CN112016840A - Method, device, equipment and storage medium for selecting index data - Google Patents

Method, device, equipment and storage medium for selecting index data Download PDF

Info

Publication number
CN112016840A
CN112016840A CN202010901797.8A CN202010901797A CN112016840A CN 112016840 A CN112016840 A CN 112016840A CN 202010901797 A CN202010901797 A CN 202010901797A CN 112016840 A CN112016840 A CN 112016840A
Authority
CN
China
Prior art keywords
data
market
index data
historical
index
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010901797.8A
Other languages
Chinese (zh)
Inventor
杨琳琳
陆彬
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Life Insurance Company of China Ltd
Original Assignee
Ping An Life Insurance Company of China Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Ping An Life Insurance Company of China Ltd filed Critical Ping An Life Insurance Company of China Ltd
Priority to CN202010901797.8A priority Critical patent/CN112016840A/en
Publication of CN112016840A publication Critical patent/CN112016840A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Human Resources & Organizations (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Development Economics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Computing Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Business, Economics & Management (AREA)
  • Biophysics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Molecular Biology (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • Educational Administration (AREA)
  • General Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Technology Law (AREA)
  • Evolutionary Biology (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Medical Informatics (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Bioinformatics & Computational Biology (AREA)

Abstract

The invention discloses a method, a device, equipment and a storage medium for selecting index data, wherein the method comprises the following steps: when a trigger instruction is received, acquiring market situation time sequence data of a target financial product in a first set time period from a preset financial platform, and extracting market situation characteristic vectors from the market situation time sequence data; acquiring historical performance data of each preset index data from a preset database, and extracting historical performance characteristic vectors from the historical performance data; inputting the market condition characteristic vector and the historical expression characteristic vector into a preset index evaluation model to obtain an estimation value of each index data; arranging all index data according to the estimated value from big to small, and sending the index data arranged before the set number to a designated terminal; the invention can select index data with higher reliability.

Description

Method, device, equipment and storage medium for selecting index data
Technical Field
The present invention relates to the field of data analysis technologies, and in particular, to a method, an apparatus, a device, and a storage medium for selecting index data.
Background
As machine learning algorithms continue to evolve, machine learning algorithms are increasingly being applied to various domains to address, for example: problems with data prediction, data classification, and data clustering; when a machine learning algorithm is used, a computer is used as an arithmetic device, so that a model is trained by the computer and a result is calculated; in order to obtain a model with higher reliability, a large amount of sample data is required to be used for model training; in addition, in order to obtain a more accurate result, a multi-dimensional feature vector needs to be input into the model. Therefore, when the computer is used as an arithmetic device to process more and more data, not only a large amount of arithmetic time is consumed, but also a large amount of CPU resources are occupied, thereby reducing the efficiency of training and using the model. In the field of financial investment, machine learning algorithm is also utilized to determine index data influencing investment strategy, and investment of financial products is carried out by referring to the index data. Since the financial data can change greatly in a short time, if the processing speed of the computer is too slow to determine the investment strategy in time, the income of the company can be directly influenced.
Disclosure of Invention
The invention aims to provide a method, a device, equipment and a storage medium for selecting index data, which not only can select the index data with higher reliability, but also can improve the efficiency of determining the index data.
According to an aspect of the present invention, there is provided a method of selecting index data, the method comprising:
when a trigger instruction is received, acquiring market situation time sequence data of a target financial product in a first set time period from a preset financial platform, and extracting market situation characteristic vectors from the market situation time sequence data;
acquiring historical performance data of each preset index data from a preset database, and extracting historical performance characteristic vectors from the historical performance data;
inputting the market condition characteristic vector and the historical expression characteristic vector into a preset index evaluation model to obtain an estimation value of each index data; the index evaluation model is obtained by training with the profitability of the target financial product in a second set time period as a target;
and arranging all index data according to the estimated values from large to small, and sending the index data arranged before the set number to the appointed terminal.
Optionally, the acquiring, from the preset financial platform, market data of the target financial product in a first set time period specifically includes:
acquiring daily market data of each transaction day of the target financial product before the current time point from a preset financial platform;
performing dimension increasing processing and standardization processing on the daily market data;
and forming the daily market data in the first set time period into market time sequence data.
Optionally, the extracting the market characteristic vector from the market time series data specifically includes:
determining a global feature vector irrelevant to time according to the market condition time sequence data by using a preset convolutional neural network model;
performing short-time Fourier transform on the quotation time sequence data to obtain a quotation time-frequency spectrogram, and performing local feature extraction on the quotation time-frequency spectrogram to obtain a local feature vector;
determining the importance of the local feature vector through the global feature vector by using an attention mechanism to obtain a time-frequency spectrogram local feature vector corresponding to the local feature vector;
and fusing the global feature vector and the time-frequency spectrogram local feature vector to obtain a market situation feature vector.
Optionally, the historical performance data of one index data includes: rolling cumulative profitability, maximum pullback, and sharp rate;
the extracting of the historical expression feature vector from the historical expression data specifically includes:
inputting the rolling accumulated yield, the maximum withdrawal and the sharp rate of index data into a preset multilayer neural network model to obtain historical performance characteristic data of the index data;
and forming historical performance characteristic data of all index data into a historical performance characteristic vector.
Optionally, the method further includes:
acquiring a training sample set; wherein the training sample set comprises a plurality of training samples, and each training sample comprises: historical market condition time series data in a first set time period before a time point, historical expression data of each index data before the time point, and historical market condition time series data in a second set time period after the time point;
aiming at a training sample, calculating an estimated value of each index data by using a factor evaluation model according to historical market time sequence data in the first set time period and historical performance data of each index data, and setting the index data with the maximum estimated value as target index data;
according to the target index data and the historical market time sequence data in the second set time period, calculating a real value corresponding to the target index data;
and calculating an error value of the estimated value and the true value of the target index data by using a preset loss function, and adjusting parameters of the index evaluation model according to the error value.
In order to achieve the above object, the present invention also provides an apparatus for selecting index data, the apparatus comprising:
the system comprises a first extraction module, a second extraction module and a third extraction module, wherein the first extraction module is used for acquiring market quotation time sequence data of a target financial product in a first set time period from a preset financial platform and extracting market quotation feature vectors from the market quotation time sequence data when a trigger instruction is received;
the second extraction module is used for acquiring historical expression data of each preset index data from a preset database and extracting historical expression characteristic vectors from the historical expression data;
the input module is used for inputting the market characteristic vector and the historical expression characteristic vector into a preset index evaluation model so as to obtain an estimation value of each index data; the index evaluation model is obtained by training with the profitability of the target financial product in a second set time period as a target;
and the sending module is used for arranging all the index data according to the estimated values from large to small and sending the index data arranged before the set number to the appointed terminal.
Optionally, the first extraction module is specifically configured to:
determining a global feature vector irrelevant to time according to the market condition time sequence data by using a preset convolutional neural network model;
performing short-time Fourier transform on the quotation time sequence data to obtain a quotation time-frequency spectrogram, and performing local feature extraction on the quotation time-frequency spectrogram to obtain a local feature vector;
determining the importance of the local feature vector through the global feature vector by using an attention mechanism to obtain a time-frequency spectrogram local feature vector corresponding to the local feature vector;
and fusing the global feature vector and the time-frequency spectrogram local feature vector to obtain a market situation feature vector.
Optionally, the apparatus further comprises:
the training module is used for acquiring a training sample set; wherein the training sample set comprises a plurality of training samples, and each training sample comprises: historical market condition time series data in a first set time period before a time point, historical expression data of each index data before the time point, and historical market condition time series data in a second set time period after the time point;
aiming at a training sample, calculating an estimated value of each index data by using a factor evaluation model according to historical market time sequence data in the first set time period and historical performance data of each index data, and setting the index data with the maximum estimated value as target index data;
according to the target index data and the historical market time sequence data in the second set time period, calculating a real value corresponding to the target index data;
and calculating an error value of the estimated value and the true value of the target index data by using a preset loss function, and adjusting parameters of the index evaluation model according to the error value.
In order to achieve the above object, the present invention further provides a computer device, which specifically includes: a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the above-described steps of the method of selecting metric data when executing the computer program.
In order to achieve the above object, the present invention also provides a computer-readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, carries out the above-mentioned steps of the method of selecting metric data.
According to the method, the device, the equipment and the storage medium for selecting the index data, when a computer processes the financial data, the characteristic extraction is firstly carried out on the financial data so as to reduce the data volume and improve the processing speed of the computer; the invention can improve the processing speed of the computer while ensuring the accuracy of the result, thereby more quickly selecting index data related to the target financial product and facilitating the later financial investment operation of enterprises.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:
fig. 1 is a schematic flow chart of an alternative method for selecting index data according to an embodiment;
fig. 2 is a schematic diagram of an alternative structure of the apparatus for selecting index data according to the second embodiment;
fig. 3 is a schematic diagram of an alternative hardware architecture of the computer device according to the third embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. 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.
Example one
An embodiment of the present invention provides a method for selecting index data, as shown in fig. 1, the method specifically includes the following steps:
step S101: when a trigger instruction is received, acquiring market situation time sequence data of a target financial product in a first set time period from a preset financial platform, and extracting market situation characteristic vectors from the market situation time sequence data.
Wherein the target financial product may be a stock, fund, future, bond. The market quotation time sequence data is formed by daily market quotation data of the target financial product on a plurality of transaction days according to the time sequence; the quotation characteristic vector is composed of a plurality of quotation characteristic data, and the quotation characteristic data are obtained according to the quotation time sequence data.
Specifically, the acquiring of the market data of the target financial product from the preset financial platform includes:
step A1: acquiring daily market data of each transaction day of the target financial product before the current time point from a preset financial platform;
step A2: performing dimension-increasing processing on the daily quotation data to expand the daily quotation data into N-dimensional data;
for example, taking the Shanghai 300-stock index as an example, the highest price, the lowest price, the average price and the yesterday difference price of the daily stock index of 5 days, 10 days, 20 days and 40 days are calculated, respectively, and the calculated data is added to the daily market data to expand the daily market data.
Step A3: and carrying out standardization processing on the daily market data.
For example, the daily quotation data is standardized by rolling z-score, the size of a rolling window is 400 transaction days, the rightmost end of the rolling window is the daily quotation data to be standardized, and the average value and the variance of all dimensional data in the rolling window are calculated to standardize the daily quotation data to be standardized.
Step A4: and forming the daily market data in the first set time period into market time sequence data.
Specifically, the extracting the market characteristic vector from the market time series data includes:
step B1: utilizing a preset Convolutional Neural Network (CNN) model to obtain the market situation time sequence data XtDetermining a time-independent global feature vector Xt″′。
Wherein, Xt∈RW×NW is the time length of the first set time period, and N is the dimension of daily market data; the CNN model comprises: 2 one-dimensional convolutional layers and 1 one-dimensional pooling layer.
In this embodiment, all daily market data of financial products (stocks or futures) needing financial operations before a certain time node t by a certain time W are acquired to form market data Xt。
Further, step B1 includes:
step B11: according to the following formula to XtPerforming convolution operation:
Figure BDA0002659978510000071
Figure BDA0002659978510000072
wherein the content of the first and second substances,
Figure BDA0002659978510000073
a feature vector output for the kth convolution kernel in the first convolution layer;
Figure BDA0002659978510000074
a feature vector output for the kth convolution kernel in the second convolution layer;
represents a convolution operation;
Figure BDA0002659978510000075
a kth one-dimensional convolution kernel for the l layer;
Figure BDA0002659978510000076
is a deviation term;
phi is the activation function.
It should be noted that, in the following description,
Figure BDA0002659978510000077
and
Figure BDA0002659978510000078
is a parameter to be learned; in the first convolution layer, h convolution kernels with the same length are adopted for convolution operation in order to determine the relation between daily market data with different dimensionalities, and the output time dimensionality of each obtained convolution kernel is consistent; in the second convolutional layer, convolutional kernels with different lengths are selected to determine the relationship between daily market data of each dimension in different time periods, and the obtained output needs to be aligned by using maximum pooling operation and processed into a global feature vector irrelevant to time.
Step B12: according to the following formula
Figure BDA0002659978510000079
Performing a max pooling operation to align and process into a time-independent global feature vector:
Figure BDA00026599785100000710
the characteristic vector formed by performing maximum pooling on the vectors calculated by all the second layer convolution kernels is XtGlobal feature vector X oft″′。
In this example, the CNN model pair X is usedtCarrying out feature extraction; wherein, the CNN model comprises: the system comprises 2 one-dimensional convolutional layers and 1 one-dimensional pooling layer, wherein the first convolutional layer is used for determining the relation among all the characteristics in the same time period, the second convolutional layer is used for determining the relation among all the characteristics in different time periods, and the pooling layer is used for extracting all the characteristics in different time periods and integrating all the characteristics into a section characteristic vector irrelevant to a time sequence.
Step B2: for the market data XtPerforming short-time Fourier transform (STFT) operation to obtain market situation time-frequency spectrogram StAnd comparing the market time-frequency spectrogram StLocal feature extraction is carried out to obtain a local feature vector St,L。
Further, step B2 includes:
step B21: the market information time sequence data X is processed according to the following formulatPerforming short-time Fourier transform (STFT) operation to obtain market time-frequency spectrogram St：
Figure BDA0002659978510000081
Wherein, STFTX(. represents a market data XtAnd performing short-time Fourier transform, wherein the short-time Fourier transform is a function of the frequency fq and the time t, and the obtained time-frequency spectrogram of the short-time Fourier transform can depict the time-frequency domain information of the market quotation.
Step B22: using a two-dimensional convolutional neural network CNN model pair StPerforming convolution operation to obtain time-frequency domain local image characteristics
Figure BDA0002659978510000082
Figure BDA0002659978510000083
Wherein, represents a convolution operation;
Figure BDA0002659978510000084
the g two-dimensional convolution kernel of the l layer;
Figure BDA0002659978510000085
is the g deviation term of the l layer;
phi is the activation function.
It should be noted that, in the following description,
Figure BDA0002659978510000086
and
Figure BDA0002659978510000087
is a parameter to be learned; the two-dimensional CNN convolution model is used to slide the convolution kernel along both time and frequency dimensions of the image.
Step B23: for the time-frequency domain local image characteristics
Figure BDA0002659978510000088
Performing maximum pooling operation to obtain local feature vector St,L。
Wherein the local feature vector St,LIs obtained by stitching together all two-dimensional convolution kernel results.
Step B3: by means of an attention mechanism, passing the global feature vector Xt"' determining the local feature vector St,LThe importance of (a) to (b),to obtain the local feature vector St,LCorresponding time-frequency spectrogram local feature vector St,aAnd combining the global feature vector Xt"' and the local feature vector S of the time-frequency spectrogramt,aFusing to obtain the financial market feature vector mktt。
Further, step B3 includes:
step B31: the ith local feature is calculated according to the following formula
Figure BDA0002659978510000091
Importance of wi：
Figure BDA0002659978510000092
Wherein f isatt(. is a non-linear mapping for computing the importance w of each local feature under the guidance of the global featurei。
Step B32: the ith local feature is calculated according to the following formula
Figure BDA0002659978510000093
Weight value of alphai：
Figure BDA0002659978510000094
Step B33: using a weight value alphaiFor local characteristics
Figure BDA0002659978510000095
Carrying out weighting and operation to obtain a local characteristic vector S of the time-frequency spectrogramt,a：
Figure BDA0002659978510000096
Step B34: global feature vector Xt"' and the local feature vector S of the time-frequency spectrogramt，aSplicingGet up to obtain the financial market characteristic vector mktt：
mktt＝(Xt″′；St,a)。
Step S102: historical performance data of each preset index data are obtained from a preset database, and historical performance characteristic vectors are extracted from the historical performance data.
The index data is used for guiding the financial operation of the financial product, and can directly reflect the state of the stock market and provide a guiding direction for the financial operation behavior. For example, the relative intensity index (RSI), the random index (KD), the trend index (DMI), the smooth iso-mean line (MACD), the energy tide (OBV), the psychographic line, the divergence ratio, and the like.
The historical performance data is calculated according to the historical market data of the financial products and preset rules; for example, historical performance data of each index data in a set time period T is calculated according to a preset rule according to daily performance data of the time point T and historical performance time series data in the set time period T after the time point T, and the historical performance data is stored in the database. In the database, including: historical performance data of each index data three months away from the current time point, historical performance data of each index data six months away from the current time point, and historical performance data of each index data twelve months away from the current time point.
Specifically, the historical performance data of one index data includes: rolling accumulated profitability crt(clinical Return), maximum withdrawal mdt(Max Drawback), the sharp rate srt(Sharpe Ratio)；
Further, in the above-mentioned case,
Figure BDA0002659978510000101
wherein r isiThe daily rate of return for the ith index data;
Figure BDA0002659978510000102
Figure BDA0002659978510000103
further, the extracting historical performance feature vectors from the historical performance data includes:
historical performance data P of index datai t∈RMInputting the data into a preset multilayer neural network model to obtain historical performance characteristic data of the index data
Figure BDA0002659978510000104
Wherein M is the number of evaluation indexes used for evaluating the historical performance of the index data, i belongs to n, and n is the total number of preset index data;
the historical performance characteristic data of all index data form a historical performance characteristic vector pfmt。
Further, in the above-mentioned case,
Figure BDA0002659978510000105
Figure BDA0002659978510000106
is the historical performance characteristic of the ith index data, the MLP is a multilayer neural network, and the operation of each layer is:
Figure BDA0002659978510000107
wherein the content of the first and second substances,
Figure BDA0002659978510000108
for the output of each fully-connected layer, sigma is the activation function sigmoid function of the neural network, and the MLP network is formed by stacking the above single-layer neural networks for multiple times, wherein the parameters of each layer
Figure BDA0002659978510000109
bpAre different and require separate training. Historical performance of each index data
Figure BDA00026599785100001010
The historical expression characteristic vector pfm can be obtained by using the same MLP to extract the characteristicst。
Step S103: inputting the market condition characteristic vector and the historical expression characteristic vector into a preset index evaluation model to obtain an estimation value of each index data; the index evaluation model is obtained by training with the profitability of the target financial product in a second set time period as a target.
It should be noted that, an estimated value of the index data is used for representing, and if a financial operation corresponding to the index data is adopted in the future, the rate of return that can be obtained based on the financial operation is obtained in the future; the estimation value is used for estimating the yield in the next time state according to the information in the current time state.
Specifically, the factor evaluation model is a multilayer neural network, and an evaluation value Q of index data is calculated according to the following formulat：
Qt＝Wqfq(mktt；pfmt)+bq；
Wherein f isqRepresenting a plurality of fully-connected layers with activation functions, wherein the activation functions are sigmoids, and the output values of the last layer of neural network do not need to be processed by the activation functions.
Further, before step S103, the method further includes:
step C1: acquiring a training sample set; wherein the training sample set comprises a plurality of training samples, and each training sample comprises: the historical market data of the index data are obtained by the historical market time sequence data in a first set time period before a time point, the historical expression data of each index data before the time point and the historical market time sequence data in a second set time period after the time point.
In this embodiment, a training sample set is formed according to the historical market time series data of the target financial product and the historical performance data of each preset index data.
Step C2: and aiming at a training sample, calculating the estimation value of each index data by using a factor evaluation model according to the historical market time sequence data in the first set time period and the historical performance data of each index data, and setting the index data with the maximum estimation value as target index data.
Step C3: and according to the target index data and the historical market time sequence data in the second set time period, calculating a true value corresponding to the target index data.
Step C4: and calculating an error value of the estimated value and the true value of the target index data by using a preset loss function, and adjusting parameters of the index evaluation model according to the error value.
Step C5: the above steps C2 to C4 are repeatedly performed until the error value is minimized, thereby obtaining the index estimation model.
It should be noted that, in this embodiment, an Xavier algorithm is used to randomly initialize each parameter in the index evaluation model;
preferably, the loss function in step C4 is as follows:
Figure BDA0002659978510000111
wherein L isi(θi) Is an error value;
Figure BDA0002659978510000112
the actual value of the target index data;
Q(s,a；θi) Is an estimated value of target index data a, s is historical market time sequence data in a first set time period before a time state point and historical expression data of all preset index data before the time state point, and thetaiEvaluating parameters in the model for the index;
preferably, in step C3Calculating the true value y of the target index data according to the following formulat：
Figure BDA0002659978510000121
Wherein R istAnd gamma is a parameter for measuring the influence of the reward value of the future time state on the reward value in the second set time period after the time state point, and the end symbol is a mark for the state, and when the target index data falls into a poor state, the exploration is stopped, and the end symbol is marked for the last state.
Step S104: and arranging all index data according to the estimated values from large to small, and sending the index data arranged before the set number to the appointed terminal.
In this embodiment, when the computer processes the financial data, the feature extraction is performed on the financial data first to reduce the data amount, so as to improve the processing speed of the computer and obtain the financial policy more quickly. Specifically, in this embodiment, a financial market characteristic vector is extracted from market time series data of a period of time, a historical expression characteristic vector is extracted from historical expression of index data, and the computer selects the index data based on the financial market characteristic vector and the historical expression characteristic vector by using a machine learning algorithm. The embodiment can improve the processing speed of the computer while ensuring the accuracy of the result, thereby more quickly selecting the index data related to the target financial product and facilitating the later investment operation of enterprises.
Example two
An embodiment of the present invention provides a device for selecting index data, as shown in fig. 2, the device specifically includes the following components:
the first extraction module 201 is configured to, when a trigger instruction is received, acquire market situation time series data of a target financial product in a first set time period from a preset financial platform, and extract market situation feature vectors from the market situation time series data;
the second extraction module 202 is configured to obtain historical performance data of each preset index data from a preset database, and extract a historical performance feature vector from the historical performance data;
the input module 203 is configured to input the market characteristic vector and the historical performance characteristic vector into a preset index evaluation model to obtain an estimation value of each index data; the index evaluation model is obtained by training with the profitability of the target financial product in a second set time period as a target;
and the sending module 204 is configured to rank all the index data according to the estimated values from large to small, and send the index data ranked before the set number to the designated terminal.
Specifically, the first extraction module 201 is specifically configured to, when the function of acquiring the market data of the target financial product in the first set time period from the preset financial platform is implemented:
acquiring daily market data of each transaction day of the target financial product before the current time point from a preset financial platform;
performing dimension increasing processing and standardization processing on the daily market data;
and forming the daily market data in the first set time period into market time sequence data.
Further, when the function of extracting the market characteristic vector from the market time series data is implemented, the first extraction module 201 is specifically configured to:
determining a global feature vector irrelevant to time according to the market condition time sequence data by using a preset convolutional neural network model;
performing short-time Fourier transform on the quotation time sequence data to obtain a quotation time-frequency spectrogram, and performing local feature extraction on the quotation time-frequency spectrogram to obtain a local feature vector;
determining the importance of the local feature vector through the global feature vector by using an attention mechanism to obtain a time-frequency spectrogram local feature vector corresponding to the local feature vector;
and fusing the global feature vector and the time-frequency spectrogram local feature vector to obtain a market situation feature vector.
Further, the historical performance data of one index data includes: rolling cumulative profitability, maximum pullback, and sharp rate;
the second extraction module 202 is specifically configured to:
inputting the rolling accumulated yield, the maximum withdrawal and the sharp rate of index data into a preset multilayer neural network model to obtain historical performance characteristic data of the index data;
and forming historical performance characteristic data of all index data into a historical performance characteristic vector.
Still further, the apparatus further comprises:
the training module is used for acquiring a training sample set; wherein the training sample set comprises a plurality of training samples, and each training sample comprises: historical market condition time series data in a first set time period before a time point, historical expression data of each index data before the time point, and historical market condition time series data in a second set time period after the time point;
aiming at a training sample, calculating an estimated value of each index data by using a factor evaluation model according to historical market time sequence data in the first set time period and historical performance data of each index data, and setting the index data with the maximum estimated value as target index data;
according to the target index data and the historical market time sequence data in the second set time period, calculating a real value corresponding to the target index data;
and calculating an error value of the estimated value and the true value of the target index data by using a preset loss function, and adjusting parameters of the index evaluation model according to the error value.
EXAMPLE III
The embodiment also provides a computer device, such as a smart phone, a tablet computer, a notebook computer, a desktop computer, a rack server, a blade server, a tower server or a rack server (including an independent server or a server cluster composed of a plurality of servers) capable of executing programs, and the like. As shown in fig. 3, the computer device 30 of the present embodiment includes at least but is not limited to: a memory 301, a processor 302 communicatively coupled to each other via a system bus. It is noted that FIG. 3 only shows the computer device 30 having components 301 and 302, but it is understood that not all of the shown components are required and that more or fewer components may be implemented instead.
In this embodiment, the memory 301 (i.e., the readable storage medium) includes a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, and the like. In some embodiments, the storage 301 may be an internal storage unit of the computer device 30, such as a hard disk or a memory of the computer device 30. In other embodiments, the memory 301 may also be an external storage device of the computer device 30, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like, provided on the computer device 30. Of course, the memory 301 may also include both internal and external storage devices for the computer device 30. In the present embodiment, the memory 301 is generally used for storing an operating system and various types of application software installed in the computer device 30. In addition, the memory 301 may also be used to temporarily store various types of data that have been output or are to be output.
Processor 302 may be a Central Processing Unit (CPU), controller, microcontroller, microprocessor, or other data Processing chip in some embodiments. The processor 302 generally serves to control the overall operation of the computer device 30.
Specifically, in the present embodiment, the processor 302 is configured to execute a program of a method for selecting index data stored in the processor 302, and the program of the method for selecting index data implements the following steps when executed:
when a trigger instruction is received, acquiring market situation time sequence data of a target financial product in a first set time period from a preset financial platform, and extracting market situation characteristic vectors from the market situation time sequence data;
acquiring historical performance data of each preset index data from a preset database, and extracting historical performance characteristic vectors from the historical performance data;
inputting the market condition characteristic vector and the historical expression characteristic vector into a preset index evaluation model to obtain an estimation value of each index data; the index evaluation model is obtained by training with the profitability of the target financial product in a second set time period as a target;
and arranging all index data according to the estimated values from large to small, and sending the index data arranged before the set number to the appointed terminal.
The specific embodiment process of the above method steps can be referred to in the first embodiment, and the detailed description of this embodiment is not repeated here.
Example four
The present embodiments also provide a computer readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a Read Only Memory (ROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), a Programmable Read Only Memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an App application mall, etc., having stored thereon a computer program that when executed by a processor implements the method steps of:
when a trigger instruction is received, acquiring market situation time sequence data of a target financial product in a first set time period from a preset financial platform, and extracting market situation characteristic vectors from the market situation time sequence data;
acquiring historical performance data of each preset index data from a preset database, and extracting historical performance characteristic vectors from the historical performance data;
inputting the market condition characteristic vector and the historical expression characteristic vector into a preset index evaluation model to obtain an estimation value of each index data; the index evaluation model is obtained by training with the profitability of the target financial product in a second set time period as a target;
and arranging all index data according to the estimated values from large to small, and sending the index data arranged before the set number to the appointed terminal.
The specific embodiment process of the above method steps can be referred to in the first embodiment, and the detailed description of this embodiment is not repeated here.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
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.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.

Claims (10)

1. A method of selecting metric data, the method comprising:
when a trigger instruction is received, acquiring market situation time sequence data of a target financial product in a first set time period from a preset financial platform, and extracting market situation characteristic vectors from the market situation time sequence data;
acquiring historical performance data of each preset index data from a preset database, and extracting historical performance characteristic vectors from the historical performance data;
inputting the market condition characteristic vector and the historical expression characteristic vector into a preset index evaluation model to obtain an estimation value of each index data; the index evaluation model is obtained by training with the profitability of the target financial product in a second set time period as a target;
and arranging all index data according to the estimated values from large to small, and sending the index data arranged before the set number to the appointed terminal.
2. The method according to claim 1, wherein the obtaining of market data of the target financial product from the predetermined financial platform within a first predetermined time period specifically comprises:
acquiring daily market data of each transaction day of the target financial product before the current time point from a preset financial platform;
performing dimension increasing processing and standardization processing on the daily market data;
and forming the daily market data in the first set time period into market time sequence data.
3. The method of selecting index data according to claim 1, wherein the extracting a market condition feature vector from the market condition time series data specifically includes:
determining a global feature vector irrelevant to time according to the market condition time sequence data by using a preset convolutional neural network model;
performing short-time Fourier transform on the quotation time sequence data to obtain a quotation time-frequency spectrogram, and performing local feature extraction on the quotation time-frequency spectrogram to obtain a local feature vector;
determining the importance of the local feature vector through the global feature vector by using an attention mechanism to obtain a time-frequency spectrogram local feature vector corresponding to the local feature vector;
and fusing the global feature vector and the time-frequency spectrogram local feature vector to obtain a market situation feature vector.
4. The method of selecting metric data of claim 1, wherein the historical performance data of a metric data comprises: rolling cumulative profitability, maximum pullback, and sharp rate;
the extracting of the historical expression feature vector from the historical expression data specifically includes:
inputting the rolling accumulated yield, the maximum withdrawal and the sharp rate of index data into a preset multilayer neural network model to obtain historical performance characteristic data of the index data;
and forming historical performance characteristic data of all index data into a historical performance characteristic vector.
5. A method of selecting metric data as defined in claim 1, the method further comprising:
acquiring a training sample set; wherein the training sample set comprises a plurality of training samples, and each training sample comprises: historical market condition time series data in a first set time period before a time point, historical expression data of each index data before the time point, and historical market condition time series data in a second set time period after the time point;
aiming at a training sample, calculating an estimated value of each index data by using a factor evaluation model according to historical market time sequence data in the first set time period and historical performance data of each index data, and setting the index data with the maximum estimated value as target index data;
according to the target index data and the historical market time sequence data in the second set time period, calculating a real value corresponding to the target index data;
and calculating an error value of the estimated value and the true value of the target index data by using a preset loss function, and adjusting parameters of the index evaluation model according to the error value.
6. An apparatus for selecting metric data, the apparatus comprising:
the system comprises a first extraction module, a second extraction module and a third extraction module, wherein the first extraction module is used for acquiring market quotation time sequence data of a target financial product in a first set time period from a preset financial platform and extracting market quotation feature vectors from the market quotation time sequence data when a trigger instruction is received;
the second extraction module is used for acquiring historical expression data of each preset index data from a preset database and extracting historical expression characteristic vectors from the historical expression data;
the input module is used for inputting the market characteristic vector and the historical expression characteristic vector into a preset index evaluation model so as to obtain an estimation value of each index data; the index evaluation model is obtained by training with the profitability of the target financial product in a second set time period as a target;
and the sending module is used for arranging all the index data according to the estimated values from large to small and sending the index data arranged before the set number to the appointed terminal.
7. The apparatus for selecting metric data of claim 6, wherein the first extraction module is specifically configured to:
determining a global feature vector irrelevant to time according to the market condition time sequence data by using a preset convolutional neural network model;
performing short-time Fourier transform on the quotation time sequence data to obtain a quotation time-frequency spectrogram, and performing local feature extraction on the quotation time-frequency spectrogram to obtain a local feature vector;
determining the importance of the local feature vector through the global feature vector by using an attention mechanism to obtain a time-frequency spectrogram local feature vector corresponding to the local feature vector;
and fusing the global feature vector and the time-frequency spectrogram local feature vector to obtain a market situation feature vector.
8. An apparatus for selecting metric data as defined in claim 6, further comprising:
the training module is used for acquiring a training sample set; wherein the training sample set comprises a plurality of training samples, and each training sample comprises: historical market condition time series data in a first set time period before a time point, historical expression data of each index data before the time point, and historical market condition time series data in a second set time period after the time point;
aiming at a training sample, calculating an estimated value of each index data by using a factor evaluation model according to historical market time sequence data in the first set time period and historical performance data of each index data, and setting the index data with the maximum estimated value as target index data;
according to the target index data and the historical market time sequence data in the second set time period, calculating a real value corresponding to the target index data;
and calculating an error value of the estimated value and the true value of the target index data by using a preset loss function, and adjusting parameters of the index evaluation model according to the error value.
9. A computer device, the computer device comprising: memory, processor and computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method of any of claims 1 to 5 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 5.
CN202010901797.8A 2020-09-01 2020-09-01 Method, device, equipment and storage medium for selecting index data Pending CN112016840A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010901797.8A CN112016840A (en) 2020-09-01 2020-09-01 Method, device, equipment and storage medium for selecting index data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010901797.8A CN112016840A (en) 2020-09-01 2020-09-01 Method, device, equipment and storage medium for selecting index data

Publications (1)

Publication Number Publication Date
CN112016840A true CN112016840A (en) 2020-12-01

Family

ID=73516426

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010901797.8A Pending CN112016840A (en) 2020-09-01 2020-09-01 Method, device, equipment and storage medium for selecting index data

Country Status (1)

Country Link
CN (1) CN112016840A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113420096A (en) * 2021-06-22 2021-09-21 平安科技（深圳）有限公司 Index system construction method, device, equipment and storage medium
CN117633605A (en) * 2024-01-25 2024-03-01 浙江鹏信信息科技股份有限公司 Data security classification capability maturity assessment method, system and readable medium

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113420096A (en) * 2021-06-22 2021-09-21 平安科技（深圳）有限公司 Index system construction method, device, equipment and storage medium
CN113420096B (en) * 2021-06-22 2024-05-10 平安科技（深圳）有限公司 Index system construction method, device, equipment and storage medium
CN117633605A (en) * 2024-01-25 2024-03-01 浙江鹏信信息科技股份有限公司 Data security classification capability maturity assessment method, system and readable medium
CN117633605B (en) * 2024-01-25 2024-04-12 浙江鹏信信息科技股份有限公司 Data security classification capability maturity assessment method, system and readable medium

Similar Documents

Publication Publication Date Title
CN110060144B (en) Method for training credit model, method, device, equipment and medium for evaluating credit
Wu et al. Portfolio management system in equity market neutral using reinforcement learning
CN110263821B (en) Training of transaction feature generation model, and method and device for generating transaction features
CN110866628A (en) System and method for multi-bounded time series prediction using dynamic time context learning
CN109657932A (en) Business risk analysis method, device, computer equipment and storage medium
CN108763277B (en) Data analysis method, computer readable storage medium and terminal device
CN109583966A (en) A kind of high value customer recognition methods, system, equipment and storage medium
US20200380524A1 (en) Transaction feature generation
CN112016840A (en) Method, device, equipment and storage medium for selecting index data
CN108804564A (en) The combined recommendation method and terminal device of financial product
CN107679987A (en) Asset Allocation strategy acquisition methods, device, computer equipment and storage medium
CN116822651A (en) Large model parameter fine adjustment method, device, equipment and medium based on incremental learning
CN109934381A (en) The real-time estimation method of fund, device, computer equipment and storage medium
CN111582932A (en) Inter-scene information pushing method and device, computer equipment and storage medium
Oktoviany et al. A machine learning-based price state prediction model for agricultural commodities using external factors
Jiang Using machine learning to analyze merger activity
CN109800138B (en) CPU testing method, electronic device and storage medium
CN113408582B (en) Training method and device for feature evaluation model
Panos et al. Large scale multi-label learning using gaussian processes
CN111144098B (en) Recall method and device for extended question
CN109886299A (en) A kind of user draws a portrait method, apparatus, readable storage medium storing program for executing and terminal device
CN109885745A (en) A kind of user draws a portrait method, apparatus, readable storage medium storing program for executing and terminal device
CN109784406A (en) A kind of user draws a portrait method, apparatus, readable storage medium storing program for executing and terminal device
Khiem Tran et al. Towards Improved Bankruptcy Prediction: Utilizing Variational Autoencoder Latent Representations in a Norwegian Context
Li et al. Dataset complexity assessment based on cumulative maximum scaled area under Laplacian spectrum

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination