WO2020037922A1 - 股指预测方法、装置及存储介质 - Google Patents

股指预测方法、装置及存储介质 Download PDF

Info

Publication number
WO2020037922A1
WO2020037922A1 PCT/CN2018/123583 CN2018123583W WO2020037922A1 WO 2020037922 A1 WO2020037922 A1 WO 2020037922A1 CN 2018123583 W CN2018123583 W CN 2018123583W WO 2020037922 A1 WO2020037922 A1 WO 2020037922A1
Authority
WO
WIPO (PCT)
Prior art keywords
stock index
index
preset time
time point
model
Prior art date
Application number
PCT/CN2018/123583
Other languages
English (en)
French (fr)
Inventor
李海疆
Original Assignee
平安科技（深圳）有限公司
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 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2020037922A1 publication Critical patent/WO2020037922A1/zh

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
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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
    • G06Q40/06—Asset management; Financial planning or analysis

Definitions

  • the present application relates to the field of computer technology and intelligent prediction technology, and in particular, to a stock index prediction method, device, and computer-readable storage medium.
  • the stock market is an investment market with high risks and returns. Predicting it to better select stocks and time to obtain maximum returns is an area that researchers pay close attention to.
  • Predictive analysis methods that have been used so far can be divided into two categories: fundamental analysis methods and technical analysis methods.
  • the fundamental analysis method starts with information such as the national economic policy and the fundamentals of the company, while the technical analysis method focuses on using historical data to bring it into mathematical models or machines for training and calculation. Among them, the national macroeconomic conditions, corporate profitability and other factors applied in the fundamental analysis method are difficult to quantify, and most of them are long-term factors. Therefore, if only the fundamental analysis method is used, the accuracy of prediction will be very inaccurate.
  • the technical analysis method mainly uses objective quantitative indicators to make predictions.
  • the more commonly used methods include time series method, wavelet analysis method, neural network method, etc., but due to the restrictions of the stock market and its own characteristics such as stocks and futures, such as various influencing factors Most of them do not have obvious correlations, and many current methods cannot achieve good results in prediction. For example, it is difficult to measure the interaction of multiple index factors using time series prediction methods alone, and it is difficult to obtain very good results when processing nonlinear characteristic data. Good prediction results, while the currently popular support vector machine prediction models of neural networks only have excellent generalization ability for solving nonlinear problems with small samples, and it is difficult to accurately predict the trend of stock indexes.
  • this application provides a stock index prediction method, an electronic device, and a computer-readable storage medium, the main purpose of which is to extract deep-level feature vectors from a variety of index factors, so as to scientifically and accurately predict the stock index return.
  • the stock index prediction method provided in this application includes the following steps:
  • Sample acquisition step selecting a time series of n index factors in a preset time period, obtaining a stock index return rate at each first preset time point in the preset time period, and according to the time series of the n index factors, Constructing a first n-dimensional vector at each first preset time point;
  • Model training step using the first n-dimensional vectors at multiple first preset time points and the stock index return rate as sample data, using a back-propagation algorithm to perform a pre-established dual-recurrent neural network model based on a gated recurrent unit Training, iteratively updating model weights to obtain stock index prediction models;
  • Prediction step collect data of the n index factors at a second preset time point, construct a second n-dimensional vector at a second preset time point based on the collected data, and input the second n-dimensional vector into the stock index prediction The model predicts the stock index return rate at the second preset time point.
  • the present application also provides an electronic device.
  • the electronic device includes a memory and a processor.
  • the memory includes a stock index prediction program. When the stock index prediction program is executed by the processor, the following steps are implemented:
  • Sample acquisition step selecting a time series of n index factors in a preset time period, obtaining a stock index return rate at each first preset time point in the preset time period, and according to the time series of the n index factors, Constructing a first n-dimensional vector at each first preset time point;
  • Model training step using the plurality of first n-dimensional vectors at a first preset time point and the stock index return rate as sample data, using a back-propagation algorithm to perform a pre-established dual-recurrent neural network model based on a gated recurrent unit Training, iteratively updating model weights to obtain stock index prediction models;
  • Prediction step collect data of the n index factors at a second preset time point, construct a second n-dimensional vector at a second preset time point based on the collected data, and input the second n-dimensional vector into the stock index prediction The model predicts the stock index return rate at the second preset time point.
  • the present application also provides a computer-readable storage medium, the computer-readable storage medium includes a stock index prediction program, and when the stock index prediction program is executed by the processor, the following steps are implemented:
  • Sample acquisition step selecting a time series of n index factors in a preset time period, obtaining a stock index return rate at each first preset time point in the preset time period, and according to the time series of the n index factors, Constructing a first n-dimensional vector at each first preset time point;
  • Model training step using the plurality of first n-dimensional vectors at a first preset time point and the stock index return rate as sample data, using a back-propagation algorithm to perform a pre-established dual-recurrent neural network model based on a gated recurrent unit Training, iteratively updating model weights to obtain stock index prediction models;
  • Prediction step collect data of the n index factors at a second preset time point, construct a second n-dimensional vector at a second preset time point based on the collected data, and input the second n-dimensional vector into the stock index prediction The model predicts the stock index return rate at the second preset time point.
  • the stock index prediction method, device, and computer-readable storage medium proposed in this application obtain the stock index return rate at each first preset time point of the preset time period by selecting a time series of n index factors of the preset time period. And construct a first n-dimensional vector for each first preset time point according to the time series of the n index factors, and then use the first n-dimensional vectors of the plurality of first preset time points and stock index returns
  • the rate is sample data.
  • the back-propagation algorithm is used to train a pre-established bicyclic neural network model based on the gating recurrent unit, iteratively updates the model weights, and obtains a stock index prediction model.
  • the double-recurrent neural network model based on the gating recurrent unit can extract deep-level features from various index factors, the present application can accurately predict the stock index.
  • FIG. 1 is a schematic diagram of an electronic device according to an embodiment of the present application.
  • FIG. 2 is a program module diagram of an embodiment of a stock index prediction program in FIG. 1;
  • FIG. 3 is a flowchart of an embodiment of a stock index prediction method of the present application.
  • FIG. 1 is a schematic diagram of an embodiment of an electronic device 1 according to the present application.
  • the electronic device 1 acquires sample data, uses a back-propagation algorithm to train a pre-established bicyclic neural network model based on a gated recurring unit, updates the model weights iteratively, obtains a stock index prediction model, and then uses the The stock index prediction model predicts the stock index return rate at a preset time point.
  • the electronic device 1 may be a terminal device having storage and computing functions, such as a server, a smart phone, a tablet computer, a portable computer, a desktop computer, and the like.
  • the server when the electronic device 1 is a server, the server may be one or more of a rack server, a blade server, a tower server, or a rack server.
  • the electronic device 1 includes a memory 11, a processor 12 and a network interface 13.
  • the memory 11 includes at least one type of readable storage medium.
  • the at least one type of readable storage medium may be a non-volatile storage medium such as a flash memory, a hard disk, a multimedia card, a card-type memory, and the like.
  • the readable storage medium may be an internal storage unit of the electronic device 1, such as a hard disk of the electronic device 1.
  • the readable storage medium may also be the external memory 11 of the electronic device 1, such as a plug-in hard disk and a smart memory card (SMC) provided on the electronic device 1. , Secure Digital (SD) Card, Flash Card (Flash Card), etc.
  • SD Secure Digital
  • Flash Card Flash Card
  • the readable storage medium of the memory 11 is generally used to store an operating system, a stock index prediction program 10, a pre-established double-loop neural network model based on a gating recurrent unit, and various index factor data collected Wait.
  • the memory 11 may also be used to temporarily store data that has been or will be output.
  • the processor 12 may be a central processing unit (CPU), a microprocessor, or other data processing chip, and is configured to run program codes or process data stored in the memory 11, for example, to execute The stock index forecasting procedure 10 and so on.
  • CPU central processing unit
  • microprocessor microprocessor
  • other data processing chip and is configured to run program codes or process data stored in the memory 11, for example, to execute The stock index forecasting procedure 10 and so on.
  • the network interface 13 may include a standard wired interface and a wireless interface (such as a WI-FI interface). It is generally used to establish a communication connection between the electronic device 1 and other electronic devices or systems.
  • FIG. 1 shows only the electronic device 1 having components 11-13 and the stock index prediction program 10, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead. .
  • the electronic device 1 may further include a display, which may also be referred to as a display screen or a display unit.
  • a display may also be referred to as a display screen or a display unit.
  • it may be an LED display, a liquid crystal display, a touch-type liquid crystal display, an organic light-emitting diode (OLED) display, or the like.
  • the display is used to display information processed in the electronic device 1 and to display a visualized user interface.
  • the electronic device 1 further includes a touch sensor.
  • An area provided by the touch sensor for a user to perform a touch operation is referred to as a touch area.
  • the touch sensor described herein may be a resistive touch sensor, a capacitive touch sensor, or the like.
  • the touch sensor includes not only a contact type touch sensor, but also a proximity type touch sensor and the like.
  • the touch sensor may be a single sensor, or may be a plurality of sensors arranged in an array, for example. The user may start the stock index prediction program 10 by touching the touch area.
  • the area of the display of the electronic device 1 may be the same as that of the touch sensor, or may be different.
  • a display and the touch sensor are stacked to form a touch display screen. The device detects a touch operation triggered by a user based on a touch display screen.
  • the electronic device 1 may further include a radio frequency (RF) circuit, a sensor, an audio circuit, and the like, which are not described herein again.
  • RF radio frequency
  • Sample acquisition step selecting a time series of n index factors in a preset time period, obtaining a stock index return rate at each first preset time point in the preset time period, and according to the time series of the n index factors, Constructing a first n-dimensional vector at each first preset time point;
  • Model training step using the plurality of first n-dimensional vectors at a first preset time point and the stock index return rate as sample data, using a back-propagation algorithm to perform a pre-established dual-recurrent neural network model based on a gated recurrent unit Training, iteratively updating model weights to obtain stock index prediction models;
  • Prediction step collect data of the n index factors at a second preset time point, construct a second n-dimensional vector at a second preset time point based on the collected data, and input the second n-dimensional vector into the stock index prediction The model predicts the stock index return rate at the second preset time point.
  • the stock index prediction program 10 may be divided into multiple modules, and the multiple modules are stored in the memory 12 and executed by the processor 13 to complete the application.
  • the module referred to in this application refers to a series of computer program instruction segments capable of performing specific functions.
  • the stock index prediction program 10 may be divided into: a sample acquisition module 110, a preprocessing module 120, a model training module 130, and a prediction module 140. among them:
  • the sample acquisition module 110 is configured to select a time series of n index factors of a preset time period, obtain a stock index return rate at each first preset time point of the preset time period, and according to the n kinds of indexes
  • the time series of factors constructs a first n-dimensional vector for each first preset time point.
  • a time series of 13 index factors explaining the trend of the Shanghai and Shenzhen 300 Index in the past two months is selected from a preset data platform, and is constructed for each first preset time point (such as the daily opening time) according to the selected time series.
  • a 13-dimensional vector and obtain the daily return rate of the Shanghai and Shenzhen 300 stock index futures as the return rate of the stock index at each first preset time point.
  • the index factor may include only one or more of the above-mentioned index factors, and may also be other statistical data that affects the trend of the stock index, which is not repeated here.
  • the pre-processing module 120 is configured to pre-process the index factor data of the time series, and use Lagrangian interpolation to repair missing values in the index factor data.
  • the preprocessing includes removing noise, and performing interpolation adjustment on missing values in the data by using Lagrange interpolation. For example, when constructing a multidimensional vector using a time series of exponential factors, it may happen that the acquisition of exponential factor data fails. At this time, Lagrange interpolation can be used to supplement the exponential factor data that failed to obtain the multidimensional vector.
  • the model training module 130 is configured to use the first n-dimensional vector and the stock index return rate of the plurality of first preset time points as sample data, and use a back-propagation algorithm to pre-establish the double-based
  • the recurrent neural network model is trained, the model weights are updated iteratively, and the stock index prediction model is obtained.
  • the pre-established bicyclic neural network model based on a gated recurrent unit includes two hidden layers, and the hidden layers are used to abstract the n-dimensional vector into a preset dimension (for example, two-dimensional) Feature vector, the number of neurons in each hidden layer is equal to the value of the preset dimension of the feature vector.
  • the formula of the back propagation algorithm is:
  • a ′ k represents the k-th model weight after the update
  • a k represents the k-th model weight before the update
  • ⁇ represents the learning rate
  • the model weight is determined, the model training is ended, and the trained stock index prediction model is obtained.
  • a 13-dimensional vector composed of 13 index factors explaining the Shanghai-Shenzhen 300 Index trend at the opening time of a certain day is input into the pre-established dual-recurrent neural network model based on a gated recurrent unit, and the model has a hidden layer
  • the output result is the feature vector abstracted from the 13-dimensional vector.
  • the feature vector reflects the deep features of the index factor data.
  • the predicted value of the Shanghai and Shenzhen 300 stock index returns can be calculated on that day.
  • the stock index return rate obtained by the sample acquisition module 110 is the true value of the stock index return rate.
  • the prediction module 140 is configured to collect data of the n index factors at a second preset time point, construct a second n-dimensional vector at the second preset time point according to the collected data, and convert the second n-dimensional vector
  • the stock index prediction model is input, and the stock index return rate at the second preset time point is predicted. For example, the data of 13 index factors explaining the trend of the Shanghai and Shenzhen 300 Index at the time of opening today is collected, and a 13-dimensional vector of today's opening time is constructed based on the collected data.
  • the stock index forecasting model predicts the stock index return of the Shanghai and Shenzhen 300 Index today.
  • this application also provides a stock index prediction method.
  • a flowchart of an embodiment of a stock index prediction method of the present application is shown.
  • the processor 12 of the electronic device 1 executes the stock index prediction program 10 stored in the memory, the following steps of the stock index prediction method are implemented:
  • Step S300 the sample acquisition module 110 selects a time series of n index factors in a preset time period, obtains a stock index return rate at each first preset time point in the preset time period, and according to the n index factors.
  • the time sequence constructs a first n-dimensional vector for each first preset time point. For example, select a time series of multiple index factors that explain the trend of the Shanghai and Shenzhen 300 Index from the preset data platform in the past two months, and according to the selected time series, each first preset time point (such as the daily opening time) Construct a multi-dimensional vector, and obtain the daily return rate of the Shanghai and Shenzhen 300 stock index futures as the return rate of the stock index at each first preset time point.
  • the various index factors may include ChinaBond corporate bond maturity yield (AAA): 10 years, risk premium, dividend yield, Slow KD indicator (SlowKD), MACD Histogram, premium rate, active purchase amount, etc.
  • AAA ChinaBond corporate bond maturity yield
  • the risk premium has a negative correlation with the dividend rate.
  • the CSI 300 dividend yield has a negative correlation with the closing price of the CSI 300 stock index.
  • This application chooses to train a dual recurrent neural network model based on a gated recurring unit, precisely to extract deeper features that are not easily detectable from the index factor, so as to improve the accuracy of stock index prediction.
  • the model training module 120 uses the first n-dimensional vectors and the stock index return rate of the plurality of first preset time points as sample data, and uses a back-propagation algorithm to apply a pre-established double-circulation nerve based on a gated circulation unit.
  • the network model is trained, the model weights are updated iteratively, and the stock index prediction model is obtained.
  • the pre-established bicyclic neural network model based on a gated recurrent unit includes two hidden layers, and the hidden layers are used to abstract the n-dimensional vector into a preset dimension (for example, two-dimensional) Feature vector, the number of neurons in each hidden layer is equal to the value of the preset dimension of the feature vector.
  • the formula of the back propagation algorithm is:
  • a ′ k represents the k-th model weight after the update
  • a k represents the k-th model weight before the update
  • ⁇ represents the learning rate
  • a multi-dimensional vector composed of multiple index factors explaining the trend of the Shanghai and Shenzhen 300 Index at the opening time of a certain day is input into the pre-established double-circulation neural network model based on the gating recurrent unit, and the output result of the hidden layer of the model is It is a feature vector abstracted from the multi-dimensional vector, which reflects the deep-level features of the index factor data.
  • the predicted value of the Shanghai and Shenzhen 300 stock index returns can be calculated on that day. Assume that the predicted value of the stock index return rate on a certain day is P, and the true value of the stock index return rate obtained by the sample acquisition module 110 is T.
  • the overall error E (PT) ⁇ 2
  • the overall error E and the model to be updated Multiply the partial derivative value of the weight by the learning rate to obtain the offset value, calculate the difference between the weight of the model to be updated and the offset value, obtain the updated model weight, and iteratively update the model weight until the offset value reaches a preset value, Determine the model weights and get the trained stock index prediction model.
  • the input data of the pre-established double-circulation neural network model based on a gated recurring unit includes a multi-dimensional vector composed of multiple index factors explaining the trend of the Shanghai and Shenzhen 300 index at the opening time of a certain day, as well as several preceding and following numbers.
  • the output of the hidden layer is a feature vector abstracted from the multiple multi-dimensional vectors mentioned above, covering the data dependency relationship of the index factors at adjacent time points, using this feature vector What we get is still the forecast value of the CSI 300 stock index returns for the current day, but this forecast value takes into account the influence of the index factor data of adjacent trading days on the stock index returns for the current day, and it is often closer to the true value of the stock index returns for the current day.
  • Step S302 the prediction module 130 collects data of the n index factors at a second preset time point, constructs a second n-dimensional vector at the second preset time point according to the collected data, and inputs the second n-dimensional vector into the The stock index prediction model is described, and the stock index return rate at the second preset time point is predicted.
  • the prediction module 130 will also collect several of the n index factors at the second preset time point (usually less than 30). Data) at adjacent preset time points, input the n-dimensional vectors constructed from these data into the stock index prediction model, then the feature vectors of the n-dimensional vectors at the preset time points output by the hidden layer will cover the adjacent The data dependency of the index factor at a preset time point, and using the feature vector as the input of the model output layer to predict the stock index return rate at the second preset time point will be more accurate.
  • an embodiment of the present application further provides a computer-readable storage medium, where the computer-readable storage medium may be a hard disk, a multimedia card, an SD card, a flash memory card, an SMC, a read-only memory (ROM), and an erasable and programmable memory. Any one or any combination of read-only memory (EPROM), portable compact disc read-only memory (CD-ROM), USB memory, etc.
  • the computer-readable storage medium may be a hard disk, a multimedia card, an SD card, a flash memory card, an SMC, a read-only memory (ROM), and an erasable and programmable memory. Any one or any combination of read-only memory (EPROM), portable compact disc read-only memory (CD-ROM), USB memory, etc.
  • the computer-readable storage medium includes a stock index prediction program, and when the stock index prediction program is executed by a processor, the following steps are implemented:
  • Sample acquisition step selecting a time series of n index factors in a preset time period, obtaining a stock index return rate at each first preset time point in the preset time period, and according to the time series of the n index factors, Constructing a first n-dimensional vector at each first preset time point;
  • Model training step using the plurality of first n-dimensional vectors at a first preset time point and the stock index return rate as sample data, using a back-propagation algorithm to perform a pre-established dual-recurrent neural network model based on a gated recurrent unit Training, iteratively updating model weights to obtain stock index prediction models;
  • Prediction step collect data of the n index factors at a second preset time point, construct a second n-dimensional vector at a second preset time point based on the collected data, and input the second n-dimensional vector into the stock index prediction The model predicts the stock index return rate at the second preset time point.

Landscapes

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

Abstract

一种股指预测方法、装置及存储介质。该方法包括：选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量（S300）；以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，对基于门控循环单元的双循环神经网络模型进行训练，得到股指预测模型（S301）；采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率（S302）。利用本方法，可以提取指数因子的深层次特征，提高股指预测准确度。

Description

股指预测方法、装置及存储介质
优先权申明
本申请要求于2018年8月21日提交中国专利局、申请号为201810954505.X，发明名称为“股指预测方法、装置及存储介质”的中国专利申请的优先权，其内容全部通过引用结合在本申请中。
技术领域
本申请涉及计算机技术和智能预测技术领域，尤其涉及一种股指预测方法、装置及计算机可读存储介质。
背景技术
股票市场作为风险和收益双高的投资市场，对其进行预测从而更好地进行选股和择时以获得最大收益是研究者密切关注的领域。沿用至今的预测分析方法可以分为两类：基本面分析法与技术分析法。基本面分析法着手点在于国家经济政策与公司的基本面等信息，而技术分析方法则侧重利用历史数据带入数学模型或机器中来训练和演算。其中，基本面分析法中应用到的国家宏观经济状况、企业盈利状况等因素均较难定量，且大多属于长期性因素，所以如果只应用基本面分析法预测的精度将很不准确。而技术分析法主要是应用客观的量化指标进行预测，较为常用的方法包括时间序列法、小波分析法、神经网络法等，但由于股市的限制以及股票、期货等自身的特性，例如各影响因素大多不具有明显的关联关系，当前诸多方法并不能在预测时达到较好的效果，比如单独使用时间序列预测方法难以衡量多种指数因子的相互作用，在处理非线性特征数据时很难得到很好的预测结果，而神经网络当前较为流行的支持向量机预测模型只对求解小样本的非线性问题具有优秀的泛化能力，难以准确地预测股指走势。
发明内容
鉴于以上原因，本申请提供一种股指预测方法、电子装置及计算机可读存储介质，其主要目的在于从多种指数因子中提取深层次的特征向量，从而 科学、准确地预测股指收益率。
为实现上述目的，本申请提供的股指预测方法包括如下步骤：
样本获取步骤：选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量；
模型训练步骤：以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型；及
预测步骤：采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。
本申请还提供一种电子装置，该电子装置包括存储器和处理器，所述存储器中包括股指预测程序，该股指预测程序被所述处理器执行时实现如下步骤：
样本获取步骤：选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量；
模型训练步骤：以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型；及
预测步骤：采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。
此外，为实现上述目的，本申请还提供一种计算机可读存储介质，该计算机可读存储介质中包括股指预测程序，该股指预测程序被所述处理器执行时实现如下步骤：
样本获取步骤：选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量；
模型训练步骤：以所述多个第一预设时间点的第一n维向量及股指收益 率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型；及
预测步骤：采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。
本申请提出的股指预测方法、装置及计算机可读存储介质，通过选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量，然后以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型，最后采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。因为基于门控循环单元的双循环神经网络模型可以从各种指数因子中提取深层次特征，所以利用本申请可以准确地对股指进行预测。
附图说明
图1为本申请电子装置一实施例的示意图；
图2为图1中股指预测程序一实施例的程序模块图；
图3为本申请股指预测方法一实施例的流程图。
本申请目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
为了使本申请的目的、技术方案和优点更加清楚明白，下面将结合若干附图及实施例，对本申请进行进一步的详细说明。应当理解的是，此处所描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所 有其他实施例，都属于本申请保护的范围。
本申请提供一种电子装置。参照图1所示，为本申请电子装置1一实施例的示意图。在该实施例中，电子装置1通过获取样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型，然后利用该股指预测模型对预设时间点的股指收益率进行预测。
所述电子装置1可以是服务器、智能手机、平板电脑、便携计算机、桌上型计算机等具有存储和运算功能的终端设备。在一个实施例中，当电子装置1为服务器时，该服务器可以是机架式服务器、刀片式服务器、塔式服务器或机柜式服务器等的一种或几种。
所述电子装置1包括存储器11、处理器12及网络接口13。
其中，所述存储器11包括至少一种类型的可读存储介质。所述至少一种类型的可读存储介质可为如闪存、硬盘、多媒体卡、卡型存储器等的非易失性存储介质。在一些实施例中，所述可读存储介质可以是所述电子装置1的内部存储单元，例如该电子装置1的硬盘。在另一些实施例中，所述可读存储介质也可以是所述电子装置1的外部存储器11，例如所述电子装置1上配备的插接式硬盘，智能存储卡(Smart Media Card，SMC)，安全数字(Secure Digital，SD)卡，闪存卡(Flash Card)等。
在本实施例中，所述存储器11的可读存储介质通常用于存储操作系统、股指预测程序10、预先建立的基于门控循环单元的双循环神经网络模型以及采集到的各种指数因子数据等。所述存储器11还可以用于暂时地存储已经输出或者将要输出的数据。
所述处理器12在一些实施例中可以是一中央处理器(Central Processing Unit，CPU)，微处理器或其他数据处理芯片，用于运行存储器11中存储的程序代码或处理数据，例如执行所述股指预测程序10等。
所述网络接口13可以包括标准的有线接口、无线接口(如WI-FI接口)。通常用于在该电子装置1与其他电子设备或系统之间建立通信连接。
图1仅示出了具有组件11-13以及所述股指预测程序10的电子装置1，但是应理解的是，并不要求实施所有示出的组件，可以替代的实施更多或者更少的组件。
可选地，该电子装置1还可以包括显示器，也可以称为显示屏或显示单元。在一些实施例中可以是LED显示器、液晶显示器、触控式液晶显示器以及有机发光二极管(Organic Light-Emitting Diode，OLED)显示器等。显示器用于显示在该电子装置1中处理的信息以及用于显示可视化的用户界面。
可选地，该电子装置1还包括触摸传感器。所述触摸传感器所提供的供用户进行触摸操作的区域称为触控区域。此外，这里所述的触摸传感器可以为电阻式触摸传感器、电容式触摸传感器等。而且，所述触摸传感器不仅包括接触式的触摸传感器，也可包括接近式的触摸传感器等。此外，所述触摸传感器可以为单个传感器，也可以为例如阵列布置的多个传感器。用户可以通过触摸所述触控区域启动所述股指预测程序10。
此外，该电子装置1的显示器的面积可以与所述触摸传感器的面积相同，也可以不同。可选地，将显示器与所述触摸传感器层叠设置，以形成触摸显示屏。该装置基于触摸显示屏侦测用户触发的触控操作。
该电子装置1还可以包括射频(Radio Frequency，RF)电路、传感器和音频电路等等，在此不再赘述。
在上述实施例中，处理器12执行存储器11中存储的股指预测程序10时实现如下步骤：
样本获取步骤：选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量；
模型训练步骤：以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型；及
预测步骤：采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。
关于上述步骤的详细介绍，请参照下述图2关于股指预测程序10实施例的程序模块图以及图3关于股指预测方法实施例的流程图的说明。
在其他实施例中，所述股指预测程序10可以被分割为多个模块，该多个模块被存储于存储器12中，并由处理器13执行，以完成本申请。本申请所称的模块是指能够完成特定功能的一系列计算机程序指令段。
参照图2所示，为图1中股指预测程序10一实施例的程序模块图。在本实施例中，所述股指预测程序10可以被分割为：样本获取模块110、预处理模块120、模型训练模块130以及预测模块140。其中：
所述样本获取模块110，用于选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量。例如，从预设的数据平台选取近两个月解释沪深300指数走势的13种指数因子的时间序列，根据选取的时间序列为每个第一预设时间点(例如每日开盘时间)构造一个13维向量，获取沪深300股指期货每日收益率作为所述每个第一预设时间点的股指收益率。其中，所述13种指数因子可以是中债企业债到期收益率(AAA)：10年、风险溢价、股息率、慢速KD指标(SlowKD)、平滑异同移动平均线直方图(Moving Average Convergence and Divergence Histogram，MACD Histogram)、布林线指标(Bollinger Bands)、移动平均线-相对强弱指标(MA of RSI(14)[m＝22])、4-period MA of 4week MA of modified OBV-(MA4*4)、CR指标、大小盘换手率比值、RSRS指标、溢价率、主动买入额等。在其他实施例中，所述指数因子也可以只包括上述指数因子中的一种或几种，还可以为其他影响股指走势的统计数据，在此不再赘述。
所述预处理模块120，用于对所述时间序列的指数因子数据进行预处理，并利用拉格朗日插值法对所述指数因子数据中的缺失值进行修补。所述预处理包括去除噪声，以及利用拉格朗日插值法对数据中的缺失值进行插补调整。例如，在利用指数因子的时间序列构造多维向量时，可能出现指数因子数据获取失败的情况，此时可利用拉格朗日插值法补齐构造多维向量需要的获取失败的指数因子数据。
所述模型训练模块130，用于以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型。其中，所述预先建立的基于门控循环单元的双循环神经网络模型包括两个隐 含层，所述隐含层用于将所述n维向量抽象为预设维度(例如，二维)的特征向量，每个所述隐含层的神经元个数与所述特征向量的预设维度的数值相等。所述反向传播算法的公式为：
Figure PCTCN2018123583-appb-000001
其中，a′ k代表更新后的第k个模型权重，a k代表更新前的第k个模型权重，η代表学习速率，0.1<η<3，
Figure PCTCN2018123583-appb-000002
代表整体误差E对a k的偏导值。
当偏置值
Figure PCTCN2018123583-appb-000003
达到预设值时，确定模型权重，结束模型训练，得到训练好的股指预测模型。
依上述例子，假设将某日开盘时间的解释沪深300指数走势的13种指数因子构成的13维向量输入所述预先建立的基于门控循环单元的双循环神经网络模型，该模型隐含层的输出结果即为由所述13维向量抽象出的特征向量，该特征向量反映出指数因子数据的深层次特征，利用该特征向量，可以计算得到当日沪深300股指收益率的预测值。在本实施例中，样本获取模块110获取的股指收益率为股指收益率的真实值，股指收益率真实值的计算公式可以为：股指收益率＝(当日收盘价-当日开盘价)/当日开盘价。假设某日股指收益率的预测值为P，该日股指收益率的真实值为T，则整体误差E＝(P-T)^2，将整体误差E与待更新模型权重的偏导值乘以学习速率，得到偏置值，计算待更新模型权重与所述偏置值的差，得到更新后的模型权重，迭代更新模型权重，直到偏置值达到预设值，确定模型权重，得到训练好的股指预测模型。
所述预测模块140，用于采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。例如，采集解释沪深300指数走势的13种指数因子在今日开盘时间点的数据，根据采集的数据构造今日开盘时间点的13维向量，将该13维向量输入所述模型训练模块130训练得到的股指预测模型，预测得到沪深300指数今日的股指收益率。
此外，本申请还提供一种股指预测方法。参照图3所示，为本申请股指预测方法一实施例的流程图。电子装置1的处理器12执行存储器中存储的股指预测程序10时实现股指预测方法的如下步骤：
步骤S300，样本获取模块110选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量。例如，从预设的数据平台选取近两个月的解释沪深300指数走势的多种指数因子的时间序列，根据选取的时间序列为每个第一预设时间点(例如每日开盘时间)构造一个多维向量，获取沪深300股指期货每日收益率作为所述每个第一预设时间点的股指收益率。其中，所述多种指数因子可以包括中债企业债到期收益率(AAA)：10年、风险溢价、股息率、慢速KD指标(SlowKD)、MACD Histogram、溢价率、主动买入额等的一种或多种。其中，通常来说，风险溢价与股息率呈负相关性，当股市的收益率高于债市时，资金将从债市流向股市；反之，当股指收益率低于债市时，资金将从股市流向债市。沪深300股息率与沪深300股指收盘价呈负相关性。当SlowKD<10时，超卖信号出现，沪深300股指可能上升；当SlowKD>10时，超卖信号消失，沪深300股指可能下降。当MACD Histogram由负转正，沪深300股指做多信号出现；MACD由正转负，沪深300股指做空信号出现。溢价率>0，市场乐观；溢价率<0，市场悲观；历史上，溢价率触及5时，为卖出信号；溢价率触及-1时，为买入信号。以上仅为在理论上对部分指数因子做的简单介绍，因为股指走势受投资者的心理影响非常大，所以指数因子与股指之间存在很高的非线性度，用单一的指数因子或结合多种指数因子解释股指走势常常并不具有很强的说服力。本申请选择对基于门控循环单元的双循环神经网络模型进行训练，正是为了从指数因子中提取不易被察觉的更深层次的特征，以此提高股指预测准确度。
步骤S301，模型训练模块120以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型。其中，所述预先建立的基于门控循环单元的双循环神经网络模型包括两个隐含层，所述隐含层用于将所述n维向量抽象为预设维度(例如，二维)的特征向量，每个所述隐含层的神经元个数与所述特征向量的预设维度的数值相等。所述反向传播算法的公式为：
Figure PCTCN2018123583-appb-000004
其中，a′ k代表更新后的第k个模型权重，a k代表更新前的第k个模型权重，η代表学习速率，0.1<η<3，
Figure PCTCN2018123583-appb-000005
代表整体误差E对a k的偏导值。当偏置值
Figure PCTCN2018123583-appb-000006
达到预设值时，确定模型权重，结束模型训练，得到股指预测模型。
例如，将某日开盘时间的解释沪深300指数走势的多种指数因子构成的多维向量输入所述预先建立的基于门控循环单元的双循环神经网络模型，该模型隐含层的输出结果即为由所述多维向量抽象出的特征向量，该特征向量反映出指数因子数据的深层次特征，利用该特征向量，可以计算得到当日沪深300股指收益率的预测值。假设某日股指收益率的预测值为P，样本获取模块110获取的该日股指收益率的真实值为T，则所述整体误差E＝(P-T)^2，将整体误差E与待更新模型权重的偏导值乘以学习速率，得到偏置值，计算待更新模型权重与所述偏置值的差，得到更新后的模型权重，迭代更新模型权重，直到偏置值达到预设值，确定模型权重，得到训练好的股指预测模型。
在一个实施例中，输入所述预先建立的基于门控循环单元的双循环神经网络模型的数据包括某日开盘时间的解释沪深300指数走势的多种指数因子构成的多维向量以及前后若干个(通常小于30个)交易日的多维向量，所述隐含层的输出结果为由上述多个多维向量抽象出的特征向量，涵盖了相邻时间点指数因子的数据依赖关系，利用该特征向量得到的仍是当日沪深300股指收益率的预测值，但该预测值考虑到了相邻交易日的指数因子数据对当日股指收益率的影响，往往更加接近当日股指收益率的真实值。
步骤S302，预测模块130采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。
为了体现基于门控循环单元的双循环神经网络模型的优势，在一个实施例中，预测模块130还将采集所述n种指数因子在所述第二预设时间点的若干个(通常小于30个)相邻预设时间点的数据，将由这些数据构造的n维向量一并输入所述股指预测模型，则隐含层输出的所述预设时间点n维向量的特征向量将涵盖相邻预设时间点指数因子的数据依赖关系，将该特征向量作为模型输出层的输入预测得到该第二预设时间点的股指收益率将更加准确。
此外，本申请实施例还提出一种计算机可读存储介质，所述计算机可读 存储介质可以是硬盘、多媒体卡、SD卡、闪存卡、SMC、只读存储器(ROM)、可擦除可编程只读存储器(EPROM)、便携式紧致盘只读存储器(CD-ROM)、USB存储器等等中的任意一种或者几种的任意组合。
所述计算机可读存储介质中包括股指预测程序，所述股指预测程序被处理器执行时实现如下步骤：
样本获取步骤：选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量；
模型训练步骤：以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型；及
预测步骤：采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。
本申请之计算机可读存储介质的具体实施方式与上述股指预测方法和电子装置1的具体实施方式大致相同，请参相关介绍，在此不再赘述。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、装置、物品或者方法不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、装置、物品或者方法所固有的要素。另外，各个实施例之间的技术方案可以相互结合，但是必须是以本领域普通技术人员能够实现为基础，当技术方案的结合出现相互矛盾或无法实现时应当认为这种技术方案的结合不存在，也不在本申请要求的保护范围之内。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的 技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在如上所述的一个存储介质中，包括若干指令用以使得电子装置执行本申请各个实施例所述的方法。
以上仅为本申请的优选实施例，并非因此限制本申请的专利范围，凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种股指预测方法，应用于电子装置，其特征在于，所述方法包括：
    样本获取步骤：选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量；
    模型训练步骤：以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型；及
    预测步骤：采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。
  2. 如权利要求1所述的股指预测方法，其特征在于，该方法还包括：
    对所述时间序列的指数因子数据进行预处理，利用拉格朗日插值法对所述指数因子数据中的缺失值进行修补。
  3. 如权利要求1所述的股指预测方法，其特征在于，所述预先建立的基于门控循环单元的双循环神经网络模型包括两个隐含层，所述隐含层用于将所述n维向量抽象为预设维度的特征向量，每个所述隐含层的神经元个数与所述特征向量的预设维度的数值相等。
  4. 如权利要求2所述的股指预测方法，其特征在于，所述预先建立的基于门控循环单元的双循环神经网络模型包括两个隐含层，所述隐含层用于将所述n维向量抽象为预设维度的特征向量，每个所述隐含层的神经元个数与所述特征向量的预设维度的数值相等。
  5. 如权利要求4所述的股指预测方法，其特征在于，所述预处理还包括去除噪声处理。
  6. 如权利要求1所述的股指预测方法，其特征在于，所述反向传播算法的公式为：
    Figure PCTCN2018123583-appb-100001
    其中，a′ k代表更新后的第k个模型权重，a k代表更新前的第k个模型权重，η代表学习速率，0.1<η<3，
    Figure PCTCN2018123583-appb-100002
    代表整体误差E对a k的偏导值。
  7. 如权利要求6所述的股指预测方法，其特征在于，当
    Figure PCTCN2018123583-appb-100003
    达到预设 值时，确定模型权重，结束所述模型训练步骤，得到所述股指预测模型。
  8. 一种电子装置，包括存储器和处理器，其特征在于，所述存储器中包括股指预测程序，所述股指预测程序被所述处理器执行时实现如下步骤：
    样本获取步骤：选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量；
    模型训练步骤：以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型；及
    预测步骤：采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。
  9. 如权利要求8所述的电子装置，其特征在于，所述股指预测程序被所述处理器执行时实现的步骤还包括：
    对所述时间序列的指数因子数据进行预处理，将缺失值用拉格朗日插值法进行修补。
  10. 如权利要求8所述的电子装置，其特征在于，所述预先建立的基于门控循环单元的双循环神经网络模型包括两个隐含层，所述隐含层用于将所述n维向量抽象为预设维度的特征向量，每个所述隐含层的神经元个数与所述特征向量的预设维度的数值相等。
  11. 如权利要求9所述的电子装置，其特征在于，所述预先建立的基于门控循环单元的双循环神经网络模型包括两个隐含层，所述隐含层用于将所述n维向量抽象为预设维度的特征向量，每个所述隐含层的神经元个数与所述特征向量的预设维度的数值相等。
  12. 如权利要求11所述的电子装置，其特征在于，所述预处理还包括去除噪声处理。
  13. 如权利要求8所述的电子装置，其特征在于，所述反向传播算法的公式为：
    Figure PCTCN2018123583-appb-100004
    其中，a′ k代表更新后的第k个模型权重，a k代表更新前的第k个模型权 重，η代表学习速率，0.1<η<3，
    Figure PCTCN2018123583-appb-100005
    代表整体误差E对a k的偏导值。
  14. 如权利要求13所述的电子装置，其特征在于，当
    Figure PCTCN2018123583-appb-100006
    达到预设值时，确定模型权重，结束所述模型训练步骤，得到所述股指预测模型。
  15. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质中包括股指预测程序，所述股指预测程序被执行时实现如下步骤：
    样本获取步骤：选取预设时间段的n种指数因子的时间序列，获取该预设时间段的每个第一预设时间点的股指收益率，并根据所述n种指数因子的时间序列为每个第一预设时间点构造一个第一n维向量；
    模型训练步骤：以所述多个第一预设时间点的第一n维向量及股指收益率为样本数据，利用反向传播算法对预先建立的基于门控循环单元的双循环神经网络模型进行训练，迭代更新模型权重，得到股指预测模型；及
    预测步骤：采集所述n种指数因子在第二预设时间点的数据，根据采集的数据构造第二预设时间点的第二n维向量，将该第二n维向量输入所述股指预测模型，预测得到该第二预设时间点的股指收益率。
  16. 如权利要求15所述的计算机可读存储介质，其特征在于，所述股指预测程序被所述处理器执行时实现的步骤还包括：
    对所述时间序列的指数因子数据进行预处理，将缺失值用拉格朗日插值法进行修补。
  17. 如权利要求15所述的计算机可读存储介质，其特征在于，所述预先建立的基于门控循环单元的双循环神经网络模型包括两个隐含层，所述隐含层用于将所述n维向量抽象为预设维度的特征向量，每个所述隐含层的神经元个数与所述特征向量的预设维度的数值相等。
  18. 如权利要求16所述的计算机可读存储介质，其特征在于，所述预先建立的基于门控循环单元的双循环神经网络模型包括两个隐含层，所述隐含层用于将所述n维向量抽象为预设维度的特征向量，每个所述隐含层的神经元个数与所述特征向量的预设维度的数值相等。
  19. 如权利要求15所述的计算机可读存储介质，其特征在于，所述反向传播算法的公式为：
    Figure PCTCN2018123583-appb-100007
    其中，a′ k代表更新后的第k个模型权重，a k代表更新前的第k个模型权 重，η代表学习速率，0.1<η<3，
    Figure PCTCN2018123583-appb-100008
    代表整体误差E对a k的偏导值。
  20. 如权利要求19所述的计算机可读存储介质，其特征在于，当
    Figure PCTCN2018123583-appb-100009
    达到预设值时，确定模型权重，结束所述模型训练步骤，得到所述股指预测模型。
PCT/CN2018/123583 2018-08-21 2018-12-25 股指预测方法、装置及存储介质 WO2020037922A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810954505.XA CN109214560A (zh) 2018-08-21 2018-08-21 股指预测方法、装置及存储介质
CN201810954505.X 2018-08-21

Publications (1)

Publication Number Publication Date
WO2020037922A1 true WO2020037922A1 (zh) 2020-02-27

Family

ID=64988948

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/123583 WO2020037922A1 (zh) 2018-08-21 2018-12-25 股指预测方法、装置及存储介质

Country Status (2)

Country Link
CN (1) CN109214560A (zh)
WO (1) WO2020037922A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111785018A (zh) * 2020-06-18 2020-10-16 重庆大学 一种基于门控循环单元的收费站下道流量预测方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108154435A (zh) * 2017-12-26 2018-06-12 浙江工业大学 一种基于循环神经网络的股指价格预测方法
CN108229750A (zh) * 2018-01-17 2018-06-29 河南理工大学 一种股票收益率预测方法
CN108241872A (zh) * 2017-12-30 2018-07-03 北京工业大学 基于多特征因子的隐马尔科夫模型的自适应股票预测方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108154435A (zh) * 2017-12-26 2018-06-12 浙江工业大学 一种基于循环神经网络的股指价格预测方法
CN108241872A (zh) * 2017-12-30 2018-07-03 北京工业大学 基于多特征因子的隐马尔科夫模型的自适应股票预测方法
CN108229750A (zh) * 2018-01-17 2018-06-29 河南理工大学 一种股票收益率预测方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111785018A (zh) * 2020-06-18 2020-10-16 重庆大学 一种基于门控循环单元的收费站下道流量预测方法

Also Published As

Publication number Publication date
CN109214560A (zh) 2019-01-15

Similar Documents

Publication Publication Date Title
US20180150783A1 (en) Method and system for predicting task completion of a time period based on task completion rates and data trend of prior time periods in view of attributes of tasks using machine learning models
US10614073B2 (en) System and method for using data incident based modeling and prediction
US11704682B2 (en) Pre-processing financial market data prior to machine learning training
US11875408B2 (en) Techniques for accurate evaluation of a financial portfolio
US11151418B2 (en) Interactive modeling application adapted for execution via distributed computer-based systems
CN108985501B (zh) 基于指数特征提取的股指预测方法、服务器及存储介质
CN115130894A (zh) 基于人工智能的生产规划方法、装置、计算机设备及介质
CN117235608B (zh) 风险检测方法、装置、电子设备及存储介质
WO2020037922A1 (zh) 股指预测方法、装置及存储介质
CN113435900A (zh) 交易风险确定方法、装置和服务器
CN112686470A (zh) 电网饱和负荷预测方法、装置及终端设备
WO2020199483A1 (zh) 金融数据的图像处理方法、装置、设备及计算机可读存储介质
CN111860998A (zh) 股票交易数据的预测方法、装置、电子设备及介质
WO2019242143A1 (zh) 股票卖出预警方法、装置及计算机可读存储介质
Balder et al. Primal–dual linear Monte Carlo algorithm for multiple stopping—an application to flexible caps
CN115827994A (zh) 一种数据处理方法、装置、设备、存储介质
CN108765153A (zh) 交易机器人金融数据计算方法及装置
CN110210959A (zh) 财务数据的分析方法、装置及存储介质
CN110598918A (zh) 一种水产品价格预测方法、装置及计算机存储介质
CN116883181B (zh) 基于用户画像的金融服务推送方法、存储介质及服务器
CN117436550B (zh) 推荐模型训练方法及装置
CN111507366B (zh) 推荐概率模型的训练方法、智能补全方法及相关装置
CN117522565A (zh) 贷款风险等级的确定方法、装置、电子设备及存储介质
CN115640461A (zh) 一种产品的推荐方法、装置、电子设备及其存储介质
CN113837861A (zh) 基于用户分组的催收方法、装置、存储介质及设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18930592

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18930592

Country of ref document: EP

Kind code of ref document: A1