CN114022171A - Method and system for predicting highest price of short-term stock - Google Patents
Method and system for predicting highest price of short-term stock Download PDFInfo
- Publication number
- CN114022171A CN114022171A CN202111079576.8A CN202111079576A CN114022171A CN 114022171 A CN114022171 A CN 114022171A CN 202111079576 A CN202111079576 A CN 202111079576A CN 114022171 A CN114022171 A CN 114022171A
- Authority
- CN
- China
- Prior art keywords
- variables
- components
- stock
- highest price
- imf
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0206—Price or cost determination based on market factors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
- G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2134—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis
- G06F18/21342—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis using statistical independence, i.e. minimising mutual information or maximising non-gaussianity
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Marketing (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- General Engineering & Computer Science (AREA)
- Operations Research (AREA)
- Technology Law (AREA)
- Human Resources & Organizations (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a method and a system for predicting the highest price of a short-term stock, wherein in the prediction method, a financial time sequence is decomposed by a CEEMDAN algorithm to obtain IMF components and residual components RES, the IMF components are denoised by an SSA algorithm, and finally the short-term highest price of the stock of an enterprise to be predicted is predicted by training a TCN (train control network) which corresponds to each component and has an attention mechanism; the invention also provides a prediction system based on the prediction method, which comprises a data acquisition module, a data decomposition module, a variable prediction module and a highest price output module and is used for realizing the prediction method. The invention combines CEEMDAN and SSA algorithm, uses TCN deep learning model and attention mechanism to capture the variation trend of the stock maximum price, improves the precision of stock maximum price prediction, and has good generalization performance.
Description
Technical Field
The invention belongs to the field of stock price prediction, and particularly relates to a method and a system for predicting the highest price of short-term stocks.
Background
With the development of the stock market and computer technology in China, more and more scholars explore the changing trend of stock prices based on deep learning technology. In the conventional stock price prediction technology, the change of stocks is assumed to be a linear process, but in reality, the stock price change characteristic is nonlinear. With the subsequent development of deep learning, a plurality of models for processing time series data are generated, and the models solve the characteristic of non-linearity of stock price change on one hand and can memorize the characteristic of change trend of the stock price on the other hand. However, most of the stock forecast prices in research are closing prices and also long-term variation trends of the closing prices, and for some scattered households, the highest price and short-term variation trends of the stocks are more concerned.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a method and a system for predicting the highest price of short-term stocks, which adopt the following technical scheme:
a method for predicting the highest price of short-term stocks comprises the following steps:
step 1: acquiring daily highest price data of each enterprise stock within a certain time to form a financial time sequence of the highest price of each enterprise stock;
step 2: performing CEEMDAN decomposition on the financial time sequence with the highest price of each enterprise stock to obtain IMF components and residual components RES of the financial time sequence;
and step 3: deleting IMF components irrelevant to the financial time sequence and denoising the residual IMF components;
and 4, step 4: standardizing the IMF components and the residual components RES after noise reduction, and respectively dividing characteristic variables and target variables to obtain the characteristic variables and the target variables of the IMF components and the characteristic variables and the target variables of the residual components RES;
and 5: training a TCN network for predicting the IMF components by using the characteristic variables and the target variables of the IMF components, and training the TCN network for predicting the residual components RES by using the characteristic variables and the target variables of the residual components RES;
step 6: extracting characteristic variables of IMF components and characteristic variables of residual components RES of financial time sequences of the highest prices of stocks of the enterprise to be predicted, inputting the extracted characteristic variables of the IMF components into the corresponding TCN network which is trained to obtain the prediction variables of the IMF components, and inputting the extracted characteristic variables of the residual components RES into the corresponding TCN network which is trained to obtain the prediction variables of the residual components RES;
and 7: and superposing the forecasting variables of the IMF components and the forecasting variables of the residual components RES to obtain a predicted financial time sequence of the highest price, so as to obtain a short-term highest price predicted value of the stock of the enterprise to be predicted.
Further, in step 3, the residual IMF component is denoised by the SSA algorithm.
Further, in step 4, when dividing the characteristic variables and the target variables, respectively dividing the normalized IMF components and residual components RES into a plurality of segment sequences with the same length according to time sequence, where a former segment sequence is used as the characteristic variable of an immediately succeeding segment sequence, and a latter segment sequence is used as the target variable of an immediately preceding segment sequence, so as to construct a plurality of groups of data samples.
Further, in step 5 and step 6, before the characteristic variables are input into the corresponding TCN networks, the characteristic variables are weighted in time sequence by the attention mechanism, and the weight value of the characteristic variables is increased according to the value closer to the current time point.
The short-term stock highest price forecasting system based on the forecasting method comprises a data acquisition module, a data decomposition module, a variable forecasting module and a highest price output module; the data acquisition module is used for inputting daily highest price data of the stock of the enterprise to be predicted within a certain period of time so as to generate a financial time sequence of the highest price; the data decomposition module is used for carrying out CEEMDAN decomposition on the financial time sequence generated by the data acquisition module to obtain an IMF component and a residual component RES, screening and denoising the IMF component, and standardizing the IMF component and the residual component RES after denoising to generate corresponding characteristic variables; the variable prediction module comprises each trained TCN network, and each TCN network outputs a corresponding prediction variable after the characteristic variable generated by the data classification module is input into the corresponding TCN network; and the highest price output module mutually superposes the prediction variables output by the variable prediction module to obtain and output a predicted financial time sequence, so that the short-term highest price prediction value of the stock of the enterprise to be predicted is obtained according to the predicted financial time sequence.
The method captures the characteristics of the financial time sequence with the highest price of the stock through the CEEMDAN decomposition algorithm, reduces the noise of the financial time sequence through the SSA algorithm, captures the variation trend of the highest price of the stock by using the TCN deep learning model and the attention mechanism, improves the precision of the prediction of the highest price of the stock, and has good generalization performance.
Drawings
FIG. 1 is a flow chart of a prediction method of the present invention;
FIG. 2 is a schematic diagram of the IMF component and the residual RES component of CEEMDAN decomposition.
Detailed Description
The present invention will now be described in further detail with reference to the accompanying drawings.
As shown in fig. 1, the prediction method of the present invention mainly includes the following steps:
(1) the daily maximum price of an enterprise stock that has been on the market in the stock market in china for more than 10 years is collected as the original financial time series data of the decomposition algorithm.
(2) The original financial time-series data is decomposed using the CEEMDAN algorithm to obtain an IMF component and a residual component RES, and as shown in FIG. 2, the obtained IMF component can be generally divided into a high-frequency component, a low-frequency component, and a trend component.
(3) And (4) screening IMF components, calculating the correlation size between each IMF component and the original financial time sequence, and deleting the IMF components which are obviously irrelevant to the fluctuation of the original financial time sequence.
(4) And decomposing and recombining the residual IMF components by utilizing an SSA algorithm to obtain the IMF components after noise reduction.
The CEEMDAN algorithm can better decompose the sequence into the sum of a limited number of intrinsic mode functions (IMF components) and a residue according to the characteristics of the original financial time sequence, has the advantage of well separating high-frequency noise in the original sequence, but can generate false IMF components along with the gradual increase of the decomposition layer number, so that a better decomposition result cannot be obtained on the low-frequency part of the sequence. However, the SSA algorithm may decompose the IMF component into a sum of a plurality of mutually uncorrelated components to achieve the purpose of denoising, and then select a proper decomposition order to reconstruct the IMF component to obtain the denoised IMF component.
(5) Normalizing the residual component RES and the denoised IMF component; the purpose of this is to eliminate the influence of magnitude between data on one hand and to accelerate the convergence speed of the model on the other hand. And after standardization, respectively dividing the characteristic variables and the target variables to obtain the characteristic variables and the target variables of the IMF components and the characteristic variables and the target variables of the residual components RES.
When the characteristic variables and the target variables are divided, the standardized IMF components and the standardized residual components RES are respectively cut into a plurality of segment sequences with the same length according to time sequence, the former segment sequence is used as the characteristic variable of the next adjacent later segment sequence, and the latter segment sequence is used as the target variable of the next adjacent previous segment sequence, so that a plurality of groups of data samples are constructed for training the TCN deep network corresponding to each component.
(6) And training a TCN network for predicting the IMF component by using the characteristic variables and the target variables of the IMF component, and training the TCN network for predicting the residual component RES by using the characteristic variables and the target variables of the residual component RES. Before the characteristic variables are input into the corresponding TCN network, the characteristic variables are weighted in time sequence through the attention mechanism, the weighting of the attention mechanism is weighting in time, and the weight value added to the characteristic variables is larger when the value is closer to the current time point.
(7) Extracting the characteristic variables of IMF components and the characteristic variables of residual components RES of financial time sequences of the highest prices of stocks of the enterprise to be predicted, inputting the extracted characteristic variables of the IMF components into the corresponding TCN networks which are trained to obtain the prediction variables of the IMF components, and inputting the extracted characteristic variables of the residual components RES into the corresponding TCN networks which are trained to obtain the prediction variables of the residual components RES.
(8) And superposing the forecasting variables of the IMF components and the forecasting variables of the residual components RES to obtain a predicted financial time sequence of the highest price, so as to obtain a short-term highest price predicted value of the stock of the enterprise to be predicted.
The invention also provides a short-term stock maximum price forecasting system which comprises a data acquisition module, a data decomposition module, a variable forecasting module and a maximum price output module. The system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for inputting daily highest price data of the stock of an enterprise to be predicted within a certain period of time in the near future so as to generate a financial time sequence of the highest price; the data decomposition module is used for carrying out CEEMDAN decomposition on the financial time sequence generated by the data acquisition module to obtain an IMF component and a residual component RES, screening and denoising the IMF component, and standardizing the IMF component and the residual component RES after denoising to generate corresponding characteristic variables; the variable prediction module comprises each trained TCN network, and each TCN network outputs a corresponding prediction variable after the characteristic variable generated by the data classification module is input into the corresponding TCN network; and the highest price output module mutually superposes the prediction variables output by the variable prediction module to obtain and output a predicted financial time sequence, so that the short-term highest price prediction value of the stock of the enterprise to be predicted is obtained according to the predicted financial time sequence.
The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may be made by those skilled in the art without departing from the principle of the invention.
Claims (5)
1. A method for predicting the highest price of short-term stocks is characterized by comprising the following steps:
step 1: acquiring daily highest price data of each enterprise stock within a certain time to form a financial time sequence of the highest price of each enterprise stock;
step 2: performing CEEMDAN decomposition on the financial time sequence with the highest price of each enterprise stock to obtain IMF components and residual components RES of the financial time sequence;
and step 3: deleting IMF components irrelevant to the financial time sequence and denoising the residual IMF components;
and 4, step 4: standardizing the IMF components and the residual components RES after noise reduction, and respectively dividing characteristic variables and target variables to obtain the characteristic variables and the target variables of the IMF components and the characteristic variables and the target variables of the residual components RES;
and 5: training a TCN network for predicting the IMF components by using the characteristic variables and the target variables of the IMF components, and training the TCN network for predicting the residual components RES by using the characteristic variables and the target variables of the residual components RES;
step 6: extracting characteristic variables of IMF components and characteristic variables of residual components RES of financial time sequences of the highest prices of stocks of the enterprise to be predicted, inputting the extracted characteristic variables of the IMF components into the corresponding TCN network which is trained to obtain the prediction variables of the IMF components, and inputting the extracted characteristic variables of the residual components RES into the corresponding TCN network which is trained to obtain the prediction variables of the residual components RES;
and 7: and superposing the forecasting variables of the IMF components and the forecasting variables of the residual components RES to obtain a predicted financial time sequence of the highest price, so as to obtain a short-term highest price predicted value of the stock of the enterprise to be predicted.
2. The method for predicting the short term stock peak of claim 1, wherein in step 3, the residual IMF component is denoised by SSA algorithm.
3. The method for predicting the short-term stock maximum price according to claim 1, wherein in the step 4, when the characteristic variables and the target variables are divided, the normalized IMF components and the normalized residual components RES are respectively divided into a plurality of segment sequences with the same length according to time sequence, the former segment sequence is used as the characteristic variable of the next adjacent next segment sequence, and the latter segment sequence is used as the target variable of the next adjacent previous segment sequence, so as to construct the plurality of groups of data samples.
4. The method of claim 1, wherein the characteristic variables are weighted in time sequence by the attention mechanism before being inputted into the corresponding TCN network in steps 5 and 6, and the weight value of the characteristic variables is increased as the value is closer to the current time point.
5. The short-term stock highest price forecasting system based on the forecasting method of any one of claims 1 to 4 is characterized by comprising a data acquisition module, a data decomposition module, a variable forecasting module and a highest price output module; the data acquisition module is used for inputting daily highest price data of the stock of the enterprise to be predicted within a certain period of time so as to generate a financial time sequence of the highest price; the data decomposition module is used for carrying out CEEMDAN decomposition on the financial time sequence generated by the data acquisition module to obtain an IMF component and a residual component RES, screening and denoising the IMF component, and standardizing the IMF component and the residual component RES after denoising to generate corresponding characteristic variables; the variable prediction module comprises each trained TCN network, and each TCN network outputs a corresponding prediction variable after the characteristic variable generated by the data classification module is input into the corresponding TCN network; and the highest price output module mutually superposes the prediction variables output by the variable prediction module to obtain and output a predicted financial time sequence, so that the short-term highest price prediction value of the stock of the enterprise to be predicted is obtained according to the predicted financial time sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111079576.8A CN114022171A (en) | 2021-09-15 | 2021-09-15 | Method and system for predicting highest price of short-term stock |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111079576.8A CN114022171A (en) | 2021-09-15 | 2021-09-15 | Method and system for predicting highest price of short-term stock |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114022171A true CN114022171A (en) | 2022-02-08 |
Family
ID=80054109
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111079576.8A Pending CN114022171A (en) | 2021-09-15 | 2021-09-15 | Method and system for predicting highest price of short-term stock |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114022171A (en) |
-
2021
- 2021-09-15 CN CN202111079576.8A patent/CN114022171A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107292446B (en) | Hybrid wind speed prediction method based on component relevance wavelet decomposition | |
| CN110212528B (en) | Power distribution network measurement data missing reconstruction method | |
| CN114120041B (en) | Small sample classification method based on double-countermeasure variable self-encoder | |
| CN112668611B (en) | Kmeans and CEEMD-PE-LSTM-based short-term photovoltaic power generation power prediction method | |
| CN112232604B (en) | Prediction method for extracting network traffic based on Prophet model | |
| Faraz et al. | Stock market prediction-by-prediction based on autoencoder long short-term memory networks | |
| CN117596191A (en) | Power Internet of things abnormality detection method, device and storage medium | |
| CN115883424B (en) | Method and system for predicting flow data between high-speed backbone networks | |
| CN117520809A (en) | Transformer fault diagnosis method based on EEMD-KPCA-CNN-BiLSTM | |
| CN110740063A (en) | Network flow characteristic index prediction method based on signal decomposition and periodic characteristics | |
| CN115733673B (en) | Data anomaly detection method based on multi-scale residual error classifier | |
| CN114022171A (en) | Method and system for predicting highest price of short-term stock | |
| CN114781450B (en) | Train rolling bearing state identification method based on parameter optimization MOMEDA-MIA-CNN | |
| Fu et al. | PQEventCog: Classification of power quality disturbances based on optimized S-transform and CNNs with noisy labeled datasets | |
| CN110705638A (en) | Credit rating prediction classification method using deep network learning fuzzy information feature technology | |
| CN116911419A (en) | Long time sequence prediction method based on trend correlation feature learning | |
| CN116561569A (en) | Industrial power load identification method based on EO feature selection and AdaBoost algorithm | |
| Yuan et al. | A novel hybrid short-term wind power prediction framework based on singular spectrum analysis and deep belief network utilized improved adaptive genetic algorithm | |
| Yu et al. | Time series cross-correlation network for wind power prediction | |
| CN112616160A (en) | Intelligent short-wave frequency cross-frequency-band real-time prediction method and system | |
| Mengjiao et al. | Wind turbine bearing fault diagnosis method based on multi-domain feature extraction | |
| CN117952646A (en) | TCN-LSTM combined electricity price prediction method based on wavelet packet decomposition | |
| Lin et al. | Data Augmentation for Industrial Multivariate Time Series via a Spatial and Frequency Domain Knowledge GAN | |
| CN117473275B (en) | Energy consumption detection method for data center | |
| You et al. | An Improved Long Short-Term Memory Neural Network Wind Power Forecast Algorithm Based on TEMD Noise Reduction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |