the target product in the next time period.

The memory may include volatile memory in a computer readable medium, Random Access Memory (RAM) and/or nonvolatile memory such as Read Only Memory (ROM) or flash memory (flash RAM), and the memory includes at least one memory chip.

According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including: a processor; and a memory for storing executable instructions for the processor; wherein the processor is configured to perform the method of predicting a transaction return rate of any of the above via execution of the executable instructions.

According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, which includes a stored computer program, wherein when the computer program runs, the apparatus on which the computer-readable storage medium is located is controlled to execute any one of the above methods for predicting a transaction yield.

Fig. 5 is a block diagram of a hardware architecture of an electronic device (or mobile device) for predicting transaction profitability according to an embodiment of the present invention. As shown in fig. 5, the electronic device may include one or more processors 502 (shown as 502a, 502b, … …, 502 n) (the processors 502 may include, but are not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), and memory 504 for storing data. Besides, the method can also comprise the following steps: a display, an input/output interface (I/O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of the I/O interface), a network interface, a keyboard, a power supply, and/or a camera. It will be understood by those skilled in the art that the structure shown in fig. 5 is only an illustration and is not intended to limit the structure of the electronic device. For example, the electronic device may also include more or fewer components than shown in FIG. 5, or have a different configuration than shown in FIG. 5.

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units may be a logical division, and in actual implementation, there may be another division, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed coupling or direct coupling or communication connection between each other may be an indirect coupling or communication connection through some interfaces, units or modules, and may be electrical or in other forms.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.

The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims

1. A method for predicting a transaction rate of return, comprising:

acquiring historical transaction data of a target product, and determining historical trend data according to historical characteristic data and a target period in the historical transaction data;

dividing the historical trend data into a plurality of initial training samples through a sliding window, wherein each initial training sample at least comprises: an initial feature sample and an initial prediction sample, the initial feature sample comprising at least one feature factor and the initial prediction sample comprising at least one predictor;

determining a model training sample by combining the characteristic factor data corresponding to the characteristic factor and the prediction factor data corresponding to the prediction factor, wherein the model training sample is used for training to obtain a transaction prediction model;

and inputting the real-time transaction characteristic data into the transaction prediction model to obtain a transaction yield prediction result of the target product in the next time period.

2. The prediction method according to claim 1, wherein the step of dividing the historical trend data into a plurality of initial training samples through a sliding window comprises:

acquiring a preset sliding segmentation time period, wherein the sliding segmentation time period is used for determining a tendency segmentation sample;

determining a plurality of trend segmentation samples in a history interval in the historical trend data as one sliding window;

sliding the sliding window in the historical trend data to determine a plurality of sample intervals to be analyzed;

dividing each sample interval to be analyzed into a characteristic interval and a prediction interval;

obtaining an interval sample in the characteristic interval to obtain the initial characteristic sample, and obtaining an interval sample in the prediction interval to obtain the initial prediction sample;

and determining a plurality of initial training samples according to the initial characteristic samples and the initial prediction samples in each sample interval to be analyzed.

3. The prediction method according to claim 2, wherein the step of determining model training samples by combining the feature factor data corresponding to the feature factor and the predictor data corresponding to the predictor comprises:

acquiring a feature calculation strategy corresponding to each feature factor, and calculating feature factor data corresponding to the feature factors respectively by adopting the feature calculation strategies to obtain a feature data set;

acquiring a factor calculation strategy corresponding to each prediction factor, and calculating prediction factor data corresponding to the prediction factors respectively by adopting the factor calculation strategies to obtain a prediction data set;

calculating the yield data of each unit interval in the model training sample by adopting a preset yield calculation strategy, and determining the prediction yield, wherein the prediction interval in the model training sample comprises a plurality of unit intervals;

and determining the model training sample by combining the characteristic data set, the prediction data set and the yield data.

4. The prediction method according to claim 3, wherein the step of calculating the yield data of each unit interval in the model training sample by using a preset yield calculation strategy to determine the predicted yield comprises:

calculating sample data of each unit interval in the prediction interval, and determining a plurality of yield rate data of the prediction interval;

calculating a yield difference value between each yield data and the last yield data in the prediction interval through a time sequence label strategy;

accumulating all the yield difference values in the prediction interval to determine a yield area;

and comparing the area of the yield rate with a preset area threshold value to determine the predicted yield rate.

5. The prediction method according to claim 2, wherein before determining the model training samples by combining the feature factor data corresponding to the feature factor and the predictor data corresponding to the predictor, the method further comprises:

calculating the characteristic change rate of each characteristic data through a preset characteristic calculation strategy;

calculating the data correlation between every two feature data through a preset correlation calculation strategy;

deleting the feature data with the feature change rate smaller than a preset change rate threshold value,

and/or the presence of a gas in the gas,

and deleting the characteristic data with the data correlation smaller than the correlation threshold.

6. The forecasting method of claim 1, wherein the step of obtaining historical transaction data of a target product and determining historical trend data according to historical feature data in the historical transaction data and a target period comprises:

carrying out hierarchical quantitative processing on the historical transaction data to determine first index data;

and analyzing the first index data according to the target period, and determining the historical trend data.

7. The forecasting method of claim 1, further comprising, before inputting real-time transaction characteristic data into the transaction forecasting model to obtain a prediction of a transaction rate of return for the target product over a next time period:

establishing an incidence relation between the transaction prediction model and a quantitative transaction platform, wherein the quantitative transaction platform provides real-time transaction data for the transaction prediction model;

and extracting real-time transaction characteristic data in the real-time transaction data.

8. The prediction method according to any one of claims 1 to 7, wherein the target product is a precious metal trading product.

9. An apparatus for predicting a transaction rate of return, comprising:

the first determining unit is used for acquiring historical transaction data of a target product and determining historical trend data according to historical characteristic data and a target period in the historical transaction data;

a first processing unit, configured to divide the historical trend data into a plurality of initial training samples through a sliding window, where each of the initial training samples at least includes: an initial feature sample and an initial prediction sample, the initial feature sample comprising at least one feature factor, the initial prediction sample comprising at least one predictor;

a second determining unit, configured to determine a model training sample by combining feature factor data corresponding to the feature factor and predictor data corresponding to the predictor, where the model training sample is used for training to obtain a transaction prediction model;

and the second processing unit is used for inputting the real-time transaction characteristic data into the transaction prediction model so as to obtain a transaction yield prediction result of the target product in the next time period.

10. A computer-readable storage medium, characterized by a computer program for storing, wherein when the computer program is executed, the computer-readable storage medium controls a device to execute the prediction method of transaction yield according to any one of claims 1 to 8.

11. An electronic device comprising one or more processors and memory storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of predicting transaction profitability of any one of claims 1-8.