WO2019205378A1 - 基于舆情因子的投资选股方法、装置及存储介质 - Google Patents

基于舆情因子的投资选股方法、装置及存储介质 Download PDF

Info

Publication number
WO2019205378A1
WO2019205378A1 PCT/CN2018/102127 CN2018102127W WO2019205378A1 WO 2019205378 A1 WO2019205378 A1 WO 2019205378A1 CN 2018102127 W CN2018102127 W CN 2018102127W WO 2019205378 A1 WO2019205378 A1 WO 2019205378A1
Authority
WO
WIPO (PCT)
Prior art keywords
factor
stock
information coefficient
stocks
prediction model
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.)
Ceased
Application number
PCT/CN2018/102127
Other languages
English (en)
French (fr)
Inventor
李正洋
毛小豪
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 Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co 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 Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Publication of WO2019205378A1 publication Critical patent/WO2019205378A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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

Definitions

  • the present application relates to the field of financial big data mining, and in particular, to an investment stock selection method based on a public opinion factor, an electronic device, and a computer readable storage medium.
  • the sensation factor of the stock (for example, the news heat and the public's emotional attitude towards the corresponding news) can be expressed to some extent as the market potential energy of the stock, when the stock's sentiment factor is abnormal (for example, when a stock has a significant negative
  • the investment strategy of the stock needs to be changed immediately.
  • the threshold is set in advance for a certain indicator of the stock, and then the stock with the weighted investment indicator meeting the threshold condition is selected, or the investment strategy is set according to the market experience of the professional investor. This method has high requirements for the professional skills of decision-making investors, and is easy to make mistakes, which may increase the risk of misuse.
  • the present application provides an investment stock selection method based on a public opinion factor, an electronic device and a computer readable storage medium, the main purpose of which is to visually display the influence of the sensation factor on the future earnings of the stock by calculating the stock score, and select the stock with a high score as the stock.
  • the present application provides an investment stock selection method based on a public opinion factor, the method comprising:
  • the first information coefficient of each sentiment factor is input into the pre-trained information coefficient prediction model, and the second information coefficient of each future public opinion factor is predicted;
  • the plurality of stocks are sorted according to the order of the scores, and the first predetermined number of stocks ranked first is selected as the target investment stocks.
  • the present application further provides an electronic device, including: a memory, a processor, and the memory stocking program based on a public opinion factor stored in the memory, and the investment stock selection based on the public opinion factor
  • the program implements the following steps when executed by the processor:
  • the first information coefficient of each sentiment factor is input into the pre-trained information coefficient prediction model, and the second information coefficient of each future public opinion factor is predicted;
  • the plurality of stocks are sorted according to the order of the scores, and the first predetermined number of stocks ranked first is selected as the target investment stocks.
  • the present application further provides a computer readable storage medium, wherein the computer readable storage medium stores an investment stock selection program based on a public opinion factor, and the investment factor selection program based on the public opinion factor is processed The steps of implementing the syllabus-based investment stock selection method as described above when executed.
  • the essay factor-based investment stock selection method, the electronic device and the computer readable storage medium proposed by the present application calculate the first information coefficient of each stock sensation factor according to the stock sentiment factor observation value and the yield rate data.
  • the first information coefficient and the information coefficient prediction model are used to predict the second information coefficient, and the ability of each lyric factor to predict the future earnings of the stock is reflected; the enthusiasm factors are weighted according to the second information coefficient, and the dynamics of each lyric factor are realized.
  • Empowerment increases the reliability of each lyric factor; calculates the stock score based on the observations of each lyric factor and the corresponding weights, selects the stock with higher score as the target investment stock, improves the authenticity of the stock score, and visually displays the sensation
  • the influence of factor observations on the future earnings of stocks helps investors control risks and increase investment returns.
  • FIG. 1 is a schematic diagram of a preferred embodiment of an electronic device of the present application.
  • FIG. 2 is a program module diagram of the investment stock selection procedure based on the sensation factor in FIG. 1;
  • FIG. 3 is a flow chart of a preferred embodiment of the investment stock selection method based on the public opinion factor of the present application.
  • the application provides an electronic device 1 .
  • FIG. 1 it is a schematic diagram of a preferred embodiment of the electronic device 1 of the present application.
  • the electronic device 1 includes a memory 11, a processor 12, a network interface 13, and a communication bus 14.
  • the communication bus 14 is used to implement connection communication between these components.
  • the network interface 13 may include a standard wired interface, a wireless interface (such as a WI-FI interface).
  • 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, or 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 an external storage device of the electronic device 1, such as a plug-in hard disk equipped on the electronic device 1, a smart memory card (SMC). , Secure Digital (SD) card, Flash Card, etc.
  • SMC smart memory card
  • SD Secure Digital
  • the readable storage medium of the memory 11 is generally used to store the essay factor-based investment stock selection program 10 or the like installed in the electronic device 1.
  • the memory 11 can also be used to temporarily store data that has been output or is about to be output.
  • the processor 12 in some embodiments, may be a Central Processing Unit (CPU), microprocessor or other data processing chip for running program code or processing data stored in the memory 11, such as performing a lyric based factor
  • the investment stock selection process is 10 and so on.
  • Figure 1 shows only the electronic device 1 with components 11-14 and the essay factor based investment stock selection program 10, but it should be understood that not all illustrated components may be implemented, alternative implementations may be more or more Less components.
  • the electronic device 1 may further include a user interface
  • the user interface may include a display, an input unit such as a keyboard, and the optional user interface may further include a standard wired interface and a wireless interface.
  • the electronic device 1 may further include a display, and in some embodiments, an LED display, a liquid crystal display, a touch liquid crystal display, and an Organic Light-Emitting Diode (OLED) touch device.
  • the display is used to display information processed in the electronic device and a user interface for displaying visualizations.
  • a memory 11 as a computer storage medium includes a public opinion factor based investment stock selection program 10, and the processor 12 executes a public opinion factor based investment stock selection program 10 stored in the memory 11. The following steps are implemented:
  • the stock sensation factor includes the sensational heat factor and the sensational emotion factor, so the lyric factor observation value of each stock is the lyric heat observation value and the lyric emotion observation value in the first preset time.
  • the user can set the first preset time and the second preset time as needed. It can be understood that N target stocks are selected from the pre-agreed M stocks.
  • the stock picking frequency may be performed once every other week or once every other month. In the examples, it is performed every other week.
  • the stock picking time is time t
  • the first preset time may be set to time t-1 (ie, the previous period of the stock picking period)
  • the second preset time may be set to time t (ie, During the current period of stock selection, the interval between each two periods is one week.
  • the corresponding yield of the M stocks in the second preset time is the weekly yield
  • the closing price of the M stocks at intervals of one week is used, specifically , the calculation formula is:
  • R is a stock week yield at time t (current period) is
  • P t is the closing stock price at time t (current period) is
  • P t-1 for the stock (previous period) of the closing price of the time t-1.
  • the information coefficient (IC) corresponding to the sensation factor at the second preset time.
  • the first information coefficient refers to an information coefficient of each sentiment factor of the stock at time t (current period).
  • the Spellman correlation coefficient (Rank IC) is used as the information coefficient of each lyric factor of the stock.
  • Rank IC that is, the cross-correlation coefficient between the ranking of all stocks on a certain factor and the ranking of all stock returns in the next period at a certain time, the calculation formula is:
  • X is the observed value of the estrous factor f at time t-1 (previous period);
  • Y is the yield of the stock at time t (current period).
  • the IC values of the lyric heat factor and the lyric emotion factor of the M stocks at time t (current period), that is, the first IC value, are respectively calculated.
  • the first information coefficient of each sentiment factor is input into the pre-trained information coefficient prediction model, and the second information coefficient of each future public opinion factor is predicted;
  • the IC value of each lyric factor can well reflect the predictive ability of each lyric factor.
  • the larger the IC value the stronger the predictive ability of the factor in the stock return.
  • the second information coefficient refers to an IC value of each stock factor of the stock at time t+1 (next period).
  • the first IC value of the enthusiasm heat factor and the lyric sentiment factor of the M stocks are input into a predetermined information coefficient prediction model, and the predicted M only The IC value of the stock's sentiment heat factor and the lyric emotion factor at time t+1 (the next period), that is, the second IC value.
  • the first IC values of the lyric heat factor and the lyric emotion factor of the M stocks are IC At and IC Bt respectively , and are input into the information coefficient prediction model in the form of [IC At , IC Bt ], and the model outputs
  • the result is [IC A(t+1) , IC B(t+1) ], where IC A(t+1) and IC B(t+1) are the lyric heat factors and lyric emotions of the M stocks, respectively.
  • the second IC value of the factor is [IC A(t+1) , IC B(t+1) ]
  • the predetermined information coefficient prediction model is obtained by training a neural network. Since the input data of the model is only the IC value of each lyric factor at a certain moment, and the data form is relatively simple, it is selected to use a three-layer feedforward neural network with a hidden layer for training. Specifically, the training steps of the information coefficient prediction model include:
  • the sample data is determined [I K , I K+1 ], where I K represents the IC value corresponding to each lyric factor at time K, and I K+1 represents the IC value corresponding to each lyric factor at time K+1 (ie, the actual value mentioned later) .
  • the time interval between every two moments may be one day or one week.
  • the sample data is the IC value of each sensation factor on each day; when the time interval between two moments is one week, the sample data needs to be collected time.
  • the data is separated by one week.
  • the data may be the average value of the IC values of the five turbulent days of each consecutive week, or the IC value of the last trading day of the five trading days of the consecutive week.
  • the sample data is divided into training sets, evaluation sets, and test sets based on cross-validation (for example, 70% of the sample data is used as the training set, 10% of the sample data is used as the evaluation set, and 20% of the sample data is used as the test. set).
  • the sample data of the training set is input into a three-layer feedforward neural network model, and the model is trained to initially determine model parameters; the test set is used to test the accuracy of the three-layer feedforward neural network model during the training process, And inputting the sample data of the test set into the trained three-layer feedforward neural network model to test the three-layer feedforward neural network model obtained by training, and the three-layer feedforward obtained by training If the neural network model satisfies the preset verification condition (for example, the model prediction accuracy is greater than or equal to less than the preset threshold), the training is completed, and the trained three-layer feedforward neural network model is set as the information coefficient prediction model.
  • the preset verification condition for example, the model prediction accuracy is greater than or equal to less than the preset threshold
  • this step can be refined into the following steps:
  • the weights corresponding to the lyric factors of the M stocks are respectively determined according to the second information coefficients corresponding to the lyric factors; and the scores of each stock are respectively calculated according to the observed values of the lyric factors and the weights of the lyric factors.
  • the information coefficient prediction model is used to predict the public opinion factor of the M stocks and the information coefficient of the public opinion factor for the next period is IC A(t+1) and IC B(t+1) , respectively.
  • the heat factor and the lyric emotion factor respectively correspond to the weights when calculating the stock score.
  • the sentiment heat factor is a positive factor, indicating that the higher the stock's heat of opinion observation, the higher the future earnings of the stock will be; the sentiment affective factor may be either a positive factor or a negative factor, which needs to be explained.
  • the value of the lyric emotion observation value is [-1, 1]. The higher the absolute value of the lyric emotion observation value, the stronger the positive (negative) emotion. When the lyric emotion observation value of a stock is negative, The future earnings of the stock have a negative impact.
  • the lyric sentiment factor is a negative factor; on the contrary, when the lyric sentiment observation value of a stock is positive, there is a positive impact on the future earnings of the stock. At this time, lyric emotion The factor is a positive factor.
  • the first weights corresponding to the public opinion heat factor and the lyric emotion factor are determined by the above steps as ⁇ A1 and ⁇ B1 respectively .
  • the first weight of the stock's lyric heat factor and the lyric emotional factor is returned.
  • the second weights ⁇ A2 and ⁇ B2 corresponding to the two factors of the stock are determined as the weights ⁇ A , ⁇ B corresponding to the two factors when calculating the stock score.
  • the calculation formula of ⁇ A and ⁇ B is:
  • the calculation formula for stocks is:
  • i the stock's lyric factor
  • i ⁇ A, B ⁇
  • ⁇ i the weight corresponding to the stock's sensation factor i
  • ⁇ i ⁇ A , ⁇ B ⁇
  • the plurality of stocks are sorted according to the order of the scores, and the first predetermined number of stocks ranked first is selected as the target investment stocks.
  • the score of the stock calculated by the above steps can reflect the trend of the stock in the future income to a certain extent, and the higher the score of the stock, the higher the future profit may be. Therefore, after obtaining the scores of each of the M stocks, the M stocks are sorted according to the order of the scores, and the first predetermined number (for example, N only) stocks with the highest scores are selected from the M stocks. As the target stock to be invested.
  • the investment portfolio including the target stock to be invested it is also necessary to separately determine the proportion of each target stock in the total investment amount, for example, to make an equal investment in the target stock, or to determine according to the score of the target stock. Corresponding proportions, stocks with high scores have a larger proportion, and stocks with lower scores have a smaller proportion.
  • the specific operations can be set according to the actual needs of investors. Finally, the final portfolio is determined based on the proportion of the target stock and the target stock.
  • the electronic device proposed by the above embodiment calculates the first information coefficient of each stocky factor of the stock according to the stock sentiment factor observation value and the yield rate data, and predicts the second information coefficient by using the first information coefficient and the information coefficient prediction model to reflect each public opinion factor.
  • the public opinion factor based investment stock selection program 10 may also be divided into one or more modules, one or more modules being stored in the memory 11 and processed by one or more The device 12 is executed to complete the application.
  • a module as referred to in this application refers to a series of computer program instructions that are capable of performing a particular function.
  • FIG. 2 it is a program module diagram of the investment stock selection program based on the sensation factor in FIG.
  • the investment stock selection program 10 based on the sentiment factor may be divided into: an acquisition module 110, an information coefficient calculation module 120, a prediction module 130, a score calculation module 140, and a stock selection module 150.
  • the functions or operational steps implemented by the modules 110-150 are similar to the above, and are not described in detail herein, by way of example, for example:
  • the obtaining module 110 is configured to obtain the sensation factor observation value of the predetermined plurality of stocks in the first preset time and the profit rate in the second preset time;
  • the information coefficient calculation module 120 is configured to calculate a first information coefficient of each public opinion factor according to the observation value and the profit rate of the sentiment factor;
  • the prediction module 130 is configured to input the first information coefficient of each sentiment factor into the pre-trained information coefficient prediction model, and predict the second information coefficient of each future sentiment factor;
  • the score calculation module 140 is configured to calculate the scores of the plurality of stocks according to the respective estrous factor observation values and the second information coefficients of the lyric factors;
  • the stock selection module 150 is configured to sort the plurality of stocks according to the order of the scores, and filter out the first predetermined number of stocks ranked first, as the target investment stocks.
  • the present application also provides an investment stock selection method based on public opinion factors.
  • FIG. 3 it is a flowchart of a preferred embodiment of the investment stock selection method based on the public opinion factor of the present application.
  • the method can be performed by a device that can be implemented by software and/or hardware.
  • the investment stock selection method based on the public opinion factor includes: steps S1-S5.
  • the stock sensation factor includes the sensational heat factor and the sensational emotion factor, so the lyric factor observation value of each stock is the lyric heat observation value and the lyric emotion observation value in the first preset time.
  • the user can set the first preset time and the second preset time as needed. It can be understood that N target stocks are selected from the pre-agreed M stocks.
  • the stock picking frequency may be performed once every other week or once every other month. In the examples, it is performed every other week.
  • the stock picking time is time t
  • the first preset time may be set to time t-1 (ie, the previous period of the stock picking period)
  • the second preset time may be set to time t (ie, During the current period of stock selection, the interval between each two periods is one week.
  • the corresponding yield of the M stocks in the second preset time is the weekly yield
  • the closing price of the M stocks at intervals of one week is used, specifically , the calculation formula is:
  • R is a stock week yield at time t (current period) is
  • P t is the closing stock price at time t (current period) is
  • P t-1 for the stock (previous period) of the closing price of the time t-1.
  • the information coefficient (IC) corresponding to the sensation factor at the second preset time.
  • the first information coefficient refers to an information coefficient of each sentiment factor of the stock at time t (current period).
  • the Spellman correlation coefficient (Rank IC) is used as the information coefficient of each lyric factor of the stock.
  • Rank IC that is, the cross-correlation coefficient between the ranking of all stocks on a certain factor and the ranking of all stock returns in the next period at a certain time, the calculation formula is:
  • X is the observed value of the estrous factor f at time t-1 (previous period);
  • Y is the yield of the stock at time t (current period).
  • the IC values of the lyric heat factor and the lyric emotion factor of the M stocks at time t (current period), that is, the first IC value, are respectively calculated.
  • the IC value of each lyric factor can well reflect the predictive ability of each lyric factor.
  • the larger the IC value the stronger the predictive ability of the factor in the stock return.
  • the second information coefficient refers to an IC value of each stock factor of the stock at time t+1 (next period).
  • the first IC value of the enthusiasm heat factor and the lyric sentiment factor of the M stocks are input into a predetermined information coefficient prediction model, and the predicted M only The IC value of the stock's sentiment heat factor and the lyric emotion factor at time t+1 (the next period), that is, the second IC value.
  • the first IC values of the lyric heat factor and the lyric emotion factor of the M stocks are IC At and IC Bt respectively , and are input into the information coefficient prediction model in the form of [IC At , IC Bt ], and the model outputs
  • the result is [IC A(t+1) , IC B(t+1) ], where IC A(t+1) and IC B(t+1) are the lyric heat factors and lyric emotions of the M stocks, respectively.
  • the second IC value of the factor is [IC A(t+1) , IC B(t+1) ]
  • the predetermined information coefficient prediction model is obtained by training a neural network. Since the input data of the model is only the IC value of each lyric factor at a certain moment, and the data form is relatively simple, it is selected to use a three-layer feedforward neural network with a hidden layer for training. Specifically, the training steps of the information coefficient prediction model include:
  • the sample data is divided into training set, evaluation set and test set, and the neural network is trained by using the sample data of the training set to obtain the information coefficient prediction model, and the accuracy of the information coefficient prediction model is tested by using the sample data of the test set.
  • the final information coefficient prediction model is obtained.
  • the information coefficient corresponding to each moment of the predetermined second predetermined number (for example, 3000) of the stocks in the third preset time (two years) is calculated.
  • the sample data [I K , I K+1 ] is determined, where I K represents the IC value corresponding to each lyric factor at time K, and I K+1 represents the IC value corresponding to each lyric factor at time K+1 (ie, , the actual value mentioned later).
  • the time interval between every two moments may be one day or one week.
  • the sample data is the IC value of each sensation factor on each day; when the time interval between two moments is one week, the sample data needs to be collected time.
  • the data is separated by one week.
  • the data may be the average value of the IC values of the five turbulent days of each consecutive week, or the IC value of the last trading day of the five trading days of the consecutive week.
  • the sample data is divided into training sets, evaluation sets, and test sets based on cross-validation (for example, 70% of the sample data is used as the training set, 10% of the sample data is used as the evaluation set, and 20% of the sample data is used as the test. set).
  • the sample data of the training set is input into a three-layer feedforward neural network model, and the model is trained to initially determine model parameters; the test set is used to test the accuracy of the three-layer feedforward neural network model during the training process, And inputting the sample data of the test set into the trained three-layer feedforward neural network model to test the three-layer feedforward neural network model obtained by training, and the three-layer feedforward obtained by training If the neural network model satisfies the preset verification condition (for example, the model prediction accuracy is greater than or equal to less than the preset threshold), the training is completed, and the trained three-layer feedforward neural network model is set as the information coefficient prediction model.
  • the preset verification condition for example, the model prediction accuracy is greater than or equal to less than the preset threshold
  • ANNs artificial neural networks
  • three-layer feedforward is based on backpropagation.
  • the neural network model is trained, including:
  • Input the training set data to the input layer of the model, go through the hidden layer, and finally reach the output layer and output the result; calculate the error between the predicted value of the model output and the actual value in the sample data, and hide the error from the output layer
  • the layer propagates backwards until it propagates to the input layer; the values of the model parameters are adjusted according to the error; the above process is iterated until convergence.
  • the model training process in order to increase the generalization ability of the model and prevent over-fitting, for example, input the actual value of the IC value at time K into the three-layer feedforward network model, and output the IC value at time K+1.
  • a random noise value is added based on the predicted value of the IC value at time K+1 (for example, multiplied by 0.01 from the sample value of the standard normal distribution as the noise value).
  • the least square method is used to minimize the error between the predicted value (including the noise value) and the actual value. .
  • the above steps of dividing the sample data into the training set, the evaluation set and the test set based on the cross-validation method may be replaced by: dividing the sample data into a training set and a test set based on the cross-validation method.
  • the number of sample data in the training set, the evaluation set, and the test set can be set as needed, and is not limited to the above-exemplified scheme.
  • this step can be refined into the following steps:
  • the score of each stock is calculated separately.
  • the information coefficient prediction model is used to predict the public opinion factor of the M stocks and the information coefficient of the public opinion factor for the next period is IC A(t+1) and IC B(t+1) , respectively.
  • the heat factor and the lyric emotion factor respectively correspond to the weights when calculating the stock score.
  • the sentiment heat factor is a positive factor, indicating that the higher the stock's heat of opinion observation, the higher the future earnings of the stock will be; the sentiment affective factor may be either a positive factor or a negative factor, which needs to be explained.
  • the value of the lyric emotion observation value is [-1, 1]. The higher the absolute value of the lyric emotion observation value, the stronger the positive (negative) emotion. When the lyric emotion observation value of a stock is negative, The future earnings of the stock have a negative impact.
  • the lyric sentiment factor is a negative factor; on the contrary, when the lyric sentiment observation value of a stock is positive, there is a positive impact on the future earnings of the stock. At this time, lyric emotion The factor is a positive factor.
  • the first weights corresponding to the public opinion heat factor and the lyric emotion factor are determined by the above steps as ⁇ A1 and ⁇ B1 respectively .
  • the first weight of the stock's lyric heat factor and the lyric emotional factor is returned.
  • the second weights ⁇ A2 and ⁇ B2 corresponding to the two factors of the stock are determined as the weights ⁇ A , ⁇ B corresponding to the two factors when calculating the stock score.
  • the calculation formula of ⁇ A and ⁇ B is:
  • the weight corresponding to all factors is 0, before the first weight of each factor is normalized, the method of smoothing is used in each factor.
  • the first weight is added to ⁇ 0 and then normalized.
  • the calculation formula of ⁇ A and ⁇ B is:
  • c is the total number of factors.
  • there are only two lyric factors considered: the sensational heat factor and the lyric emotional factor, so ⁇ 0 1/2.
  • the weighting strategy such as factor is adopted, that is, the weight corresponding to each factor is the same.
  • the calculation formula for stocks is:
  • i the stock's lyric factor
  • i ⁇ A, B ⁇
  • ⁇ i the weight corresponding to the stock's sensation factor i
  • ⁇ i ⁇ A , ⁇ B ⁇
  • the score of the stock calculated by the above steps can reflect the trend of the stock in the future income to a certain extent, and the higher the score of the stock, the higher the future profit may be. Therefore, after obtaining the scores of each of the M stocks, the M stocks are sorted according to the order of the scores, and the first predetermined number (for example, N only) stocks with the highest scores are selected from the M stocks. As the target stock to be invested.
  • the investment portfolio including the target stock to be invested it is also necessary to separately determine the proportion of each target stock in the total investment amount, for example, to make an equal investment in the target stock, or to determine according to the score of the target stock. Corresponding proportions, stocks with high scores have a larger proportion, and stocks with lower scores have a smaller proportion.
  • the specific operations can be set according to the actual needs of investors. Finally, the final portfolio is determined based on the proportion of the target stock and the target stock.
  • the solution of the present application is also applicable to optimizing an existing investment portfolio, separately calculating the scores of the stocks in the existing investment portfolio, selecting the stocks with higher scores as the target stocks, and the proportion of the target stocks. Make adjustments to get an optimized portfolio.
  • the investment stock selection method based on the public opinion factor proposed in the above embodiment calculates the first information coefficient of each stocky factor of the stock according to the stock sentiment factor observation value and the yield rate data, and predicts the second information by using the first information coefficient and the information coefficient prediction model.
  • the coefficient reflects the ability of each sentiment factor to predict the future earnings of the stock;
  • the weight of each sentiment factor is weighted according to the second information coefficient, realizing the dynamic empowerment of each sentiment factor, and improving the reliability of each sentiment factor;
  • Factor observations and corresponding weights are used to calculate stock scores, and stocks with higher scores are selected as target investment stocks, which improves the authenticity of stock scores, and visually demonstrates the impact of the public opinion factor observations on the future earnings of stocks, which helps investors Control risk and increase investment income.
  • the embodiment of the present application further provides a computer readable storage medium, where the computer-readable storage medium stores an investment stock selection program based on a public opinion factor, and the investment stock selection program based on the public opinion factor is executed by the processor. Implement the following operations:
  • the first information coefficient of each sentiment factor is input into the pre-trained information coefficient prediction model, and the second information coefficient of each future public opinion factor is predicted;
  • the plurality of stocks are sorted according to the order of the scores, and the first predetermined number of stocks ranked first is selected as the target investment stocks.
  • the specific implementation manner of the computer readable storage medium of the present application is substantially the same as the specific implementation method of the above-mentioned public opinion factor based investment stock selection method, and details are not described herein again.
  • a disk including a number of instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the methods described in the various embodiments of the present application.
  • a terminal device which may be a mobile phone, a computer, a server, or a network device, etc.

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • Educational Administration (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Technology Law (AREA)
  • Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本申请提出一种基于舆情因子的投资选股方法，该方法包括：获取预先确定的多只股票在第一预设时间内的舆情因子观测值及第二预设时间内的收益率；利用所述舆情因子观测值及收益率，计算各舆情因子的第一信息系数；分别将各舆情因子的第一信息系数输入预先训练好的信息系数预测模型中，预测未来各舆情因子的第二信息系数；分别计算所述多只股票的评分；及，根据评分的高低顺序对所述多只股票进行排序，筛选出排序靠前的第一预设数量的股票，作为目标投资股票。本申请还提出一种电子装置及存储介质。本申请通过计算股票评分，直观地展示舆情因子对股票未来收益的影响。

Description

基于舆情因子的投资选股方法、装置及存储介质
本申请基于巴黎公约申明享有2018年4月26日递交的申请号为CN 2018103823117、名称为“基于舆情因子的投资选股方法、装置及存储介质”的中国专利申请的优先权，该中国专利申请的整体内容以参考的方式结合在本申请中。
技术领域
本申请涉及金融大数据挖掘领域，尤其涉及一种基于舆情因子的投资选股方法、电子装置及计算机可读存储介质。
背景技术
股票的舆情因子(例如，新闻热度及大众对相应新闻的情绪态度)在一定程度上可以表达为该股票的市场潜在能量，当一只股票的舆情因子异常(例如，当一只股票有重大负面新闻出现或者获得新一轮投资)时，需要即时更改该股票的投资策略。一般都是预先为股票的某个指标设置阈值，然后等权重投资指标满足阈值条件的股票，或者根据专业投资人员市场经验对相应股票设定投资策略。这种方式对做决策投资人员的专业技能要求较高，很容易出错，可能会增加误操作的风险。
发明内容
本申请提供一种基于舆情因子的投资选股方法、电子装置及计算机可读存储介质，其主要目的在于通过计算股票评分，直观地展示舆情因子对股票未来收益的影响，选择评分高的股票作为目标投资股票，减小投资风险、提高投资收益。
为实现上述目的，本申请提供一种基于舆情因子的投资选股方法，该方法包括：
获取预先确定的多只股票在第一预设时间内的舆情因子观测值及第二预设时间内的收益率；
根据所述舆情因子观测值及收益率，计算各舆情因子的第一信息系数；
分别将各舆情因子的第一信息系数输入预先训练好的信息系数预测模型中，预测未来各舆情因子的第二信息系数；
根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分；及
根据评分的高低顺序对所述多只股票进行排序，筛选出排序靠前的第一预设数量的股票，作为目标投资股票。
此外，为实现上述目的，本申请还提供一种电子装置，该电子装置包括：存储器、处理器，所述存储器上存储有基于舆情因子的投资选股程序，所述基于舆情因子的投资选股程序被所述处理器执行时实现如下步骤：
获取预先确定的多只股票在第一预设时间内的舆情因子观测值及第二预设时间内的收益率；
根据所述舆情因子观测值及收益率，计算各舆情因子的第一信息系数；
分别将各舆情因子的第一信息系数输入预先训练好的信息系数预测模型中，预测未来各舆情因子的第二信息系数；
根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分；及
根据评分的高低顺序对所述多只股票进行排序，筛选出排序靠前的第一预设数量的股票，作为目标投资股票。
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，所述计算机可读存储介质上存储有基于舆情因子的投资选股程序，所述基于舆情因子的投资选股程序被处理器执行时实现如上所述的基于舆情因子的投资选股方法的步骤。
相较于现有技术，本申请提出的基于舆情因子的投资选股方法、电子装置及计算机可读存储介质，根据股票舆情因子观测值及收益率数据，计算股票各舆情因子的第一信息系数，利用第一信息系数及信息系数预测模型预测第二信息系数，反映各舆情因子对股票未来收益的预测能力；根据第二信息系数对各舆情因子进行赋权，实现了对各舆情因子的动态赋权，提高了各舆 情因子的可靠性；根据各舆情因子观测值及相应的权重计算股票评分，选择评分较高的股票作为目标投资股票，提高了股票评分的真实性，直观地展示了舆情因子观测值对股票未来收益的影响，有助于投资者控制风险、提高投资收益。
附图说明
图1为本申请电子装置较佳实施例的示意图；
图2为图1中基于舆情因子的投资选股程序的程序模块图；
图3为本申请基于舆情因子的投资选股方法较佳实施例的流程图。
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
应当理解，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。
本申请提供一种电子装置1。参照图1所示，为本申请电子装置1较佳实施例的示意图。
在本实施例中，该电子装置1包括存储器11、处理器12，网络接口13及通信总线14。其中，通信总线14用于实现这些组件之间的连接通信。
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)。
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的外部存储设备，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安装于所述电子装置1的基于舆情因子的投资选股程序10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。
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，例如执行基于舆情因子的投资选股程序10等。
图1仅示出了具有组件11-14以及基于舆情因子的投资选股程序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。
可选的，该电子装置1还可以包括用户接口，用户接口可以包括显示器(Display)、输入单元比如键盘(Keyboard)，可选的用户接口还可以包括标准的有线接口、无线接口。
可选地，该电子装置1还可以包括显示器，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)触摸器等。显示器用于显示在电子装置中处理的信息以及用于显示可视化的用户界面。
在图1所示的装置实施例中，作为一种计算机存储介质的存储器11中包括基于舆情因子的投资选股程序10，处理器12执行存储器11中存储的基于舆情因子的投资选股程序10时实现以下步骤：
获取预先确定的多只股票在第一预设时间内的舆情因子观测值及第二预设时间内的收益率；
在本实施例中，股票的舆情因子包括舆情热度因子及舆情情感因子，故各只股票的舆情因子观测值即为第一预设时间内的舆情热度观测值及舆情情感观测值。
假设有预先确定的M只股票，用户可根据需要设置第一预设时间及第二预设时间。可以理解的是，从预先约定的M只股票中选择N只目标股票，作为下一期投资的对象，选股频率可以是每隔一周进行一次，也可以是每隔一个月进行一次，在本实施例中为每隔一周进行一次。假设选股时间为t时刻，相应地，例如，第一预设时间可以设置为t-1时刻(即，选股当期的上一期)，第二预设时间可以设置为t时刻(即，选股当期)，每两期之间的时间间隔为一周。
需要说明的是，在获取第一预设时间内的各舆情因子观测值及第二预设 时间内的收益率时，对于舆情因子观测值而言，不需要通过收集、处理各只股票相关的舆情信息，而是直接从一些网站上手动提取，或者通过程序对外接口直接提取，例如，从通联数据或新浪股吧提取预先确定的M只股票在第一预设时间内的舆情热度观测值及舆情情感观测值。
对于收益率而言，则需要通过计算得到。在本实施例中，M只股票在第二预设时间内对应的收益率为周收益率，在计算周收益率时，采用的是M只股票的以一周为时间间隔的收盘价，具体地，计算公式为：
R＝(P t-P t-1)/P t-1
其中，R为股票在t时刻(当期)的周收益率，P t为股票在t时刻(当期)的收盘价，P t-1为股票在t-1时刻(上一期)的收盘价。
根据所述舆情因子观测值及收益率，计算各舆情因子的第一信息系数；
为了直观地展示出各舆情因子对股票收益的预测能力，需根据M只股票在第一预设时间内的各舆情因子观测值及第二预设时间内的收益率，计算M只股票的各舆情因子在第二预设时间对应的信息系数(Information Coefficient，IC)。具体地，上述第一信息系数指在t时刻(当期)股票各舆情因子的信息系数。
在本实施例中，将斯皮尔曼相关系数(Rank IC)作为股票的各舆情因子的信息系数。Rank IC，即在某时刻，全部股票在某因子上的排序与下一期全部股票收益的排序之间的截面相关系数，计算公式为：
Figure PCTCN2018102127-appb-000001
其中，
Figure PCTCN2018102127-appb-000002
为t-1时刻(上一期)各股票的舆情因子f的观测值排名，X为t-1时刻(上一期)舆情因子f的观测值；
Figure PCTCN2018102127-appb-000003
为t时刻(当期)各股票的收益率排名，Y为t时刻(当期)股票的收益率。
利用上述步骤，分别计算M只股的舆情热度因子及舆情情感因子在t时刻(当期)的IC值，即第一IC值。
分别将各舆情因子的第一信息系数输入预先训练好的信息系数预测模型中，预测未来各舆情因子的第二信息系数；
可以理解的是，各舆情因子的IC值能够很好地反映各舆情因子的预测能力，IC值越大，就说表明该因子其在该期对股票收益的预测能力越强。鉴于 本申请的目的在于选择下一期的目标股票，则需要了解各舆情因子在下一期对股票收益的预测能力，也就是各舆情因子在下一期的IC值。具体地，上述第二信息系数指股票的各舆情因子在t+1时刻(下一期)的IC值。
在确定M只股票的舆情热度因子及舆情情感因子的第一IC值后，将M只股票的舆情热度因子及舆情情感因子的第一IC值输入预先确定的信息系数预测模型中，预测M只股票的舆情热度因子及舆情情感因子在t+1时刻(下一期)的IC值，即第二IC值。
具体地，假设M只股票的舆情热度因子及舆情情感因子的第一IC值分别为IC At、IC Bt，将其以[IC At,IC Bt]的形式输入信息系数预测模型中，模型输出的结果为[IC A(t+1),IC B(t+1)]，其中，IC A(t+1)、IC B(t+1)分别为该M只股票的舆情热度因子及舆情情感因子的第二IC值。
在本实施例中，所述预先确定的信息系数预测模型通过训练神经网络得到。鉴于模型的输入数据只是各舆情因子在某个时刻的IC值，数据形式比较简单，故选用用含有一层隐藏层的三层前馈神经网络进行训练。具体地，该信息系数预测模型的训练步骤包括：
在计算得到预先确定的第二预设数量(例如，3000只)的股票的各舆情因子在第三预设时间(两年)内的每个时刻对应的信息系数后，确定样本数据[I K，I K+1]，其中，I K表示各舆情因子在K时刻对应的IC值，I K+1表示各舆情因子在K+1时刻对应的IC值(即，后面提到的实际值)。
具体地，每两个时刻之间的时间间隔可以为一天，也可以为一周。例如，当两个时刻之间的时间间隔为一天时，则样本数据为各舆情因子在每一天的IC值；当两个时刻之间的时间间隔为一周时，则样本数据需要采集的是时间间隔为一周的数据，该数据可以是各舆情因子在连续一周的五个交易日的IC值的平均值，也可以是取连续一周的五个交易日中最后一个交易日的IC值。
基于交叉验证法(cross-validation)将样本数据划分为训练集、评估集和测试集(例如，70％的样本数据作为训练集，10％的样本数据作为评估集，20％的样本数据作为测试集)。
将训练集的样本数据输入至三层前馈神经网络模型，对模型进行训练，初步确定模型参数；所述测试集用于在训练过程中对三层前馈神经网络模型的准确率进行测试，将所述测试集的样本数据输入训练得到的所述三层前馈 神经网络模型中，以对训练得到的所述三层前馈神经网络模型进行测试，当训练得到的所述三层前馈神经网络模型满足预设验证条件(例如，模型预测准确率大于或等于小于预设阈值)，则训练完成，将训练完成的三层前馈神经网络模型设置为信息系数预测模型。
根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分；
具体地，该步骤可以细化为以下步骤：
根据各舆情因子对应的第二信息系数，分别确定所述M只股票的各舆情因子对应的权重；根据所述舆情因子观测值及各舆情因子所占的权重，分别计算每只股票的评分。
在t时刻，利用信息系数预测模型预测得到M只股票的舆情热度因子、舆情情感因子未来一期的信息系数分别为IC A(t+1)、IC B(t+1)后，需要确定舆情热度因子、舆情情感因子在计算股票评分时分别对应的权重。
在确定各舆情因子对应的权重之前，需判断各舆情因子的因子种类，其中，因子种类包括：正向因子和负向因子。具体地，舆情热度因子为正向因子，表示股票的舆情热度观测值越高，该股票的未来收益会比较高；舆情情感因子既可能是正向因子，又可能是负向因子，需要说明的是，舆情情感观测值的取值范围为[-1,1]，舆情情感观测值的绝对值越高，正面(负面)的情感越强烈，当某只股票的舆情情感观测值为负时，对该股票的未来收益存在负面影响，此时，舆情情感因子为负向因子；相反，当某只股票的舆情情感观测值为正时，对该股票的未来收益存在正面影响，此时，舆情情感因子为正向因子。
具体地，对于正向因子，当IC i(t+1)＞0时，该因子i对应的第一权重ω i1＝IC i(t+1)，否则，ω i1＝0，表示该因子在未来一期失效；对于负向因子，当IC i(t+1)＜0时，该因子i对应的第一权重ω i1＝-IC i(t+1)，否则，ω i1＝0，表示该因子在未来一期失效。
利用上述步骤确定舆情热度因子、舆情情感因子对应的第一权重分别为ω A1、ω B1，为了便于后续计算股票对应的评分，对股票的舆情热度因子及舆情情感因子对应的第一权重进行归一化，根据归一化结果确定股票的两个因子对应的第二权重ω A2、ω B2，作为计算股票评分时两个因子对应的权重ω A、 ω B。在本实施例中，ω A、ω B的计算公式为：
ω A＝ω A2＝ω A1/(ω A1+ω B1)
ω B＝ω B2＝ω B1/(ω A1+ω B1)
分别获取M只股票的舆情热度观测值X A、舆情情感观测值X B、舆情热度因子A对应的权重ω A及舆情情感因子B对应的权重ω B，根据预设的计算公式，计算M只股票的评分。具体地，股票的评分的计算公式为：
S＝∑ω i*X i
其中，i为股票的舆情因子，i＝{A,B}，ω i为股票的舆情因子i对应的权重，ω i＝{ω A,ω B}，X i为t时刻股票的各舆情因子的观测值，X i＝{X A,X B}。
根据评分的高低顺序对所述多只股票进行排序，筛选出排序靠前的第一预设数量的股票，作为目标投资股票。
可以理解的是，利用上述步骤计算得到的股票的评分，在一定程度上可以反映出股票在未来收益的走势，股票的评分越高，其未来收益可能越高。因此，在得到M只股票中的每只股票评分后，按照评分高低顺序对M只股票进行排序，从M只股票中筛选出评分最高的第一预设数量(例如，N只)的股票，作为待投资的目标股票。
在其他实施例中，确定包含待投资的目标股票的投资组合后，还需要分别确定各只目标股票占投资总额的比重，例如，对目标股票进行等权投资，或者，根据目标股票的评分确定相应的比重，评分高的股票对应的比重较大，评分低的股票对应的比重较小)，具体操作可根据投资者的实际需求进行设置。最后，根据目标股票及目标股票对应的比重确定最终投资组合。
上述实施例提出的电子装置，根据股票舆情因子观测值及收益率数据，计算股票各舆情因子的第一信息系数，利用第一信息系数及信息系数预测模型预测第二信息系数，反映各舆情因子对股票未来收益的预测能力；根据第二信息系数对各舆情因子进行赋权，实现了对各舆情因子的动态赋权，提高了各舆情因子的可靠性；根据各舆情因子观测值及相应的权重计算股票评分，选择评分较高的股票作为目标投资股票，提高了股票评分的真实性，直观地展示了舆情因子观测值对股票未来收益的影响，有助于投资者控制风险、提高投资收益。
可选地，在其他的实施例中，基于舆情因子的投资选股程序10还可以被分割为一个或者多个模块，一个或者多个模块被存储于存储器11中，并由一个或多个处理器12所执行，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中基于舆情因子的投资选股程序的程序模块图。在本实施例中，基于舆情因子的投资选股程序10可以被分割为：获取模块110、信息系数计算模块120、预测模块130、评分计算模块140及选股模块150。所述模块110-150所实现的功能或操作步骤均与上文类似，此处不再详述，示例性地，例如其中：
获取模块110，用于获取预先确定的多只股票在第一预设时间内的舆情因子观测值及第二预设时间内的收益率；
信息系数计算模块120，用于根据所述舆情因子观测值及收益率，计算各舆情因子的第一信息系数；
预测模块130，用于分别将各舆情因子的第一信息系数输入预先训练好的信息系数预测模型中，预测未来各舆情因子的第二信息系数；
评分计算模块140，用于根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分；及
选股模块150，用于根据评分的高低顺序对所述多只股票进行排序，筛选出排序靠前的第一预设数量的股票，作为目标投资股票。
此外，本申请还提供一种基于舆情因子的投资选股方法。参照图3所示，为本申请基于舆情因子的投资选股方法较佳实施例的流程图。该方法可以由一个装置执行，该装置可以由软件和/或硬件实现。
在本实施例中，基于舆情因子的投资选股方法包括：步骤S1-S5。
S1、获取预先确定的多只股票在第一预设时间内的舆情因子观测值及第二预设时间内的收益率；
在本实施例中，股票的舆情因子包括舆情热度因子及舆情情感因子，故各只股票的舆情因子观测值即为第一预设时间内的舆情热度观测值及舆情情感观测值。
假设有预先确定的M只股票，用户可根据需要设置第一预设时间及第二预设时间。可以理解的是，从预先约定的M只股票中选择N只目标股票，作 为下一期投资的对象，选股频率可以是每隔一周进行一次，也可以是每隔一个月进行一次，在本实施例中为每隔一周进行一次。假设选股时间为t时刻，相应地，例如，第一预设时间可以设置为t-1时刻(即，选股当期的上一期)，第二预设时间可以设置为t时刻(即，选股当期)，每两期之间的时间间隔为一周。
需要说明的是，在获取第一预设时间内的各舆情因子观测值及第二预设时间内的收益率时，对于舆情因子观测值而言，不需要通过收集、处理各只股票相关的舆情信息，而是直接从一些网站上手动提取，或者通过程序对外接口直接提取，例如，从通联数据或新浪股吧提取预先确定的M只股票在第一预设时间内的舆情热度观测值及舆情情感观测值。
对于收益率而言，则需要通过计算得到。在本实施例中，M只股票在第二预设时间内对应的收益率为周收益率，在计算周收益率时，采用的是M只股票的以一周为时间间隔的收盘价，具体地，计算公式为：
R＝(P t-P t-1)/P t-1
其中，R为股票在t时刻(当期)的周收益率，P t为股票在t时刻(当期)的收盘价，P t-1为股票在t-1时刻(上一期)的收盘价。
S2、根据所述舆情因子观测值及收益率，计算各舆情因子的第一信息系数；
为了直观地展示出各舆情因子对股票收益的预测能力，需根据M只股票在第一预设时间内的各舆情因子观测值及第二预设时间内的收益率，计算M只股票的各舆情因子在第二预设时间对应的信息系数(Information Coefficient，IC)。具体地，上述第一信息系数指在t时刻(当期)股票各舆情因子的信息系数。
在本实施例中，将斯皮尔曼相关系数(Rank IC)作为股票的各舆情因子的信息系数。Rank IC，即在某时刻，全部股票在某因子上的排序与下一期全部股票收益的排序之间的截面相关系数，计算公式为：
Figure PCTCN2018102127-appb-000004
其中，
Figure PCTCN2018102127-appb-000005
为t-1时刻(上一期)各股票的舆情因子f的观测值排名，X为t-1时刻(上一期)舆情因子f的观测值；
Figure PCTCN2018102127-appb-000006
为t时刻(当期)各股票的收益率排名，Y为t时 刻(当期)股票的收益率。
利用上述步骤，分别计算M只股的舆情热度因子及舆情情感因子在t时刻(当期)的IC值，即第一IC值。
S3、分别将各舆情因子的第一信息系数输入预先训练好的信息系数预测模型中，预测未来各舆情因子的第二信息系数；
可以理解的是，各舆情因子的IC值能够很好地反映各舆情因子的预测能力，IC值越大，就说表明该因子其在该期对股票收益的预测能力越强。鉴于本申请的目的在于选择下一期的目标股票，则需要了解各舆情因子在下一期对股票收益的预测能力，也就是各舆情因子在下一期的IC值。具体地，上述第二信息系数指股票的各舆情因子在t+1时刻(下一期)的IC值。
在确定M只股票的舆情热度因子及舆情情感因子的第一IC值后，将M只股票的舆情热度因子及舆情情感因子的第一IC值输入预先确定的信息系数预测模型中，预测M只股票的舆情热度因子及舆情情感因子在t+1时刻(下一期)的IC值，即第二IC值。
具体地，假设M只股票的舆情热度因子及舆情情感因子的第一IC值分别为IC At、IC Bt，将其以[IC At,IC Bt]的形式输入信息系数预测模型中，模型输出的结果为[IC A(t+1),IC B(t+1)]，其中，IC A(t+1)、IC B(t+1)分别为该M只股票的舆情热度因子及舆情情感因子的第二IC值。
在本实施例中，所述预先确定的信息系数预测模型通过训练神经网络得到。鉴于模型的输入数据只是各舆情因子在某个时刻的IC值，数据形式比较简单，故选用用含有一层隐藏层的三层前馈神经网络进行训练。具体地，该信息系数预测模型的训练步骤包括：
分别采集预先确定的第二预设数量的股票在第三预设时间内各舆情因子在每个时刻的历史观测值、及各只股票在每个时刻的历史收益率，计算各舆情因子在每个时刻对应的信息系数，以获取样本数据；及
将样本数据划分为训练集、评估集和测试集，并利用训练集的样本数据对神经网络进行训练，得到信息系数预测模型，利用测试集的样本数据对信息系数预测模型的准确率进行测试，得到最终的信息系数预测模型。
同理，根据Rank IC的计算公式计算得到预先确定的第二预设数量(例如，3000只)的股票的各舆情因子在第三预设时间(两年)内的每个时刻对应的 信息系数后，确定样本数据[I K，I K+1]，其中，I K表示各舆情因子在K时刻对应的IC值，I K+1表示各舆情因子在K+1时刻对应的IC值(即，后面提到的实际值)。
具体地，每两个时刻之间的时间间隔可以为一天，也可以为一周。例如，当两个时刻之间的时间间隔为一天时，则样本数据为各舆情因子在每一天的IC值；当两个时刻之间的时间间隔为一周时，则样本数据需要采集的是时间间隔为一周的数据，该数据可以是各舆情因子在连续一周的五个交易日的IC值的平均值，也可以是取连续一周的五个交易日中最后一个交易日的IC值。
基于交叉验证法(cross-validation)将样本数据划分为训练集、评估集和测试集(例如，70％的样本数据作为训练集，10％的样本数据作为评估集，20％的样本数据作为测试集)。
将训练集的样本数据输入至三层前馈神经网络模型，对模型进行训练，初步确定模型参数；所述测试集用于在训练过程中对三层前馈神经网络模型的准确率进行测试，将所述测试集的样本数据输入训练得到的所述三层前馈神经网络模型中，以对训练得到的所述三层前馈神经网络模型进行测试，当训练得到的所述三层前馈神经网络模型满足预设验证条件(例如，模型预测准确率大于或等于小于预设阈值)，则训练完成，将训练完成的三层前馈神经网络模型设置为信息系数预测模型。
鉴于反向传播算法(Back propagation)是目前用来训练人工神经网络(Artificial Neural Network，ANN)的最常用且最有效的算法，因此，本实施例中，基于反向传播法对三层前馈神经网络模型进行训练，具体包括：
将训练集数据输入到模型的输入层，经过隐藏层，最后达到输出层并输出结果；计算模型输出的预测值与样本数据中的实际值之间的误差，并将该误差从输出层向隐藏层反向传播，直至传播到输入层；根据误差调整模型参数的值；不断迭代上述过程，直至收敛。
需要说明的是，为了提高模型训练速度，在训练过程中输入数据时，不会每次只输入一个时刻的数据进行训练，而是每次固定输入batch size＝k的样本数据进行训练，其中，k的大小可以根据需要设置，在本实施例中可设置为1024。
优选地，在模型训练过程中，为了增加模型的泛化能力和防止过拟合， 例如，将K时刻的IC值的实际值输入三层前馈网络模型中，输出K+1时刻的IC值的预测值时，在K+1时刻的IC值的预测值的基础上增加一个随机噪声值(例如，用从标准正态分布的采样值乘以0.01来当作噪声值)。进一步地，在计算K+1时刻的IC值的预测值(含噪声值)与实际值之间的误差时，利用最小二乘法来最小化预测值(含噪声值)与实际值之间的误差。
需要注意的是，上述基于交叉验证法将样本数据划分为训练集、评估集和测试集的步骤可替换为：基于交叉验证法将样本数据划分为训练集和测试集。且训练集、评估集和测试集中样本数据的数量可根据需要设置，并不限于上述例举的方案。
S4、根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分；
具体地，该步骤可以细化为以下步骤：
根据各舆情因子对应的第二信息系数，分别确定所述M只股票的各舆情因子对应的权重；及
根据所述舆情因子观测值及各舆情因子所占的权重，分别计算每只股票的评分。
在t时刻，利用信息系数预测模型预测得到M只股票的舆情热度因子、舆情情感因子未来一期的信息系数分别为IC A(t+1)、IC B(t+1)后，需要确定舆情热度因子、舆情情感因子在计算股票评分时分别对应的权重。
在确定各舆情因子对应的权重之前，需判断各舆情因子的因子种类，其中，因子种类包括：正向因子和负向因子。具体地，舆情热度因子为正向因子，表示股票的舆情热度观测值越高，该股票的未来收益会比较高；舆情情感因子既可能是正向因子，又可能是负向因子，需要说明的是，舆情情感观测值的取值范围为[-1,1]，舆情情感观测值的绝对值越高，正面(负面)的情感越强烈，当某只股票的舆情情感观测值为负时，对该股票的未来收益存在负面影响，此时，舆情情感因子为负向因子；相反，当某只股票的舆情情感观测值为正时，对该股票的未来收益存在正面影响，此时，舆情情感因子为正向因子。
具体地，对于正向因子，当IC i(t+1)＞0时，该因子i对应的第一权重ω i1＝IC i(t+1)，否则，ω i1＝0，表示该因子在未来一期失效；对于负向因子， 当IC i(t+1)＜0时，该因子i对应的第一权重ω i1＝-IC i(t+1)，否则，ω i1＝0，表示该因子在未来一期失效。
利用上述步骤确定舆情热度因子、舆情情感因子对应的第一权重分别为ω A1、ω B1，为了便于后续计算股票对应的评分，对股票的舆情热度因子及舆情情感因子对应的第一权重进行归一化，根据归一化结果确定股票的两个因子对应的第二权重ω A2、ω B2，作为计算股票评分时两个因子对应的权重ω A、ω B。在本实施例中，ω A、ω B的计算公式为：
ω A＝ω A2＝ω A1/(ω A1+ω B1)
ω B＝ω B2＝ω B1/(ω A1+ω B1)
在其他实施例中，为了防止所有的因子都失效，即所有因子对应的权重均为0的情况，在对各因子的第一权重进行归一化之前，采用smoothing的方法，在每个因子的第一权重的基础上加上ω 0，然后再进行归一化处理。此时，ω A、ω B的计算公式为：
ω A＝ω A2＝(ω A1+ω 0)/(ω A1+ω B1+1)
ω B＝ω B2＝(ω B1+ω 0)/(ω A1+ω B1+1)
ω 0＝1/c
其中，c为因子总数。在本实施例中，考虑的舆情因子只有两个：舆情热度因子及舆情情感因子，故ω 0＝1/2。
也就是说，当舆情热度因子、舆情情感因子都失效时，采用的就是因子等权重策略，即每个因子对应的权重都相同。
分别获取M只股票的舆情热度观测值X A、舆情情感观测值X B、舆情热度因子A对应的权重ω A及舆情情感因子B对应的权重ω B，根据预设的计算公式，计算M只股票的评分。具体地，股票的评分的计算公式为：
S＝∑ω i*X i
其中，i为股票的舆情因子，i＝{A,B}，ω i为股票的舆情因子i对应的权重，ω i＝{ω A,ω B}，X i为t时刻股票的各舆情因子的观测值，X i＝{X A,X B}。
S5、根据评分的高低顺序对所述多只股票进行排序，筛选出排序靠前的第一预设数量的股票，作为目标投资股票。
可以理解的是，利用上述步骤计算得到的股票的评分，在一定程度上可以反映出股票在未来收益的走势，股票的评分越高，其未来收益可能越高。 因此，在得到M只股票中的每只股票评分后，按照评分高低顺序对M只股票进行排序，从M只股票中筛选出评分最高的第一预设数量(例如，N只)的股票，作为待投资的目标股票。
在其他实施例中，确定包含待投资的目标股票的投资组合后，还需要分别确定各只目标股票占投资总额的比重，例如，对目标股票进行等权投资，或者，根据目标股票的评分确定相应的比重，评分高的股票对应的比重较大，评分低的股票对应的比重较小)，具体操作可根据投资者的实际需求进行设置。最后，根据目标股票及目标股票对应的比重确定最终投资组合。
需要说明的是，本申请的方案还适用于对已有投资组合进行优化，分别计算已有投资组合中各股票的评分，选择评分较高的股票作为目标股票，并对目标股票所占的比重进行调整，得到优化后的投资组合。
上述实施例提出的基于舆情因子的投资选股方法，根据股票舆情因子观测值及收益率数据，计算股票各舆情因子的第一信息系数，利用第一信息系数及信息系数预测模型预测第二信息系数，反映各舆情因子对股票未来收益的预测能力；根据第二信息系数对各舆情因子进行赋权，实现了对各舆情因子的动态赋权，提高了各舆情因子的可靠性；根据各舆情因子观测值及相应的权重计算股票评分，选择评分较高的股票作为目标投资股票，提高了股票评分的真实性，直观地展示了舆情因子观测值对股票未来收益的影响，有助于投资者控制风险、提高投资收益。
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上存储有基于舆情因子的投资选股程序，所述基于舆情因子的投资选股程序被处理器执行时实现如下操作：
获取预先确定的多只股票在第一预设时间内的舆情因子观测值及第二预设时间内的收益率；
根据所述舆情因子观测值及收益率，计算各舆情因子的第一信息系数；
分别将各舆情因子的第一信息系数输入预先训练好的信息系数预测模型中，预测未来各舆情因子的第二信息系数；
根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分；及
根据评分的高低顺序对所述多只股票进行排序，筛选出排序靠前的第一预设数量的股票，作为目标投资股票。
本申请之计算机可读存储介质的具体实施方式与上述基于舆情因子的投资选股方法的具体实施方式大致相同，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。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、装置、物品或者方法中还存在另外的相同要素。
上述本申请实施例序号仅仅为了描述，不代表实施例的优劣。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(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，或者网络设备等)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. 一种基于舆情因子的投资选股方法，应用于电子装置，其特征在于，该方法包括：
    获取预先确定的多只股票在第一预设时间内的舆情因子观测值及第二预设时间内的收益率；
    根据所述舆情因子观测值及收益率，计算各舆情因子的第一信息系数；
    分别将各舆情因子的第一信息系数输入预先训练好的信息系数预测模型中，预测未来各舆情因子的第二信息系数；
    根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分；及
    根据评分的高低顺序对所述多只股票进行排序，筛选出排序靠前的第一预设数量的股票，作为目标投资股票。
  2. 根据权利要求1所述的基于舆情因子的投资选股方法，其特征在于，所述第一信息系数的计算公式为：
    Figure PCTCN2018102127-appb-100001
    其中，
    Figure PCTCN2018102127-appb-100002
    为t-1时刻各股票的舆情因子f的观测值排名，
    Figure PCTCN2018102127-appb-100003
    为t时刻各股票的收益率排名。
  3. 根据权利要求2所述的基于舆情因子的投资选股方法，其特征在于，所述预先确定的信息系数预测模型的训练步骤包括：
    分别采集预先确定的第二预设数量的股票在第三预设时间内各舆情因子在每个时刻的历史观测值、及各只股票在每个时刻的历史收益率，计算各舆情因子在每个时刻对应的信息系数，以获取样本数据；及
    将样本数据划分为训练集、评估集和测试集，并利用训练集的样本数据对神经网络进行训练，得到信息系数预测模型，利用测试集的样本数据对信息系数预测模型的准确率进行测试，得到最终的信息系数预测模型。
  4. 根据权利要求2所述的基于舆情因子的投资选股方法，其特征在于，所述“根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分”的步骤包括：
    根据各舆情因子对应的第二信息系数，分别确定所述M只股票的各舆情因子对应的权重；及
    根据所述舆情因子观测值及各舆情因子所占的权重，分别计算每只股票 的评分。
  5. 根据权利要求4所述的基于舆情因子的投资选股方法，其特征在于，所述预先确定的信息系数预测模型的训练步骤包括：
    分别采集预先确定的第二预设数量的股票在第三预设时间内各舆情因子在每个时刻的历史观测值、及各只股票在每个时刻的历史收益率，计算各舆情因子在每个时刻对应的信息系数，以获取样本数据；及
    将样本数据划分为训练集、评估集和测试集，并利用训练集的样本数据对神经网络进行训练，得到信息系数预测模型，利用测试集的样本数据对信息系数预测模型的准确率进行测试，得到最终的信息系数预测模型。
  6. 如权利要求4所述的基于舆情因子的投资选股方法，其特征在于，所述各只股票的评分的计算公式为：
    S＝Σω i*X i
    其中，i为股票的舆情因子，ω i为股票的舆情因子i对应的权重，X i为t时刻股票的各舆情因子的观测值。
  7. 根据权利要求6所述的基于舆情因子的投资选股方法，其特征在于，所述预先确定的信息系数预测模型的训练步骤包括：
    分别采集预先确定的第二预设数量的股票在第三预设时间内各舆情因子在每个时刻的历史观测值、及各只股票在每个时刻的历史收益率，计算各舆情因子在每个时刻对应的信息系数，以获取样本数据；及
    将样本数据划分为训练集、评估集和测试集，并利用训练集的样本数据对神经网络进行训练，得到信息系数预测模型，利用测试集的样本数据对信息系数预测模型的准确率进行测试，得到最终的信息系数预测模型。
  8. 一种电子装置，其特征在于，该电子装置包括：存储器、处理器，所述存储器上存储有基于舆情因子的投资选股程序，所述基于舆情因子的投资选股程序被所述处理器执行时实现以下步骤：
    获取预先确定的多只股票在第一预设时间内的舆情因子观测值及第二预设时间内的收益率；
    根据所述舆情因子观测值及收益率，计算各舆情因子的第一信息系数；
    分别将各舆情因子的第一信息系数输入预先训练好的信息系数预测模型中，预测未来各舆情因子的第二信息系数；
    根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分；及
    根据评分的高低顺序对所述多只股票进行排序，筛选出排序靠前的第一预设数量的股票，作为目标投资股票。
  9. 根据权利要求8所述的电子装置，其特征在于，所述第一信息系数的计算公式为：
    Figure PCTCN2018102127-appb-100004
    其中，
    Figure PCTCN2018102127-appb-100005
    为t-1时刻各股票的舆情因子f的观测值排名，
    Figure PCTCN2018102127-appb-100006
    为t时刻各股票的收益率排名。
  10. 根据权利要求9所述的电子装置，其特征在于，所述预先确定的信息系数预测模型的训练步骤包括：
    分别采集预先确定的第二预设数量的股票在第三预设时间内各舆情因子在每个时刻的历史观测值、及各只股票在每个时刻的历史收益率，计算各舆情因子在每个时刻对应的信息系数，以获取样本数据；及
    将样本数据划分为训练集、评估集和测试集，并利用训练集的样本数据对神经网络进行训练，得到信息系数预测模型，利用测试集的样本数据对信息系数预测模型的准确率进行测试，得到最终的信息系数预测模型。
  11. 根据权利要求9所述的电子装置，其特征在于，所述“根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分”的步骤包括：
    根据各舆情因子对应的第二信息系数，分别确定所述M只股票的各舆情因子对应的权重；及
    根据所述舆情因子观测值及各舆情因子所占的权重，分别计算每只股票的评分。
  12. 根据权利要求11所述的电子装置，其特征在于，所述预先确定的信息系数预测模型的训练步骤包括：
    分别采集预先确定的第二预设数量的股票在第三预设时间内各舆情因子在每个时刻的历史观测值、及各只股票在每个时刻的历史收益率，计算各舆情因子在每个时刻对应的信息系数，以获取样本数据；及
    将样本数据划分为训练集、评估集和测试集，并利用训练集的样本数据 对神经网络进行训练，得到信息系数预测模型，利用测试集的样本数据对信息系数预测模型的准确率进行测试，得到最终的信息系数预测模型。
  13. 根据权利要求11所述的电子装置，其特征在于，所述各只股票的评分的计算公式为：
    S＝Σω i*X i
    其中，i为股票的舆情因子，ω i为股票的舆情因子i对应的权重，X i为t时刻股票的各舆情因子的观测值。
  14. 根据权利要求13所述的电子装置，其特征在于，所述预先确定的信息系数预测模型的训练步骤包括：
    分别采集预先确定的第二预设数量的股票在第三预设时间内各舆情因子在每个时刻的历史观测值、及各只股票在每个时刻的历史收益率，计算各舆情因子在每个时刻对应的信息系数，以获取样本数据；及
    将样本数据划分为训练集、评估集和测试集，并利用训练集的样本数据对神经网络进行训练，得到信息系数预测模型，利用测试集的样本数据对信息系数预测模型的准确率进行测试，得到最终的信息系数预测模型。
  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上存储有基于舆情因子的投资选股程序，所述基于舆情因子的投资选股程序被处理器执行时实现如下步骤：
    获取预先确定的多只股票在第一预设时间内的舆情因子观测值及第二预设时间内的收益率；
    根据所述舆情因子观测值及收益率，计算各舆情因子的第一信息系数；
    分别将各舆情因子的第一信息系数输入预先训练好的信息系数预测模型中，预测未来各舆情因子的第二信息系数；
    根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分；及
    根据评分的高低顺序对所述多只股票进行排序，筛选出排序靠前的第一预设数量的股票，作为目标投资股票。
  16. 根据权利要求15所述的计算机可读存储介质，其特征在于，所述第一信息系数的计算公式为：
    Figure PCTCN2018102127-appb-100007
    其中，
    Figure PCTCN2018102127-appb-100008
    为t-1时刻各股票的舆情因子f的观测值排名，
    Figure PCTCN2018102127-appb-100009
    为t时刻各股票的收益率排名。
  17. 根据权利要求16所述的计算机可读存储介质，其特征在于，所述预先确定的信息系数预测模型的训练步骤包括：
    分别采集预先确定的第二预设数量的股票在第三预设时间内各舆情因子在每个时刻的历史观测值、及各只股票在每个时刻的历史收益率，计算各舆情因子在每个时刻对应的信息系数，以获取样本数据；及
    将样本数据划分为训练集、评估集和测试集，并利用训练集的样本数据对神经网络进行训练，得到信息系数预测模型，利用测试集的样本数据对信息系数预测模型的准确率进行测试，得到最终的信息系数预测模型。
  18. 根据权利要求16所述的计算机可读存储介质，其特征在于，所述“根据各舆情因子观测值及各舆情因子的第二信息系数，分别计算所述多只股票的评分”的步骤包括：
    根据各舆情因子对应的第二信息系数，分别确定所述M只股票的各舆情因子对应的权重；及
    根据所述舆情因子观测值及各舆情因子所占的权重，分别计算每只股票的评分。
  19. 如权利要求18所述的计算机可读存储介质，其特征在于，所述各只股票的评分的计算公式为：
    S＝Σω i*X i
    其中，i为股票的舆情因子，ω i为股票的舆情因子i对应的权重，X i为t时刻股票的各舆情因子的观测值。
  20. 根据权利要求19所述的计算机可读存储介质，其特征在于，所述预先确定的信息系数预测模型的训练步骤包括：
    分别采集预先确定的第二预设数量的股票在第三预设时间内各舆情因子在每个时刻的历史观测值、及各只股票在每个时刻的历史收益率，计算各舆情因子在每个时刻对应的信息系数，以获取样本数据；及
    将样本数据划分为训练集、评估集和测试集，并利用训练集的样本数据对神经网络进行训练，得到信息系数预测模型，利用测试集的样本数据对信息系数预测模型的准确率进行测试，得到最终的信息系数预测模型。
PCT/CN2018/102127 2018-04-26 2018-08-24 基于舆情因子的投资选股方法、装置及存储介质 Ceased WO2019205378A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN2018103823117 2018-04-26
CN201810382311.7A CN108710994A (zh) 2018-04-26 2018-04-26 基于舆情因子的投资选股方法、装置及存储介质

Publications (1)

Publication Number Publication Date
WO2019205378A1 true WO2019205378A1 (zh) 2019-10-31

Family

ID=63867391

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/102127 Ceased WO2019205378A1 (zh) 2018-04-26 2018-08-24 基于舆情因子的投资选股方法、装置及存储介质

Country Status (2)

Country Link
CN (1) CN108710994A (zh)
WO (1) WO2019205378A1 (zh)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109584076A (zh) * 2018-12-06 2019-04-05 平安科技（深圳）有限公司 股票因子收益分析方法、装置、计算机设备以及存储介质
CN115423621A (zh) * 2022-08-05 2022-12-02 中国银联股份有限公司 一种交易异常检测方法、装置、电子设备及存储介质
CN116522917B (zh) * 2023-03-06 2024-01-26 北京麦克斯泰科技有限公司 舆情信息热度评分方法、装置、计算机设备和存储介质
TWI869983B (zh) * 2023-08-21 2025-01-11 藍羿閔 一種建構投資組合之方法
CN118735691A (zh) * 2024-09-03 2024-10-01 上海大智慧信息科技有限公司 基于资产监控与动态平衡的智能化投后管理系统及方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105373853A (zh) * 2015-11-18 2016-03-02 百度在线网络技术（北京）有限公司 股票舆情指数预测方法以及装置
CN106022522A (zh) * 2016-05-20 2016-10-12 南京大学 一种基于互联网公开的大数据预测股票的方法及系统
CN107403017A (zh) * 2017-08-09 2017-11-28 上海数旦信息技术有限公司 一种智能分析实时新闻对金融市场影响的方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105373853A (zh) * 2015-11-18 2016-03-02 百度在线网络技术（北京）有限公司 股票舆情指数预测方法以及装置
CN106022522A (zh) * 2016-05-20 2016-10-12 南京大学 一种基于互联网公开的大数据预测股票的方法及系统
CN107403017A (zh) * 2017-08-09 2017-11-28 上海数旦信息技术有限公司 一种智能分析实时新闻对金融市场影响的方法

Also Published As

Publication number Publication date
CN108710994A (zh) 2018-10-26

Similar Documents

Publication Publication Date Title
WO2019205378A1 (zh) 基于舆情因子的投资选股方法、装置及存储介质
EP4016432A1 (en) Method and apparatus for training fusion ordering model, search ordering method and apparatus, electronic device, storage medium, and program product
CN113742492B (zh) 保险方案生成方法、装置、电子设备及存储介质
US8924409B1 (en) Presentation of match quality in auto-complete suggestions
US9002759B2 (en) Data mining technique with maintenance of fitness history
US9208251B2 (en) Computer-based evaluation tool for selecting personalized content for users
US10509837B2 (en) Modeling actions for entity-centric search
US20110093457A1 (en) Method for calculating resource points of resource information and distributing points
Bothos et al. Using social media to predict future events with agent-based markets
AU2018101512A4 (en) A comprehensive stock trend predicting method based on neural networks
WO2014134155A1 (en) Candidate and job trending in employment system
US10191985B1 (en) System and method for auto-curation of Q and A websites for search engine optimization
US20220091713A1 (en) Systems and methods for generating dynamic interface options using machine learning models
CN114819967A (zh) 数据处理方法、装置、电子设备及计算机可读存储介质
CN111061948A (zh) 一种用户标签推荐方法、装置、计算机设备及存储介质
WO2019019346A1 (zh) 资产配置策略获取方法、装置、计算机设备和存储介质
EP2648147A1 (en) Inherited user rating
CN118446783B (zh) 推荐策略确定方法、装置、计算机设备、存储介质和程序产品
CN107809370B (zh) 用户推荐方法及装置
JP7777856B2 (ja) インフルエンサー選定支援システム、インフルエンサー選定支援方法、及び、コンピュータプログラム
US20160171608A1 (en) Methods and systems for finding similar funds
US20150235129A1 (en) Systems, methods and non-transitory computer readable storage media for tracking and evaluating predictions regarding relationships
WO2016203652A1 (ja) データ分析に係るシステム、制御方法、制御プログラム、および、その記録媒体
AU2018100320A4 (en) A New System for Stock Volatility Prediction by Using Long Short-Term Memory with Sentimental Indicators
CN105593890A (zh) 预测性首次公开募股分析

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: 18916687

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 05/02/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 18916687

Country of ref document: EP

Kind code of ref document: A1