WO2023159756A1 - 价格数据的处理方法和装置、电子设备、存储介质 - Google Patents

价格数据的处理方法和装置、电子设备、存储介质 Download PDF

Info

Publication number
WO2023159756A1
WO2023159756A1 PCT/CN2022/090661 CN2022090661W WO2023159756A1 WO 2023159756 A1 WO2023159756 A1 WO 2023159756A1 CN 2022090661 W CN2022090661 W CN 2022090661W WO 2023159756 A1 WO2023159756 A1 WO 2023159756A1
Authority
WO
WIPO (PCT)
Prior art keywords
target
data
reference value
processing
report
Prior art date
Application number
PCT/CN2022/090661
Other languages
English (en)
French (fr)
Inventor
刘羲
舒畅
陈又新
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 WO2023159756A1 publication Critical patent/WO2023159756A1/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
    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Definitions

  • the present application relates to the technical field of artificial intelligence and macroeconomics, and in particular to a method and device for processing price data, electronic equipment, and a storage medium.
  • the machine learning model represented by linear regression is used to predict the price data
  • the original data input into the machine learning model for prediction is a kind of nonlinear data, through which the machine
  • the prediction of price data by the learning model can easily lead to inaccurate estimation and prediction results. Therefore, how to improve the accuracy of price data prediction has become an urgent technical problem to be solved.
  • the embodiment of the present application proposes a method for processing price data, the method comprising:
  • the raw data includes target report data and target transaction data;
  • the embodiment of the present application proposes a price data processing device, the device comprising:
  • An acquisition module configured to acquire raw data to be predicted; wherein, the raw data includes target report data and target transaction data;
  • the first building module is used to build indicator factor features according to the target transaction data
  • the second building block is used to construct public opinion factor features according to the target report data
  • a screening module configured to perform screening processing on a plurality of said index factor characteristics and a plurality of said public opinion factor characteristics to obtain a plurality of quantitative transaction characteristics
  • a feature extraction module configured to perform feature extraction on multiple quantitative transaction features through the preset first neural network model to obtain multiple distributed feature vectors
  • the forecasting module is used to input a plurality of the distributed feature vectors into the preset second neural network model to perform price forecasting processing to obtain target price data.
  • the embodiment of the present application provides an electronic device, including:
  • the program is stored in the memory, and the processor executes the at least one program to realize the processing method of price data; wherein, the processing method of price data includes: obtaining the original data to be predicted; wherein, the original The data includes target report data and target transaction data; constructing index factor features according to the target transaction data; constructing public opinion factor features according to the target report data; screening a plurality of the index factor features and a plurality of the public opinion factor features processing to obtain multiple quantitative transaction features; perform feature extraction on multiple quantitative transaction features through the preset first neural network model to obtain multiple distributed feature vectors; input multiple distributed feature vectors to the preset In the second neural network model set up, the price prediction process is carried out to obtain the target price data.
  • the embodiment of the present application provides a storage medium, the storage medium is a computer-readable storage medium, and the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to make the computer Executing a processing method for price data; wherein, the processing method for price data includes: obtaining raw data to be predicted; wherein, the raw data includes target report data and target transaction data; constructing index factor features according to the target transaction data Construct public opinion factor features according to the target report data; filter and process a plurality of said index factor features and a plurality of said public opinion factor features to obtain multiple quantitative transaction features; performing feature extraction on each of the quantified transaction features to obtain multiple distributed feature vectors; inputting the multiple distributed feature vectors into a preset second neural network model for price prediction processing to obtain target price data.
  • the embodiment of the present application proposes a price data processing method and device, electronic equipment, and storage media.
  • the accuracy of price data prediction is improved, and through a screening mechanism, it is avoided when quantitative transaction features are used.
  • the problem of excessive irrelevant data during feature extraction, and predicting the price data through the first neural network model and the second neural network model, improve the speed of price data prediction and the accuracy of price data prediction.
  • Fig. 1 is the flowchart of the processing method of the price data that the embodiment of the present application provides;
  • Fig. 2 is a specific method flowchart of step S300 in Fig. 1;
  • Fig. 3 is a specific method flowchart of step S320 in Fig. 2;
  • Fig. 4 is the flow chart of the specific method of step S500 in Fig. 1;
  • FIG. 5 is a flowchart of a specific method of step S530 in FIG. 4;
  • FIG. 6 is a flowchart of a specific method of step S600 in FIG. 1;
  • FIG. 7 is a block diagram of a device for processing price data provided in an embodiment of the present application.
  • FIG. 8 is a schematic diagram of a hardware structure of an electronic device provided by an embodiment of the present application.
  • Artificial Intelligence It is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence; artificial intelligence is a branch of computer science. Intelligence attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a manner similar to human intelligence. Research in this field includes robotics, language recognition, image recognition, natural language processing, and expert systems. Artificial intelligence can simulate the information process of human consciousness and thinking. Artificial intelligence is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
  • Natural language processing uses computers to process, understand and use human languages (such as Chinese, English, etc.). NLP belongs to a branch of artificial intelligence and is an interdisciplinary subject between computer science and linguistics. Known as computational linguistics. Natural language processing includes syntax analysis, semantic analysis, text understanding, etc. Natural language processing is often used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis and opinion mining. Deal with related data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research and linguistics research related to language computing, etc.
  • CNN Convolutional Neural Networks
  • a convolutional neural network is a feedforward neural network that consists of several convolutional and pooling layers.
  • the basic structure of CNN consists of an input layer, a convolutional layer, a pooling layer (also called a sampling layer), a fully connected layer, and an output layer.
  • a convolutional layer and pooling layers are used, and the convolutional layer and the pooling layer are alternately set, that is, a convolutional layer is connected to a pooling layer, and the pooling layer is connected to a convolutional layer, and so on.
  • CNN Since each neuron of the output feature map in the convolutional layer is locally connected to its input, and the weighted sum of the corresponding connection weight and the local input is added to the bias value to obtain the input value of the neuron, the process is equivalent to Because of the convolution process, CNN is also named after it.
  • the convolutional neural network is evolved from the multi-layer perceptron (MLP). Due to its structural characteristics of local area connection, weight sharing, and downsampling, the convolutional neural network performs well in the field of image processing. Compared with other neural networks, the particularity of convolutional neural network mainly lies in two aspects of weight sharing and local connection. Weight sharing makes the network structure of convolutional neural network more similar to biological neural network.
  • Local connections are not like traditional neural networks, where each neuron in layer n-1 is connected to all neurons in layer n, but between neurons in layer n-1 and some neurons in layer n connect.
  • the role of these two features is to reduce the complexity of the network model and reduce the number of weights.
  • Zero padding Padding the edges of the input matrix with zero values allows us to filter the edges of the input image matrix.
  • One of the great benefits of zero padding is that it allows us to control the size of the feature maps.
  • the use of zero padding is also called general convolution, and the use of zero padding is called strict convolution.
  • Recurrent Neural Network is a kind of recursive neural network (recursive neural network) that takes sequence data as input, performs recursion in the evolution direction of the sequence, and all nodes (circular units) are connected in chains. network).
  • LSTM Long-Short Term Memory
  • RNN cyclic neural network
  • All RNNs are Has a chained form of repeating neural network modules.
  • LSTM applications include: text generation, machine translation, speech recognition, image description generation, and video tagging.
  • Word segmentation processing is to automatically add spaces or other boundary marks between words in the text.
  • English words are naturally separated by spaces, and it is easy to divide words according to spaces, but sometimes it is necessary to treat multiple words as one word, such as some nouns such as "New York", which need to be treated as one word. Since there are no spaces in Chinese, word segmentation is a problem that needs to be specially solved. Whether it is English or Chinese, the principle of word segmentation is similar.
  • Chinese automatic word segmentation is to let the computer system automatically add spaces or other boundary marks between words in the Chinese text.
  • a commonly used Chinese word segmentation tool is Jieba.
  • the embodiments of the present application provide a price data processing method and device, electronic equipment, and a storage medium, which can improve the accuracy of price data prediction by acquiring various raw data to predict price trends.
  • the price data processing method and device, electronic equipment, and storage medium provided in the embodiments of the present application are specifically described through the following embodiments. First, the price data processing method in the embodiments of the present application is described.
  • AI artificial intelligence
  • the embodiments of the present application may acquire and process relevant data based on artificial intelligence technology.
  • artificial intelligence is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. .
  • Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation/interaction systems, and mechatronics.
  • Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning/deep learning.
  • the price data processing method provided in the embodiment of the present application relates to the technical field of artificial intelligence.
  • the price data processing method provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server, and can also be software running on the terminal or the server.
  • the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, etc.
  • the server end can be configured as an independent physical server, or can be configured as a server cluster or a distributed system composed of multiple physical servers, or It can be configured as a cloud that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.
  • the server; the software may be an application for realizing the processing method of price data, etc., but is not limited to the above forms.
  • the application can be used in numerous general purpose or special purpose computer system environments or configurations. Examples: personal computers, server computers, handheld or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, including A distributed computing environment for any of the above systems or devices, etc.
  • This application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer.
  • program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types.
  • the application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network.
  • program modules may be located in both local and remote computer storage media including storage devices.
  • Fig. 1 is an optional flow chart of the processing method of price data provided by the embodiment of the present application.
  • the method in Fig. 1 may include but not limited to steps S100 to S600. The six steps will be described in detail below in conjunction with Fig. 1 introduce.
  • Step S100 obtaining raw data to be predicted; wherein, the raw data includes target report data and target transaction data;
  • Step S200 constructing index factor features according to the target transaction data
  • Step S300 constructing public opinion factor features according to the target report data
  • Step S400 performing screening processing on multiple index factor features and multiple public opinion factor features to obtain multiple quantitative transaction features
  • Step S500 extract multiple quantitative transaction features through the preset first neural network model, and obtain multiple distributed feature vectors
  • Step S600 input multiple distributed feature vectors into the preset second neural network model to perform price prediction processing, and obtain target price data.
  • the method for processing price data in the embodiment of the present application obtains the original data to be predicted; wherein, the original data includes target report data and target transaction data, and then constructs index factor features based on target transaction data, and constructs public opinion factor features based on target report data , and then filter and process the obtained multiple index factor features and multiple public opinion factor features to obtain multiple quantitative transaction features, and then perform feature extraction on multiple quantitative transaction features through the preset first neural network model to obtain multiple Distributed eigenvectors. Finally, multiple distributed eigenvectors are input into the preset second neural network model for price prediction processing to obtain target price data.
  • the accuracy of price data prediction is improved, and, through the screening mechanism, the problem of too much irrelevant data in the feature extraction of quantitative trading features is avoided, and the accuracy of price data prediction is further improved.
  • the technical solution of the embodiment of the present application predicts the price data by combining the first neural network model and the second neural network model, which improves the speed of price data prediction and the accuracy of price data prediction.
  • the original data can be obtained by writing a web crawler, setting a data source, and crawling data with a goal. It is also possible to conduct data query through some public websites to obtain raw data.
  • the technical solution of the embodiment of the present application can realize the price data prediction of a stock in the stock market, a fund in the fund market, or a future in the futures market, etc., and, compared with the prior art, adopt a machine learning model
  • the prediction method, the target price data obtained by the price data processing method in the embodiment of the present application is more accurate.
  • multiple corresponding target price data are obtained through multiple price predictions on a certain stock, a fund in the fund market, or a future in the futures market, thereby obtaining the price trend of the stock, fund or futures.
  • the object to be predicted is a stock in the stock market, that is, the technical solution of the embodiment of the present application is to predict the price data of a stock in the stock market, then in the application scenario of stock price data prediction, the original The way of data can be to go to the website of the company corresponding to the stock to be tested to check some target report data, and go to the trading platform to check the target report data and target transaction data.
  • the target report data includes at least one of: target industry report data, target company report data, target news data, and target comment data.
  • the target transaction data includes at least one of the following: opening price, closing price, highest price, lowest price, and trading volume.
  • the original data is obtained by acquiring target report data and target transaction data, and then comprehensively considers the impact of the target report data and target transaction data on the future price trend, thereby improving the accuracy of price data prediction.
  • target report data includes but is not limited to at least one of the following: stock A corresponds to the company's industry research report data (ie, target industry report data), stock A corresponds to the company's research report data (ie, target company report data), News data related to stock A company (ie target news data) and comment data on stock A in the stock bar (ie target comment data).
  • the target transaction data includes but is not limited to at least one of the following: the historical opening price of stock A, the historical closing price of stock A, the historical highest price of stock A, the historical lowest price of stock A, the historical trading volume of stock A, and the historical closing price of stock A. Current price.
  • index factor features includes index factor construction, and index factors include but are not limited to at least one of the following: OBV (On Balance Volume, energy tide) factor, CCI (commoditychannelindex, homeopathic index) factor, KDJ (Stochastic indicators) factors and a series of technical indicators.
  • step S300 may include but not limited to step S310 to step S340 , and these four steps will be described in detail below in conjunction with FIG. 2 .
  • Step S310 performing sentiment classification on the target report data to obtain the report sentiment category
  • Step S320 performing text feature extraction on the target report data to obtain a first reference value for representing the value of text information
  • Step S330 evaluating the readability of the target report data to obtain a second reference value for representing the readability value of the research report
  • Step S340 obtaining public opinion factor features according to the report emotion category, the first reference value and the second reference value.
  • sentiment classification is performed on target report data by natural language processing (NLP) to obtain report sentiment categories.
  • Reported sentiment categories include, but are not limited to: Negative Sentiment, Positive Sentiment, and Neutral Sentiment.
  • step S320 may include but not limited to include steps S321 to S326:
  • Step S321 classifying the target report data to obtain the target word segmentation set, target sentence set, target paragraph set and target grammar set;
  • Step S322 performing statistical scoring processing on the target word segmentation set to obtain a word segmentation score value
  • Step S323 performing statistical scoring processing on the target sentence set to obtain a sentence score value
  • Step S324 performing statistical scoring processing on the target paragraph set to obtain a paragraph scoring value
  • Step S325 performing statistical scoring processing on the target grammar set to obtain a grammar score value
  • step S326 a first reference value is obtained according to the word segmentation score value, the sentence score value, the paragraph score value and the grammar score value.
  • the natural language processing NLP model is used to classify the target report data according to words, sentences, paragraphs, and grammars to obtain a target word segmentation set, a target sentence set, a target paragraph set, and a target grammar set.
  • word segmentation processing may be performed on each sentence in the obtained target sentence set to obtain the target word segmentation set.
  • a statistical scoring process is performed on the target word segmentation set to obtain a word segmentation score value. It mainly includes the following steps: counting the word frequency of nouns, conjunctions, function words and other types in the target word segmentation set, counting and calculating the ratio of conjunctions and function words in the target word segmentation set, and calculating the number of four-character idioms in the target word segmentation set Statistics, statistics on the relative proportion of words and phrases in the target word segmentation set, calculation of the proportion of nouns in the target word segmentation set, etc. Then score according to the ratio and quantity obtained from the statistics, and obtain the word segmentation score value.
  • step S323 of some embodiments similar to step S322, in this example, performing statistical scoring on the target sentence set to obtain a sentence score value mainly includes the following steps:
  • the average sentence length of each sentence in the target sentence set is counted, the proportion of non-text information in each sentence in the target sentence set is counted, the structural composition of each sentence in the target sentence set is statistically analyzed, and so on. Then score the average sentence length, the proportion of non-text information and the structure of the sentence to obtain the score value of the sentence.
  • step S324 of some embodiments similar to the aforementioned step S322, in this embodiment, statistical scoring processing is performed on the target paragraph set to obtain a paragraph score value, which mainly includes the following steps:
  • step S325 of some embodiments similar to the aforementioned step S322, in this embodiment, statistical scoring processing is performed on the target grammar set to obtain a grammar score value, which mainly includes the following steps:
  • step S326 of some embodiments the word segmentation score, sentence score, paragraph score and grammar score obtained in steps S322 to S325 are summed to obtain a first reference value.
  • the readability of the target report data can be evaluated through a natural language processing (NLP) model to obtain a second reference value used to characterize the readability of the research report.
  • the second reference value is a value ranging from 0 to 1. If the NLP model judges that the target report data is read smoothly and has a high value, the second reference value can be determined as a maximum value of 1. If other situations occur, the NLP model can deduct points according to the actual situation until the second reference value is 0 and terminated.
  • step S340 of some embodiments the characteristics of the public opinion factor are obtained according to the reported emotion category, the first reference value and the second reference value, specifically through the following steps:
  • the target reference value is positively or negatively processed to obtain the characteristics of public opinion factors.
  • the report emotion category if the report emotion category is neutral emotion or positive emotion, when the target reference value is positively processed, the obtained public opinion factor feature is a positive number; if the report emotion category is negative emotion, then When the target reference value is reversed, the characteristic of the public opinion factor is negative.
  • the report emotion category can also be numerically processed, and the specific operation is: if the report emotion category is neutral emotion or positive emotion, then the report emotion category is +1, and then the report emotion category and the target reference value Perform multiplication to obtain the characteristics of the public opinion factor; if the reported emotion category is negative emotion, then the reported emotion category is -1, and then multiply the reported emotion category and the target reference value to obtain the public opinion factor characteristics.
  • the second reference value is a coefficient between 0 and 1.
  • the first reference value may be a specific value, or a vector, which is used to characterize the specific situation of the target word segmentation set, target paragraph set, target sentence set and target grammar set in the target report data.
  • step S400 of some embodiments since the acquired target report data includes comment data (target comment data) of objects to be predicted (such as stocks), however, comment data usually has little impact on financial investment, so a large amount of useless Filter out the comment data to avoid too much useless data that is subsequently input into the first neural network, which will affect the accuracy of stock price data prediction. Moreover, when constructing index factors, some index factors are not effective or ineffective in specific fields. In order to improve the accuracy of stock price data forecasting, it is necessary to filter the characteristics of index factors and public opinion factors to retain effective features. , to obtain multiple quantitative trading characteristics. Specifically, the validity test may be performed on the characteristics of the index factors and the characteristics of the public opinion factors, and the characteristics of the index factors and the characteristics of the public opinion factors that have passed the test are used as the quantitative trading characteristics.
  • the validity test may be performed on the characteristics of the index factors and the characteristics of the public opinion factors, and the characteristics of the index factors and the characteristics of the public opinion factors that have passed the test are used as the quantitative trading characteristics.
  • the first neural network model includes: an input layer and a convolutional layer; step S500 may include but is not limited to include steps S510 to S530:
  • Step S510 input multiple quantitative trading features into the first neural network model
  • Step S520 preprocessing each quantitative transaction feature through the input layer to obtain corresponding standardized data
  • Step S530 performing convolution processing on the standardized data through the convolution layer to obtain multiple distributed feature vectors.
  • the first neural network model is a CNN convolutional neural network model
  • the CNN convolutional neural network model has very high efficiency in feature extraction. Extraction can improve the overall efficiency of forecasting financial investment trends.
  • step S510 of some embodiments since this application is a processing method for price data, the acquired raw data is mainly financial time series data, but there is no two-dimensional convolution neural network in the storage structure of general financial time series data.
  • Preprocessing is performed through the input layer of the CNN convolutional neural network, specifically:
  • each standardized data contains characteristic data of N trading days before the trading day to be tested. For example, in the application scenario of stock price trend prediction, use the raw data of the first 5 trading days to predict the stock price of the 6th trading day, then quantify the matrix with transaction characteristics of 49*5, and then preprocess the matrix , to get the corresponding standardized data.
  • step S530 may include but not limited to step S531 to step S532:
  • Step S531 performing format conversion on the standardized data through the convolution layer to obtain standard format data
  • step S532 the standard format data is convoluted through the convolution kernel of the convolution layer to obtain multiple distributed feature vectors.
  • convolution processing is performed through the convolutional layer of the CNN convolutional neural network model, specifically:
  • the function of the convolution layer is to extract the features of a local area, and different convolution kernels are equivalent to different feature extractors.
  • the specific operation is: take 49 features as the width, N trading days before the trading day to be tested as the length, set the step size of the filter, that is, the time interval when sliding, to 1, and set the zero padding parameter to 1.
  • the input data is convolved with a one-dimensional convolution kernel with a size of 3.
  • the number of convolution kernels is 32, that is, each convolution kernel slides horizontally to extract features according to the window size of 49*3, and trains 32 convolutions. Kernel, a total of 32 different features are extracted. After one-dimensional convolution, 32 1*N distributed feature vectors are obtained.
  • the obtained 32 1*N distributed feature vectors can be directly spliced into vectors, and then the spliced vectors can be passed through the fully connected layer of the CNN convolutional neural network model, and the target price data can also be obtained.
  • the target price data can also be obtained.
  • step S600 may include but not limited to step S610 to step S620:
  • Step S610 inputting multiple distributed feature vectors into the recurrent neural network
  • step S620 the multiple distributed feature vectors are iteratively processed through the recurrent layer in the recurrent neural network to obtain target price data.
  • the distributed eigenvectors belong to time series data, it is more appropriate to use the cyclic neural network for trend prediction, and the obtained target price data is also more accurate.
  • the second neural network model adopts an LSTM cycle neural network model.
  • the specific operation is:
  • the multiple distributed feature vectors obtained by the CNN convolutional neural network model are input into the LSTM cyclic neural network model, and the multiple distributed eigenvectors are iteratively processed through the cyclic layer in the cyclic neural network to obtain the target price data.
  • the embodiment of the present application also provides a price data processing device, which can realize the above-mentioned price data processing method, and the device includes: an acquisition module 700, a first building module 800, a second building module 900, and a screening module 1000 , feature extraction module 1100 and prediction module 1200 .
  • An acquisition module 700 configured to acquire raw data to be predicted; wherein, the raw data includes target report data and target transaction data;
  • the first construction module 800 is used to construct the index factor feature according to the target transaction data
  • the second construction module 900 is used to construct public opinion factor features according to the target report data
  • a screening module 1000 configured to perform screening processing on multiple index factor features and multiple public opinion factor features to obtain multiple quantitative trading features
  • the feature extraction module 1100 is used to perform feature extraction on multiple quantitative transaction features through the preset first neural network model to obtain multiple distributed feature vectors;
  • the prediction module 1200 is configured to input a plurality of distributed feature vectors into a preset second neural network model to perform price prediction processing to obtain target price data.
  • the price data processing device of the embodiment of the present application obtains the original data to be predicted; wherein the original data includes target report data and target transaction data, and then constructs index factor features based on target transaction data, and constructs public opinion factor features based on target report data , and then filter and process the obtained multiple index factor features and multiple public opinion factor features to obtain multiple quantitative transaction features, and then perform feature extraction on multiple quantitative transaction features through the preset first neural network model to obtain multiple Distributed eigenvectors. Finally, multiple distributed eigenvectors are input into the preset second neural network model for price prediction processing to obtain target price data.
  • the accuracy of price data prediction is improved, and, through the screening mechanism, the problem of too much irrelevant data in the feature extraction of quantitative trading features is avoided, and the accuracy of price data prediction is further improved.
  • the technical solution of the embodiment of the present application predicts the price data by combining the first neural network model and the second neural network model, which improves the speed of price data prediction and the accuracy of price data prediction.
  • the price data processing device in the embodiment of the present application corresponds to the aforementioned price data processing method, and the specific implementation is basically the same as the above-mentioned specific embodiment of the price data processing method, and will not be repeated here.
  • the embodiment of the present application also provides an electronic device, the electronic device includes: a memory, a processor, a program stored in the memory and operable on the processor, and a data bus for realizing connection and communication between the processor and the memory , when the program is executed by the processor, a method for processing price data is realized.
  • the processing method of price data includes: obtaining the original data to be predicted; wherein, the original data includes target report data and target transaction data; constructing index factor characteristics according to target transaction data; constructing public opinion factor characteristics according to target report data; The index factor features and multiple public opinion factor features are screened to obtain multiple quantitative transaction features; multiple quantitative transaction features are extracted through the preset first neural network model to obtain multiple distributed feature vectors; multiple The distributed eigenvectors are input into the preset second neural network model for price prediction processing to obtain target price data.
  • the electronic device may be any intelligent terminal including a tablet computer, a vehicle-mounted computer, and the like.
  • the electronic device of the embodiment of the present application is used to execute the aforementioned price data processing method, by obtaining the raw data to be predicted; wherein, the raw data includes target report data and target transaction data, and then constructs index factor features according to the target transaction data, Construct public opinion factor features based on the target report data, and then filter and process the obtained multiple index factor features and multiple public opinion factor features to obtain multiple quantitative transaction features, and then use the preset first neural network model to analyze multiple quantitative transactions Feature extraction is performed to obtain multiple distributed feature vectors, and finally the multiple distributed feature vectors are input into the preset second neural network model for price prediction processing to obtain target price data.
  • the accuracy of price data prediction is improved, and, through the screening mechanism, the problem of too much irrelevant data in the feature extraction of quantitative trading features is avoided, and the accuracy of price data prediction is further improved.
  • the technical solution of the embodiment of the present application predicts the price data by combining the first neural network model and the second neural network model, which improves the speed of price data prediction and the accuracy of price data prediction.
  • FIG. 8 illustrates a hardware structure of an electronic device in another embodiment.
  • the electronic device includes:
  • the processor 1300 may be implemented by a general-purpose CPU (Central Processing Unit, central processing unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute related programs to realize The technical scheme provided by the embodiment of the present application;
  • a general-purpose CPU Central Processing Unit, central processing unit
  • a microprocessor an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute related programs to realize The technical scheme provided by the embodiment of the present application;
  • ASIC Application Specific Integrated Circuit
  • the memory 1400 may be implemented in the form of a read-only memory (ReadOnlyMemory, ROM), a static storage device, a dynamic storage device, or a random access memory (RandomAccessMemory, RAM).
  • the memory 1400 can store operating systems and other application programs. When implementing the technical solutions provided by the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 1400, and are invoked by the processor 1300 to execute the implementation of this application.
  • the processing method of the price data of the example
  • Input/output interface 1500 used to realize information input and output
  • the communication interface 1600 is used to realize the communication and interaction between the device and other devices, and the communication can be realized through a wired method (such as USB, network cable, etc.), or can be realized through a wireless method (such as a mobile network, WIFI, Bluetooth, etc.);
  • bus 1700 for transferring information between various components of the device (eg, processor 1300, memory 1400, input/output interface 1500, and communication interface 1600);
  • the processor 1300 , the memory 1400 , the input/output interface 1500 and the communication interface 1600 are connected to each other within the device through the bus 1700 .
  • the embodiment of the present application also provides a storage medium, the storage medium is a computer-readable storage medium for computer-readable storage, the storage medium stores one or more programs, and one or more programs can be processed by one or more Implemented by the controller to realize a processing method of price data.
  • the processing method of price data includes: obtaining the original data to be predicted; wherein, the original data includes target report data and target transaction data; constructing index factor characteristics according to target transaction data; constructing public opinion factor characteristics according to target report data; The index factor features and multiple public opinion factor features are screened to obtain multiple quantitative transaction features; multiple quantitative transaction features are extracted through the preset first neural network model to obtain multiple distributed feature vectors; multiple The distributed eigenvectors are input into the preset second neural network model for price prediction processing to obtain target price data.
  • the computer readable storage medium can be nonvolatile or volatile.
  • the storage medium of the embodiment of the present application is used to execute the aforementioned price data processing method, by obtaining the raw data to be predicted; wherein, the raw data includes target report data and target transaction data, and then constructs index factor features according to the target transaction data, Construct public opinion factor features based on the target report data, and then filter and process the obtained multiple index factor features and multiple public opinion factor features to obtain multiple quantitative transaction features, and then use the preset first neural network model to analyze multiple quantitative transactions Feature extraction is performed to obtain multiple distributed feature vectors, and finally the multiple distributed feature vectors are input into the preset second neural network model for price prediction processing to obtain target price data.
  • the accuracy of price data prediction is improved, and, through the screening mechanism, the problem of too much irrelevant data in the feature extraction of quantitative trading features is avoided, and the accuracy of price data prediction is further improved.
  • the technical solution of the embodiment of the present application predicts the price data by combining the first neural network model and the second neural network model, which improves the speed of price data prediction and the accuracy of price data prediction.
  • memory can be used to store non-transitory software programs and non-transitory computer-executable programs.
  • the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage devices.
  • the memory optionally includes memory located remotely from the processor, and these remote memories may be connected to the processor via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
  • the device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • At least one (item) means one or more, and “multiple” means two or more.
  • “And/or” is used to describe the association relationship of associated objects, indicating that there can be three types of relationships, for example, “A and/or B” can mean: only A exists, only B exists, and A and B exist at the same time , where A and B can be singular or plural.
  • the character “/” generally indicates that the contextual objects are an “or” relationship.
  • At least one of the following” or similar expressions refer to any combination of these items, including any combination of single or plural items.
  • At least one item (piece) of a, b or c can mean: a, b, c, "a and b", “a and c", “b and c", or "a and b and c ", where a, b, c can be single or multiple.
  • the disclosed devices and methods may be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the above units is only a logical function division.
  • multiple units or components can be combined or can be Integrate into another system, or some features may be ignored, or not implemented.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.
  • the units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.
  • the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
  • the technical solution of the present application is essentially or part of the contribution to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including multiple instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method in each embodiment of the present application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), magnetic disk or optical disc, etc., which can store programs. medium.
  • ROM read-only memory
  • RAM random access memory
  • magnetic disk or optical disc etc., which can store programs. medium.

Abstract

本申请实施例提供了一种价格数据的处理方法和装置、电子设备、存储介质，属于人工智能技术领域。该方法包括：获取待预测的原始数据；其中，原始数据包括目标报告数据和目标交易数据；根据目标交易数据构建指标因子特征；根据目标报告数据构建舆论因子特征；对多个指标因子特征和多个舆论因子特征进行筛选处理，得到多个量化交易特征；通过预设的第一神经网络模型对多个量化交易特征进行特征提取，得到多个分布式特征向量；将多个分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。本申请实施例的技术方案，能够提高价格数据预测的准确性。

Description

价格数据的处理方法和装置、电子设备、存储介质
本申请要求于2022年2月22日提交中国专利局、申请号为202210160991.4，发明名称为“价格数据的处理方法和装置、电子设备、存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能及宏观经济技术领域，尤其涉及一种价格数据的处理方法和装置、电子设备、存储介质。
背景技术
通常，商家或者价格评估机构会对产品的未来价格进行预测。相关技术中，采取线性回归为代表的机器学习模型对价格数据进行预测。
技术问题
以下是发明人意识到的现有技术的技术问题：采取线性回归为代表的机器学习模型对价格数据进行预测，输入到机器学习模型中进行预测的原始数据是一种非线性数据，通过该机器学习模型对价格数据进行预测容易导致估计预测结果不准确，因此，如何提高价格数据预测的准确性，成为了亟待解决的技术问题。
技术解决方案
第一方面，本申请实施例提出了一种价格数据的处理方法，所述方法包括：
获取待预测的原始数据；其中，所述原始数据包括目标报告数据和目标交易数据；
根据所述目标交易数据构建指标因子特征；
根据所述目标报告数据构建舆论因子特征；
对多个所述指标因子特征和多个所述舆论因子特征进行筛选处理，得到多个量化交易特征；
通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个分布式特征向量；
将多个所述分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。
第二方面，本申请实施例提出了一种价格数据的处理装置，所述装置包括：
获取模块，用于获取待预测的原始数据；其中，所述原始数据包括目标报告数据和目标交易数据；
第一构建模块，用于根据所述目标交易数据构建指标因子特征；
第二构建模块，用于根据所述目标报告数据构建舆论因子特征；
筛选模块，用于对多个所述指标因子特征和多个所述舆论因子特征进行筛选处理，得到多个量化交易特征；
特征提取模块，用于通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个分布式特征向量；
预测模块，用于将多个所述分布式特征向量输入至预设的第二神经网络模型中进行价格 预测处理，得到目标价格数据。
第三方面，本申请实施例提出了一种电子设备，包括：
至少一个存储器；
至少一个处理器；
至少一个程序；
所述程序被存储在所述存储器中，处理器执行所述至少一个程序以实现价格数据的处理方法；其中，所述价格数据的处理方法包括：获取待预测的原始数据；其中，所述原始数据包括目标报告数据和目标交易数据；根据所述目标交易数据构建指标因子特征；根据所述目标报告数据构建舆论因子特征；对多个所述指标因子特征和多个所述舆论因子特征进行筛选处理，得到多个量化交易特征；通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个分布式特征向量；将多个所述分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。
第四方面，本申请实施例提出了一种存储介质，所述存储介质为计算机可读存储介质，所述计算机可读存储介质存储有计算机可执行指令，所述计算机可执行指令用于使计算机执行价格数据的处理方法；其中，所述价格数据的处理方法包括：获取待预测的原始数据；其中，所述原始数据包括目标报告数据和目标交易数据；根据所述目标交易数据构建指标因子特征；根据所述目标报告数据构建舆论因子特征；对多个所述指标因子特征和多个所述舆论因子特征进行筛选处理，得到多个量化交易特征；通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个分布式特征向量；将多个所述分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。
有益效果
本申请实施例提出一种价格数据的处理方法和装置、电子设备、存储介质，通过获取多种原始数据，提高了价格数据预测的准确性，并通过筛选机制，避免出现在对量化交易特征进行特征提取时不相关数据过多的问题，并通过第一神经网络模型和第二神经网络模型对价格数据进行预测，提高了对价格数据预测的速度以及提高了价格数据预测的准确性。
附图说明
图1是本申请实施例提供的价格数据的处理方法的流程图；
图2是图1中的步骤S300的具体方法流程图；
图3是图2中的步骤S320的具体方法流程图；
图4是图1中的步骤S500的具体方法的流程图；
图5是图4中的步骤S530的具体方法的流程图；
图6是图1中的步骤S600的具体方法的流程图；
图7是本申请实施例提供的价格数据的处理装置的模块框图；
图8是本申请实施例提供的电子设备的硬件结构示意图。
本发明的实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本申请，并不用于限定本申请。
需要说明的是，虽然在装置示意图中进行了功能模块划分，在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于装置中的模块划分，或流程图中的顺序执行所示出或描述的步骤。说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的 对象，而不必用于描述特定的顺序或先后次序。
除非另有定义，本文所使用的所有的技术和科学术语与属于本申请的技术领域的技术人员通常理解的含义相同。本文中所使用的术语只是为了描述本申请实施例的目的，不是旨在限制本申请。
首先，对本申请中涉及的若干名词进行解析：
人工智能(artificial intelligence，AI)：是研究、开发用于模拟、延伸和扩展人的智能的理论、方法、技术及应用系统的一门新的技术科学；人工智能是计算机科学的一个分支，人工智能企图了解智能的实质，并生产出一种新的能以人类智能相似的方式做出反应的智能机器，该领域的研究包括机器人、语言识别、图像识别、自然语言处理和专家系统等。人工智能可以对人的意识、思维的信息过程的模拟。人工智能还是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能，感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用系统。
自然语言处理(natural language processing，NLP)：NLP用计算机来处理、理解以及运用人类语言(如中文、英文等)，NLP属于人工智能的一个分支，是计算机科学与语言学的交叉学科，又常被称为计算语言学。自然语言处理包括语法分析、语义分析、篇章理解等。自然语言处理常用于机器翻译、手写体和印刷体字符识别、语音识别及文语转换、信息意图识别、信息抽取与过滤、文本分类与聚类、舆情分析和观点挖掘等技术领域，它涉及与语言处理相关的数据挖掘、机器学习、知识获取、知识工程、人工智能研究和与语言计算相关的语言学研究等。
卷积神经网络模型(Convolutional Neural Networks，CNN)：卷积神经网络是一种前馈神经网络，它由若干卷积层和池化层组成。CNN的基本结构由输入层、卷积层(convolutional layer)、池化层(pooling layer，也称为取样层)、全连接层及输出层构成。卷积层和池化层一般会取若干个，采用卷积层和池化层交替设置，即一个卷积层连接一个池化层，池化层后再连接一个卷积层，依此类推。由于卷积层中输出特征图的每个神经元与其输入进行局部连接，并通过对应的连接权值与局部输入进行加权求和再加上偏置值，得到该神经元输入值，该过程等同于卷积过程，CNN也由此而得名。卷积神经网络由多层感知机(MLP)演变而来，由于其具有局部区域连接、权值共享、降采样的结构特点，使得卷积神经网络在图像处理领域表现出色。卷积神经网络相比于其他神经网络的特殊性主要在于权值共享与局部连接两个方面。权值共享使得卷积神经网络的网络结构更加类似于生物神经网络。局部连接不像传统神经网络那样，第n-1层的每一神经元都与第n层的所有神经元连接，而是第n-1层的神经元与第n层的部分神经元之间连接。这两个特点的作用在于降低了网络模型的复杂度，减少了权值的数目。
零填充(zero padding)：在输入矩阵的边缘使用零值进行填充，这样我们就可以对输入图像矩阵的边缘进行滤波。零填充的一大好处是可以让我们控制特征图的大小。使用零填充的也叫做泛卷积，不适用零填充的叫做严格卷积。
循环神经网络(Recurrent Neural Network，RNN)是一类以序列(sequence)数据为输入，在序列的演进方向进行递归(recursion)且所有节点(循环单元)按链式连接的递归神经网络(recursive neural network)。
长短期记忆神经网络(Long-Short Term Memory，LSTM)是一种特殊的循环神经网络(RNN)，LSTM是为了解决一般的循环神经网络存在的长期依赖问题而专门设计出来的，所有的RNN都具有一种重复神经网络模块的链式形式。原始的RNN在训练中，随着训练时间的加长以及网络层数的增多，很容易出现梯度爆炸或者梯度消失的问题，导致无法处理较长序列数据，从而无法获取长距离数据的信息。LSTM应用的领域包括：文本生成、机器翻译、语音识别、生成图像描述和视频标记等。
分词处理：分词处理就是将文本中的词与词之间自动加上空格或者其他边界标记。英文单词天然有空格隔开容易按照空格分词，但是也有时候需要把多个单词做为一个分词，比如 一些名词如“New York”，需要做为一个词看待。而中文由于没有空格，分词就是一个需要专门去解决的问题了。无论是英文还是中文，分词的原理都是类似的。汉语自动分词就是让计算机系统在汉语文本中的词与词之间自动加上空格或其他边界标记。常用的中文分词工具有Jieba。
随着经济与科技的不断发展与人工智能技术的不断进步，智能化的价格数据的处理方法得到了广泛的应用。
通常，商家或者价格评估机构会对产品的未来价格进行预测。相关技术中，采取线性回归为代表的机器学习模型对价格数据进行预测，然而，输入到机器学习模型中进行预测的原始数据是一种非线性数据，通过该机器学习模型对价格数据进行预测容易导致估计预测结果不准确，因此，如何提高价格数据预测的准确性，成为了亟待解决的技术问题。
基于此，本申请实施例提供了一种价格数据的处理方法和装置、电子设备、存储介质，通过获取多种原始数据对价格走势进行预测，能够提高价格数据预测的准确性。
本申请实施例提供的价格数据的处理方法和装置、电子设备、存储介质，具体通过如下实施例进行说明，首先描述本申请实施例中的价格数据的处理方法。
本申请实施例可以基于人工智能技术对相关的数据进行获取和处理。其中，人工智能(Artificial Intelligence，AI)是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能，感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用系统。
人工智能基础技术一般包括如传感器、专用人工智能芯片、云计算、分布式存储、大数据处理技术、操作/交互系统、机电一体化等技术。人工智能软件技术主要包括计算机视觉技术、机器人技术、生物识别技术、语音处理技术、自然语言处理技术以及机器学习/深度学习等几大方向。
本申请实施例提供的价格数据的处理方法，涉及人工智能技术领域。本申请实施例提供的价格数据的处理方法可应用于终端中，也可应用于服务器端中，还可以是运行于终端或服务器端中的软件。在一些实施例中，终端可以是智能手机、平板电脑、笔记本电脑、台式计算机等；服务器端可以配置成独立的物理服务器，也可以配置成多个物理服务器构成的服务器集群或者分布式系统，还可以配置成提供云服务、云数据库、云计算、云函数、云存储、网络服务、云通信、中间件服务、域名服务、安全服务、CDN以及大数据和人工智能平台等基础云计算服务的云服务器；软件可以是实现价格数据的处理方法的应用等，但并不局限于以上形式。
本申请可用于众多通用或专用的计算机系统环境或配置中。例如：个人计算机、服务器计算机、手持设备或便携式设备、平板型设备、多处理器系统、基于微处理器的系统、置顶盒、可编程的消费电子设备、网络PC、小型计算机、大型计算机、包括以上任何系统或设备的分布式计算环境等等。本申请可以在由计算机执行的计算机可执行指令的一般上下文中描述，例如程序模块。一般地，程序模块包括执行特定任务或实现特定抽象数据类型的例程、程序、对象、组件、数据结构等等。也可以在分布式计算环境中实践本申请，在这些分布式计算环境中，由通过通信网络而被连接的远程处理设备来执行任务。在分布式计算环境中，程序模块可以位于包括存储设备在内的本地和远程计算机存储介质中。
下面结合附图对本申请实施例的价格数据的处理方法进行详细描述。
图1是本申请实施例提供的价格数据的处理方法的一个可选的流程图，图1中的方法可以包括但不限于包括步骤S100至步骤S600，下面结合图1对这六个步骤进行详细介绍。
步骤S100，获取待预测的原始数据；其中，原始数据包括目标报告数据和目标交易数据；
步骤S200，根据目标交易数据构建指标因子特征；
步骤S300，根据目标报告数据构建舆论因子特征；
步骤S400，对多个指标因子特征和多个舆论因子特征进行筛选处理，得到多个量化交易特征；
步骤S500，通过预设的第一神经网络模型对多个量化交易特征进行特征提取，得到多个 分布式特征向量；
步骤S600，将多个分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。
本申请实施例的价格数据的处理方法，通过获取待预测的原始数据；其中，原始数据包括目标报告数据和目标交易数据，然后根据目标交易数据构建指标因子特征，根据目标报告数据构建舆论因子特征，然后对得到的多个指标因子特征和多个舆论因子特征进行筛选处理，得到多个量化交易特征，再通过预设的第一神经网络模型对多个量化交易特征进行特征提取，得到多个分布式特征向量，最后将多个分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。通过获取多种原始数据，提高了价格数据预测的准确性，并且，通过筛选机制，避免出现在对量化交易特征进行特征提取时不相关数据过多的问题，进一步提高了价格数据预测的准确性，同时，本申请实施例的技术方案，通过结合第一神经网络模型和第二神经网络模型对价格数据进行预测，提高了对价格数据预测的速度以及提高了价格数据预测的准确性。
在一些实施例的步骤S100中，可以通过编写网络爬虫，设置好数据源之后进行有目标性的爬取数据，得到原始数据。也可以通过一些公开的网站进行数据查询，得到原始数据。本申请实施例的技术方案可以实现对股票市场的一只股票、基金市场的一只基金或者期货市场的一只期货等等的价格数据预测，并且，相比较于现有技术中采取机器学习模型的预测方法，本申请实施例的价格数据的处理方法得到的目标价格数据更加准确。并且，通过多次对某一个股票、基金市场的一只基金或者期货市场的一只期货进行价格预测，得到多个对应的目标价格数据，从而得到该股票、基金或者期货的价格走势。
如，待预测的对象是股票市场的一只股票，即本申请实施例的技术方案是对股票市场的一只股票的价格数据进行预测，则在股票的价格数据预测的应用场景中，获取原始数据的方式可以是前往待测股票对应的公司网站查看一些目标报告数据，前往交易平台查看目标报告数据和目标交易数据。
在一些实施例中，目标报告数据包括如下至少之一：目标行业报告数据、目标公司报告数据、目标新闻数据和目标评论数据。
目标交易数据包括以下至少之一：开盘价、收盘价、最高价、最低价、成交量。
本申请实施例通过获取目标报告数据和目标交易数据得到原始数据，然后综合考虑目标报告数据和目标交易数据对价格未来走势的影响，从而提高了对价格数据预测的准确性。
例如，在股票的价格数据预测的应用场景中，需要对某一个股票A的未来走势进行预测，则需要得到股票A一段时间的价格数据，需要获取股票A的原始数据，包括目标报告数据和目标交易数据，其中目标报告数据包括但不限于以下至少一种：股票A对应公司所在行业的研报数据(即目标行业报告数据)、股票A对应公司的研报数据(即目标公司报告数据)、和股票A公司相关的新闻数据(即目标新闻数据)以及股吧中对股票A的评论数据(即目标评论数据)。目标交易数据包括但不限于以下至少之一：股票A的历史开盘价、股票A的历史收盘价、股票A的历史最高价、股票A的历史最低价、股票A的历史成交量以及股票A的当前价格。
在一些实施例的步骤S200中，构建指标因子特征包括指标因子构造，指标因子包括但不限于如下至少之一：OBV(On Balance Volume，能量潮)因子、CCI(commoditychannelindex，顺势指标)因子、KDJ(随机指标)因子等等一系列的技术指标。
请参阅图2，在一些实施例中，步骤S300可以包括但不限于包括步骤S310至步骤S340，下面结合图2对这四个步骤进行详细介绍。
步骤S310，对目标报告数据进行情感分类，得到报告情感类别；
步骤S320，对目标报告数据进行文本特征提取，得到用于表征文本信息价值的第一参考值；
步骤S330，对目标报告数据进行可读性评估，得到用于表征研报可读价值的第二参考值；
步骤S340，根据报告情感类别、第一参考值和第二参考值得到舆论因子特征。
在一些实施例的步骤S310中，通过自然语言处理NLP对目标报告数据进行情感分类，得到报告情感类别。报告情感类别包括但不限于：消极情感、积极情感和中性情感。
请参阅图3，在一些实施例中，步骤S320可以包括但不限于包括步骤S321至步骤S326：
步骤S321，对目标报告数据进行分类处理，得到目标分词集、目标语句集、目标段落集和目标语法集；
步骤S322，对目标分词集进行统计评分处理，得到分词评分值；
步骤S323，对目标语句集进行统计评分处理，得到语句评分值；
步骤S324，对目标段落集进行统计评分处理，得到段落评分值；
步骤S325，对目标语法集进行统计评分处理，得到语法评分值；
步骤S326，根据分词评分值、语句评分值、段落评分值和语法评分值得到第一参考值。
具体地，在一些实施例的步骤S321中，通过自然语言处理NLP模型对目标报告数据按照词语、语句、段落、语法进行分类，得到目标分词集、目标语句集、目标段落集和目标语法集。其中，得到目标分词集时，可以对得到的目标语句集中的每一条语句进行分词处理，得到该目标分词集。
在一些实施例的步骤S322中，对目标分词集进行统计评分处理，得到分词评分值。主要包括以下步骤：对目标分词集中的名词、连接词、虚词等种类的词频进行统计、对目标分词集中的连接词和虚词的比例进行统计并计算比例，对目标分词集中的四字成语数量进行统计、对目标分词集中单字与词组的相对比例进行统计、对目标分词集中的名词所占比例进行计算等等。然后根据统计得到的比例和数量进行评分，得到分词评分值。
在一些实施例的步骤S323中，与步骤S322类似，在本实例中，对目标语句集进行统计评分，得到语句评分值，主要包括以下步骤：
对目标语句集中的每一条语句的平均句长进行统计、对目标语句集中的每一条语句中非文本类信息比例进行统计、对目标语句集中的每一条语句的结构组成进行统计分析等等。然后对统计得到的平均句长、非文本类信息比例和语句结构组成进行评分，得到语句评分值。
在一些实施例的步骤S324中，与前述步骤S322类似，在本实施例中，对目标段落集进行统计评分处理，得到段落评分值，主要包括以下步骤：
对目标段落集的段落数进行统计、对目标段落集中的每一个段落所包含的语句数进行统计、计算段落的平均句子数、计算段落的SMOG指数等等。然后对前述统计的段落数、平均句子数和SMOG指数进行评分，得到段落评分值。
在一些实施例的步骤S325中，与前述步骤S322类似，在本实施例中，对目标语法集进行统计评分处理，得到语法评分值，主要包括以下步骤：
对目标语法集中的语法树高度进行统计、对目标语法集中的语法树的节点数进行统计、对语法树的名词短语比例进行统计、对语法树的动词短语比例进行统计、对语法树的形容词短语比例进行统计等等。然后，将前述统计得到的所有比例和数量都进行评分，得到语法评分值。
在一些实施例的步骤S326中，将步骤S322至步骤S325得到的分词评分值、语句评分值、段落评分值和语法评分值进行求和处理，得到第一参考值。
在一些实施例的步骤S330中，可以通过自然语言处理NLP模型对目标报告数据进行可读性评估，得到用于表征研报可读价值的第二参考值。第二参考值是一个范围0至1之间的值，如果NLP模型判断目标报告数据阅读流畅且具备的价值较高，可以将第二参考值确定为最大值1。如果出现其他情况，NLP模型可以根据实际情况进行扣分，直到第二参考值为0时终止。
在一些实施例的步骤S340中，根据报告情感类别、第一参考值和第二参考值得到舆论因子特征，具体通过以下步骤实现：
将第一参考值和第二参考值进行相乘处理，得到目标参考值；
根据报告情感类别对目标参考值进行取正或者取反处理，得到舆论因子特征。
具体地，对于报告情感类别，如果报告情感类别是中性情感或者积极情感，则对目标参考值进行取正处理时，得到的舆论因子特征为正数；如果报告情感类别为消极的情感，则对目标参考值进行取反处理时，得到舆论因子特征为负数。
在一些实施例中，还可以对报告情感类别进行数值化处理，具体操作为：如果报告情感类别是中性情感或者积极情感，则报告情感类别为+1，然后将报告情感类别和目标参考值进行相乘处理，得到舆论因子特征；如果报告情感类别是消极情感，则报告情感类别是-1，然后将报告情感类别和目标参考值进行相乘处理，得到舆论因子特征。
需要说明的是，第二参考值是一个0至1之间的系数。第一参考值可以是一个具体的值，也可以是一个向量，该向量用于表征目标分词集、目标段落集、目标语句集和目标语法集在目标报告数据中的具体情况。
在一些实施例的步骤S400中，由于获取的目标报告数据包括待预测对象(如股票)的评论数据(目标评论数据)，然而，评论数据通常对于金融投资的影响较小，因此需要将大量无用的评论数据筛选掉，避免后续输入第一神经网络中的无用数据过多，影响股票的价格数据预测的准确性。并且，在指标因子构造时，有些指标因子在特定领域的效力不大或者是无效的，为了提高股票价格数据预测的准确性，需要将指标因子特征和舆论因子特征进行筛选处理，保留有效的特征，得到多个量化交易特征。具体可以是对指标因子特征和舆论因子特征进行有效性检验，将通过检验测试的指标因子特征和舆论因子特征作为量化交易特征。
请参阅图4，在一些实施例中，第一神经网络模型包括：输入层和卷积层；步骤S500可以包括但不限于包括步骤S510至步骤S530：
步骤S510，将多个量化交易特征输入到第一神经网络模型中；
步骤S520，通过输入层对每一量化交易特征进行预处理，得到对应的标准化数据；
步骤S530，通过卷积层对标准化数据进行卷积处理，得到多个分布式特征向量。
具体地，在本实施例中，第一神经网络模型为CNN卷积神经网络模型，CNN卷积神经网络模型在特征提取上具有非常高的效率，因此，首先通过CNN卷积神经网络模型进行特征提取，能够提高整体对金融投资走势预测的效率。
在一些实施例的步骤S510中，由于本申请为价格数据的处理方法，获取的原始数据主要是金融时间序列数据，但是，一般的金融时间序列数据的存储结构中并不存在二维卷积神经网络所需的有意义的二维空间关系，且使用二维卷积网络处理金融时间序列数据时，存在纵向移动会导致时间上信息的损失的问题。因此本申请实施例选择使用一维卷积神经网络模型进行特征提取与预测。首先需要对每一个量化交易特征进行预处理，得到标准化数据。
通过CNN卷积神经网络的输入层进行预处理，具体为：
原始数据经过前叙步骤的指标因子构造和舆论评估后，对得到的指标因子特征和舆论因子特征进行筛选处理，最终保留49个量化交易特征。然后将量化交易特征经过输入层的(0，1)标准化的与处理后，得到标准化数据。每一个标准化数据包含待测交易日之前N个交易日的特征数据。例如，在股票股价的走势预测的应用场景中，使用前5个交易日的原始数据预测第6个交易日的股价，那么，量化交易特征为49*5的矩阵，然后将该矩阵进行预处理，得到对应的标准化数据。
请参阅图5，在一些实施例中，步骤S530可以包括但不限于步骤S531至步骤S532：
步骤S531，通过卷积层对标准化数据进行格式转换，得到标准格式数据；
步骤S532，通过卷积层的卷积核对标准格式数据进行卷积，得到多个分布式特征向量。
具体地，在本实施例中，通过CNN卷积神经网络模型的卷积层进行卷积处理，具体为：
卷积层的作用是提取一个局部区域的特征，不同的卷积核相当于不同的特征提取器。在本实施例中，由于使用的是一维卷积，为了使用一维卷积，需要对标准化数据进行格式转换，得到标准格式数据。具体操作为：将49个特征作为宽，待测交易日之前N个交易日为长，滤波器的步长即滑动时的时间间隔设置为1，使用零填充参数设置为1。以尺寸为3的一维卷 积核对输入数据进行卷积处理，卷积核个数为32，即每一个卷积核对输入空间按照49*3的窗口大小横向滑动提取特征，训练32个卷积核，共提取出32种不同的特征。其中经过一维卷积过后，得到32个1*N的分布式特征向量。
如在上述在股票的价格数据预测的应用场景中，使用前5个交易日的原始数据预测第6个交易日的股价，经过格式转换和卷积后得到32个1*5的分布式特征向量。
在一些实施例中，可以将得到的32个1*N个分布式特征向量直接进行向量拼接，然后将拼接得到的向量通过CNN卷积神经网络模型的全连接层，也可以得到目标价格数据，但是，由于在金融领域，数据多为金融时间序列数据，单独使用CNN卷积神经网络模型很难捕捉到时序信息，导致得到的目标价格数据不太准确。
请参阅图6，在一些实施例中，步骤S600可以包括但不限于步骤S610至步骤S620：
步骤S610，将多个分布式特征向量输入至循环神经网络中；
步骤S620，通过循环神经网络中的循环层对多个分布式特征向量进行循环迭代处理，得到目标价格数据。
具体地，在本实施例中，由于分布式特征向量是属于时间序列数据，因此，使用循环神经网络进行走势预测更加合适，得到的目标价格数据也更加准确。
在本实施例中，第二神经网络模型采取LSTM循环神经网络模型。具体操作为：
将CNN卷积神经网络模型得到的多个分布式特征向量输入至LSTM循环神经网络模型中，通过循环神经网络中的循环层对多个分布式特征向量进行循环迭代处理，得到目标价格数据。
本申请实施例的技术方案，通过结合CNN卷积神经网络模型和LSTM循环神经网络模型对金融投资的走势进行预测，不仅能够快速捕捉到原始数据之间的交互特征，还能捕捉到原始数据的时序信息，从而提高了金融投资走势预测的准确性，提高了预测效率。
请参阅图7，本申请实施例还提供一种价格数据的处理装置，可以实现上述价格数据的处理方法，该装置包括：获取模块700、第一构建模块800、第二构建模块900、筛选模块1000、特征提取模块1100和预测模块1200。
获取模块700，用于获取待预测的原始数据；其中，原始数据包括目标报告数据和目标交易数据；
第一构建模块800，用于根据目标交易数据构建指标因子特征；
第二构建模块900，用于根据目标报告数据构建舆论因子特征；
筛选模块1000，用于对多个指标因子特征和多个舆论因子特征进行筛选处理，得到多个量化交易特征；
特征提取模块1100，用于通过预设的第一神经网络模型对多个量化交易特征进行特征提取，得到多个分布式特征向量；
预测模块1200，用于将多个分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。
本申请实施例的价格数据的处理装置，通过获取待预测的原始数据；其中，原始数据包括目标报告数据和目标交易数据，然后根据目标交易数据构建指标因子特征，根据目标报告数据构建舆论因子特征，然后对得到的多个指标因子特征和多个舆论因子特征进行筛选处理，得到多个量化交易特征，再通过预设的第一神经网络模型对多个量化交易特征进行特征提取，得到多个分布式特征向量，最后将多个分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。通过获取多种原始数据，提高了价格数据预测的准确性，并且，通过筛选机制，避免出现在对量化交易特征进行特征提取时不相关数据过多的问题，进一步提高了价格数据预测的准确性，同时，本申请实施例的技术方案，通过结合第一神经网络模型和第二神经网络模型对价格数据进行预测，提高了对价格数据预测的速度以及提高了价格数据预测的准确性。
需要说明的是，本申请实施例的价格数据的处理装置与前述的价格数据的处理方法相对 应，具体的实施方式与上述价格数据的处理方法的具体实施例基本相同，在此不再赘述。
本申请实施例还提供了一种电子设备，电子设备包括：存储器、处理器、存储在存储器上并可在处理器上运行的程序以及用于实现处理器和存储器之间的连接通信的数据总线，程序被处理器执行时实现一种价格数据的处理方法。其中，价格数据的处理方法包括：获取待预测的原始数据；其中，原始数据包括目标报告数据和目标交易数据；根据目标交易数据构建指标因子特征；根据目标报告数据构建舆论因子特征；对多个指标因子特征和多个舆论因子特征进行筛选处理，得到多个量化交易特征；通过预设的第一神经网络模型对多个量化交易特征进行特征提取，得到多个分布式特征向量；将多个分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。该电子设备可以为包括平板电脑、车载电脑等任意智能终端。
本申请实施例的电子设备，用于执行前述的价格数据的处理方法，通过获取待预测的原始数据；其中，原始数据包括目标报告数据和目标交易数据，然后根据目标交易数据构建指标因子特征，根据目标报告数据构建舆论因子特征，然后对得到的多个指标因子特征和多个舆论因子特征进行筛选处理，得到多个量化交易特征，再通过预设的第一神经网络模型对多个量化交易特征进行特征提取，得到多个分布式特征向量，最后将多个分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。通过获取多种原始数据，提高了价格数据预测的准确性，并且，通过筛选机制，避免出现在对量化交易特征进行特征提取时不相关数据过多的问题，进一步提高了价格数据预测的准确性，同时，本申请实施例的技术方案，通过结合第一神经网络模型和第二神经网络模型对价格数据进行预测，提高了对价格数据预测的速度以及提高了价格数据预测的准确性。
请参阅图8，图8示意了另一实施例的电子设备的硬件结构，电子设备包括：
处理器1300，可以采用通用的CPU(CentralProcessingUnit，中央处理器)、微处理器、应用专用集成电路(ApplicationSpecificIntegratedCircuit，ASIC)、或者一个或多个集成电路等方式实现，用于执行相关程序，以实现本申请实施例所提供的技术方案；
存储器1400，可以采用只读存储器(ReadOnlyMemory，ROM)、静态存储设备、动态存储设备或者随机存取存储器(RandomAccessMemory，RAM)等形式实现。存储器1400可以存储操作系统和其他应用程序，在通过软件或者固件来实现本说明书实施例所提供的技术方案时，相关的程序代码保存在存储器1400中，并由处理器1300来调用执行本申请实施例的价格数据的处理方法；
输入/输出接口1500，用于实现信息输入及输出；
通信接口1600，用于实现本设备与其他设备的通信交互，可以通过有线方式(例如USB、网线等)实现通信，也可以通过无线方式(例如移动网络、WIFI、蓝牙等)实现通信；
总线1700，在设备的各个组件(例如处理器1300、存储器1400、输入/输出接口1500和通信接口1600)之间传输信息；
其中处理器1300、存储器1400、输入/输出接口1500和通信接口1600通过总线1700实现彼此之间在设备内部的通信连接。
本申请实施例还提供了一种存储介质，存储介质为计算机可读存储介质，用于计算机可读存储，存储介质存储有一个或者多个程序，一个或者多个程序可被一个或者多个处理器执行，以实现一种价格数据的处理方法。其中，价格数据的处理方法包括：获取待预测的原始数据；其中，原始数据包括目标报告数据和目标交易数据；根据目标交易数据构建指标因子特征；根据目标报告数据构建舆论因子特征；对多个指标因子特征和多个舆论因子特征进行筛选处理，得到多个量化交易特征；通过预设的第一神经网络模型对多个量化交易特征进行特征提取，得到多个分布式特征向量；将多个分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。
该计算机可读存储介质可以是非易失性，也可以是易失性。本申请实施例的存储介质，用于执行前述的价格数据的处理方法，通过获取待预测的原始数据；其中，原始数据包括目 标报告数据和目标交易数据，然后根据目标交易数据构建指标因子特征，根据目标报告数据构建舆论因子特征，然后对得到的多个指标因子特征和多个舆论因子特征进行筛选处理，得到多个量化交易特征，再通过预设的第一神经网络模型对多个量化交易特征进行特征提取，得到多个分布式特征向量，最后将多个分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。通过获取多种原始数据，提高了价格数据预测的准确性，并且，通过筛选机制，避免出现在对量化交易特征进行特征提取时不相关数据过多的问题，进一步提高了价格数据预测的准确性，同时，本申请实施例的技术方案，通过结合第一神经网络模型和第二神经网络模型对价格数据进行预测，提高了对价格数据预测的速度以及提高了价格数据预测的准确性。
存储器作为一种非暂态计算机可读存储介质，可用于存储非暂态软件程序以及非暂态性计算机可执行程序。此外，存储器可以包括高速随机存取存储器，还可以包括非暂态存储器，例如至少一个磁盘存储器件、闪存器件、或其他非暂态固态存储器件。在一些实施方式中，存储器可选包括相对于处理器远程设置的存储器，这些远程存储器可以通过网络连接至该处理器。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
本申请实施例描述的实施例是为了更加清楚的说明本申请实施例的技术方案，并不构成对于本申请实施例提供的技术方案的限定，本领域技术人员可知，随着技术的演变和新应用场景的出现，本申请实施例提供的技术方案对于类似的技术问题，同样适用。
本领域技术人员可以理解的是，图1至图8中示出的技术方案并不构成对本申请实施例的限定，可以包括比图示更多或更少的步骤，或者组合某些步骤，或者不同的步骤。
以上所描述的装置实施例仅仅是示意性的，其中作为分离部件说明的单元可以是或者也可以不是物理上分开的，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
本领域普通技术人员可以理解，上文中所公开方法中的全部或某些步骤、系统、设备中的功能模块/单元可以被实施为软件、固件、硬件及其适当的组合。
本申请的说明书及上述附图中的术语“第一”、“第二”、“第三”、“第四”等(如果存在)是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
应当理解，在本申请中，“至少一个(项)”是指一个或者多个，“多个”是指两个或两个以上。“和/或”，用于描述关联对象的关联关系，表示可以存在三种关系，例如，“A和/或B”可以表示：只存在A，只存在B以及同时存在A和B三种情况，其中A，B可以是单数或者复数。字符“/”一般表示前后关联对象是一种“或”的关系。“以下至少一项(个)”或其类似表达，是指这些项中的任意组合，包括单项(个)或复数项(个)的任意组合。例如，a，b或c中的至少一项(个)，可以表示：a，b，c，“a和b”，“a和c”，“b和c”，或“a和b和c”，其中a，b，c可以是单个，也可以是多个。
在本申请所提供的几个实施例中，应该理解到，所揭露的装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，上述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
上述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元 上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。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或者该技术方案的全部或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一个存储介质中，包括多指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例的方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(Read-Only Memory，简称ROM)、随机存取存储器(Random Access Memory，简称RAM)、磁碟或者光盘等各种可以存储程序的介质。
以上参照附图说明了本申请实施例的优选实施例，并非因此局限本申请实施例的权利范围。本领域技术人员不脱离本申请实施例的范围和实质内所作的任何修改、等同替换和改进，均应在本申请实施例的权利范围之内。

Claims (20)

  1. 一种价格数据的处理方法，其中，所述方法包括：
    获取待预测的原始数据；其中，所述原始数据包括目标报告数据和目标交易数据；
    根据所述目标交易数据构建指标因子特征；
    根据所述目标报告数据构建舆论因子特征；
    对多个所述指标因子特征和多个所述舆论因子特征进行筛选处理，得到多个量化交易特征；
    通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个分布式特征向量；
    将多个所述分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。
  2. 根据权利要求1所述的方法，其中，所述根据所述目标报告数据构建舆论因子特征，包括：
    对所述目标报告数据进行情感分类，得到报告情感类别；
    对所述目标报告数据进行文本特征提取，得到用于表征文本信息价值的第一参考值；
    对所述目标报告数据进行可读性评估，得到用于表征研报可读价值的第二参考值；
    根据所述报告情感类别、所述第一参考值和所述第二参考值得到所述舆论因子特征。
  3. 根据权利要求2所述的方法，其中，所述对所述目标报告数据进行文本特征提取，得到用于表征文本信息价值的第一参考值，包括：
    对所述目标报告数据进行分类处理，得到目标分词集、目标语句集、目标段落集和目标语法集；
    对所述目标分词集进行统计评分处理，得到分词评分值；
    对所述目标语句集进行统计评分处理，得到语句评分值；
    对所述目标段落集进行统计评分处理，得到段落评分值；
    对所述目标语法集进行统计评分处理，得到语法评分值；
    根据所述分词评分值、所述语句评分值、所述段落评分值和所述语法评分值得到所述第一参考值。
  4. 根据权利要求2至3任意一项所述的方法，其中，所述根据所述报告情感类别、所述第一参考值和所述第二参考值得到所述舆论因子特征，包括：
    将所述第一参考值和所述第二参考值进行相乘处理，得到目标参考值；
    根据所述报告情感类别对所述目标参考值进行取正处理或者取反处理，得到所述舆论因子特征。
  5. 根据权利要求1至3任意一项所述的方法，其中，所述第一神经网络模型包括：输入层和卷积层；所述通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个所述分布式特征向量，包括：
    将多个所述量化交易特征输入到所述第一神经网络模型中；
    通过所述输入层对每一所述量化交易特征进行预处理，得到对应的标准化数据；
    通过所述卷积层对所述标准化数据进行卷积处理，得到多个所述分布式特征向量。
  6. 根据权利要求5所述的方法，其中，所述通过所述卷积层对所述标准化数据进行卷积处理，得到多个所述分布式特征向量，包括：
    通过所述卷积层对所述标准化数据进行格式转换，得到标准格式数据；
    通过所述卷积层的卷积核对所述标准格式数据进行卷积，得到多个所述分布式特征向量。
  7. 根据权利要求1至3任意一项所述的方法，其中，所述第二神经网络模型包括循环神经网络；所述将多个所述分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据，包括：
    将多个所述分布式特征向量输入至所述循环神经网络中；
    通过所述循环神经网络中的循环层对多个所述分布式特征向量进行循环迭代处理，得到所述目标价格数据。
  8. 一种价格数据的处理装置，其中，所述装置包括：
    获取模块，用于获取待预测的原始数据；其中，所述原始数据包括目标报告数据和目标交易数据；
    第一构建模块，用于根据所述目标交易数据构建指标因子特征；
    第二构建模块，用于根据所述目标报告数据构建舆论因子特征；
    筛选模块，用于对多个所述指标因子特征和多个所述舆论因子特征进行筛选处理，得到多个量化交易特征；
    特征提取模块，用于通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个分布式特征向量；
    预测模块，用于将多个所述分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。
  9. 一种电子设备，其中，包括：
    至少一个存储器；
    至少一个处理器；
    至少一个程序；
    所述程序被存储在所述存储器中，处理器执行所述至少一个程序以实现一种价格数据的处理方法；
    其中，所述价格数据的处理方法包括：
    获取待预测的原始数据；其中，所述原始数据包括目标报告数据和目标交易数据；
    根据所述目标交易数据构建指标因子特征；
    根据所述目标报告数据构建舆论因子特征；
    对多个所述指标因子特征和多个所述舆论因子特征进行筛选处理，得到多个量化交易特征；
    通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个分布式特征向量；
    将多个所述分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。
  10. 根据权利要求9所述的一种电子设备，其中，所述根据所述目标报告数据构建舆论因子特征，包括：
    对所述目标报告数据进行情感分类，得到报告情感类别；
    对所述目标报告数据进行文本特征提取，得到用于表征文本信息价值的第一参考值；
    对所述目标报告数据进行可读性评估，得到用于表征研报可读价值的第二参考值；
    根据所述报告情感类别、所述第一参考值和所述第二参考值得到所述舆论因子特征。
  11. 根据权利要求10所述的一种电子设备，其中，所述对所述目标报告数据进行文本特征提取，得到用于表征文本信息价值的第一参考值，包括：
    对所述目标报告数据进行分类处理，得到目标分词集、目标语句集、目标段落集和目标语法集；
    对所述目标分词集进行统计评分处理，得到分词评分值；
    对所述目标语句集进行统计评分处理，得到语句评分值；
    对所述目标段落集进行统计评分处理，得到段落评分值；
    对所述目标语法集进行统计评分处理，得到语法评分值；
    根据所述分词评分值、所述语句评分值、所述段落评分值和所述语法评分值得到所述第一参考值。
  12. 根据权利要求10和11任意一项所述的一种电子设备，其中，所述根据所述报告情感类别、所述第一参考值和所述第二参考值得到所述舆论因子特征，包括：
    将所述第一参考值和所述第二参考值进行相乘处理，得到目标参考值；
    根据所述报告情感类别对所述目标参考值进行取正处理或者取反处理，得到所述舆论因子特征。
  13. 根据权利要求9至11任意一项所述的一种电子设备，其中，所述第一神经网络模型包括：输入层和卷积层；所述通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个所述分布式特征向量，包括：
    将多个所述量化交易特征输入到所述第一神经网络模型中；
    通过所述输入层对每一所述量化交易特征进行预处理，得到对应的标准化数据；
    通过所述卷积层对所述标准化数据进行卷积处理，得到多个所述分布式特征向量。
  14. 根据权利要求13所述的一种电子设备，其中，所述通过所述卷积层对所述标准化数据进行卷积处理，得到多个所述分布式特征向量，包括：
    通过所述卷积层对所述标准化数据进行格式转换，得到标准格式数据；
    通过所述卷积层的卷积核对所述标准格式数据进行卷积，得到多个所述分布式特征向量。
  15. 一种存储介质，所述存储介质为计算机可读存储介质，其中，所述计算机可读存储介质存储有计算机可执行指令，所述计算机可执行指令用于使计算机执行一种价格数据的处理方法；
    其中，所述价格数据的处理方法包括：
    获取待预测的原始数据；其中，所述原始数据包括目标报告数据和目标交易数据；
    根据所述目标交易数据构建指标因子特征；
    根据所述目标报告数据构建舆论因子特征；
    对多个所述指标因子特征和多个所述舆论因子特征进行筛选处理，得到多个量化交易特征；
    通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个分布式特征向量；
    将多个所述分布式特征向量输入至预设的第二神经网络模型中进行价格预测处理，得到目标价格数据。
  16. 根据权利要求15所述的一种存储介质，其中，所述根据所述目标报告数据构建舆论因子特征，包括：
    对所述目标报告数据进行情感分类，得到报告情感类别；
    对所述目标报告数据进行文本特征提取，得到用于表征文本信息价值的第一参考值；
    对所述目标报告数据进行可读性评估，得到用于表征研报可读价值的第二参考值；
    根据所述报告情感类别、所述第一参考值和所述第二参考值得到所述舆论因子特征。
  17. 根据权利要求16所述的一种存储介质，其中，所述对所述目标报告数据进行文本特征提取，得到用于表征文本信息价值的第一参考值，包括：
    对所述目标报告数据进行分类处理，得到目标分词集、目标语句集、目标段落集和目标语法集；
    对所述目标分词集进行统计评分处理，得到分词评分值；
    对所述目标语句集进行统计评分处理，得到语句评分值；
    对所述目标段落集进行统计评分处理，得到段落评分值；
    对所述目标语法集进行统计评分处理，得到语法评分值；
    根据所述分词评分值、所述语句评分值、所述段落评分值和所述语法评分值得到所述第一参考值。
  18. 根据权利要求16和17任意一项所述的一种存储介质，其中，所述根据所述报告情感类别、所述第一参考值和所述第二参考值得到所述舆论因子特征，包括：
    将所述第一参考值和所述第二参考值进行相乘处理，得到目标参考值；
    根据所述报告情感类别对所述目标参考值进行取正处理或者取反处理，得到所述舆论因子特征。
  19. 根据权利要求15至17任意一项所述的一种存储介质，其中，所述第一神经网络模型包括：输入层和卷积层；所述通过预设的第一神经网络模型对多个所述量化交易特征进行特征提取，得到多个所述分布式特征向量，包括：
    将多个所述量化交易特征输入到所述第一神经网络模型中；
    通过所述输入层对每一所述量化交易特征进行预处理，得到对应的标准化数据；
    通过所述卷积层对所述标准化数据进行卷积处理，得到多个所述分布式特征向量。
  20. 根据权利要求19所述的一种电子设备，其中，所述通过所述卷积层对所述标准化数据进行卷积处理，得到多个所述分布式特征向量，包括：
    通过所述卷积层对所述标准化数据进行格式转换，得到标准格式数据；
    通过所述卷积层的卷积核对所述标准格式数据进行卷积，得到多个所述分布式特征向量。
PCT/CN2022/090661 2022-02-22 2022-04-29 价格数据的处理方法和装置、电子设备、存储介质 WO2023159756A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210160991.4 2022-02-22
CN202210160991.4A CN114519613B (zh) 2022-02-22 2022-02-22 价格数据的处理方法和装置、电子设备、存储介质

Publications (1)

Publication Number Publication Date
WO2023159756A1 true WO2023159756A1 (zh) 2023-08-31

Family

ID=81598141

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/090661 WO2023159756A1 (zh) 2022-02-22 2022-04-29 价格数据的处理方法和装置、电子设备、存储介质

Country Status (2)

Country Link
CN (1) CN114519613B (zh)
WO (1) WO2023159756A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116957635A (zh) * 2023-09-20 2023-10-27 中国华能集团清洁能源技术研究院有限公司 电力价格获取方法、装置、电子设备及存储介质

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115545790B (zh) * 2022-10-20 2023-06-09 北京宽客进化科技有限公司 价格数据预测方法、装置、电子设备及存储介质

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109360097A (zh) * 2018-09-28 2019-02-19 中山大学 基于深度学习的股票预测方法、装置、设备及存储介质
US20200097820A1 (en) * 2017-05-16 2020-03-26 Samsung Electronics Co., Ltd. Method and apparatus for classifying class, to which sentence belongs, using deep neural network
CN113919945A (zh) * 2021-09-17 2022-01-11 平安科技（深圳）有限公司 基于神经网络的数据分析方法、装置、设备及存储介质

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11537845B2 (en) * 2017-04-12 2022-12-27 Yodlee, Inc. Neural networks for information extraction from transaction data
CN110390408B (zh) * 2018-04-16 2024-03-05 北京京东尚科信息技术有限公司 交易对象预测方法和装置
CN108647823A (zh) * 2018-05-10 2018-10-12 北京航空航天大学 基于深度学习的股票数据分析方法和装置
CN111368088A (zh) * 2020-03-31 2020-07-03 成都信息工程大学 一种基于深度学习的文本情感分类方法
CN113435204A (zh) * 2021-02-02 2021-09-24 上海卡方信息科技有限公司 基于新闻信息的股票价格涨跌预测方法
CN113643060A (zh) * 2021-08-12 2021-11-12 工银科技有限公司 产品价格的预测方法和装置

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200097820A1 (en) * 2017-05-16 2020-03-26 Samsung Electronics Co., Ltd. Method and apparatus for classifying class, to which sentence belongs, using deep neural network
CN109360097A (zh) * 2018-09-28 2019-02-19 中山大学 基于深度学习的股票预测方法、装置、设备及存储介质
CN113919945A (zh) * 2021-09-17 2022-01-11 平安科技（深圳）有限公司 基于神经网络的数据分析方法、装置、设备及存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
PEI MANRU; ZHANG LIWEN: "Research on Stock Trend Prediction Based on Investor Sentiment and Deep Learning", FINANCIAL DEVELOPMENT, no. 01, 15 March 2020 (2020-03-15), pages 78 - 97, XP009549397, ISSN: 2096-7160 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116957635A (zh) * 2023-09-20 2023-10-27 中国华能集团清洁能源技术研究院有限公司 电力价格获取方法、装置、电子设备及存储介质
CN116957635B (zh) * 2023-09-20 2023-12-26 中国华能集团清洁能源技术研究院有限公司 电力价格获取方法、装置、电子设备及存储介质

Also Published As

Publication number Publication date
CN114519613B (zh) 2023-07-25
CN114519613A (zh) 2022-05-20

Similar Documents

Publication Publication Date Title
CN108334605B (zh) 文本分类方法、装置、计算机设备及存储介质
WO2022007823A1 (zh) 一种文本数据处理方法及装置
CN113641820B (zh) 基于图卷积神经网络的视角级文本情感分类方法及系统
CN110069709B (zh) 意图识别方法、装置、计算机可读介质及电子设备
CN111143576A (zh) 一种面向事件的动态知识图谱构建方法和装置
WO2023159756A1 (zh) 价格数据的处理方法和装置、电子设备、存储介质
CN110781686B (zh) 一种语句相似度计算方法、装置及计算机设备
WO2024022354A1 (zh) 结合rpa及ai实现ia的对象推荐方法、装置及存储介质
CN109086265B (zh) 一种语义训练方法、短文本中多语义词消歧方法
CN111475622A (zh) 一种文本分类方法、装置、终端及存储介质
WO2023236977A1 (zh) 一种数据处理方法及相关设备
CN111428513A (zh) 一种基于卷积神经网络的虚假评论分析方法
CN113449204A (zh) 基于局部聚合图注意力网络的社会事件分类方法、装置
CN114547299A (zh) 一种基于复合网络模型的短文本情感分类方法及装置
CN110826327A (zh) 情感分析方法、装置、计算机可读介质及电子设备
Kumar et al. Emotion analysis of news and social media text for stock price prediction using svm-lstm-gru composite model
WO2021238568A1 (zh) 一种更新参数的方法、装置及存储介质
AlAjlan et al. Machine learning approach for threat detection on social media posts containing Arabic text
CN113656563A (zh) 一种神经网络搜索方法及相关设备
CN115659951B (zh) 基于标签嵌入的语句情感分析方法、装置以及设备
CN110852066B (zh) 一种基于对抗训练机制的多语言实体关系抽取方法及系统
US20210241147A1 (en) Method and device for predicting pair of similar questions and electronic equipment
CN114491076A (zh) 基于领域知识图谱的数据增强方法、装置、设备及介质
CN113821571A (zh) 基于bert和改进pcnn的食品安全关系抽取方法
CN110309285B (zh) 自动问答方法、装置、电子设备和存储介质

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: 22928028

Country of ref document: EP

Kind code of ref document: A1