CN114168575A - Public opinion analysis method and system in financial field - Google Patents
Public opinion analysis method and system in financial field Download PDFInfo
- Publication number
- CN114168575A CN114168575A CN202111383790.2A CN202111383790A CN114168575A CN 114168575 A CN114168575 A CN 114168575A CN 202111383790 A CN202111383790 A CN 202111383790A CN 114168575 A CN114168575 A CN 114168575A
- Authority
- CN
- China
- Prior art keywords
- data
- financial
- public opinion
- event
- field
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a public opinion analysis method and a system in the financial field, which mainly comprise data formulation and collection, training data marking, deep learning model construction and public opinion analysis, and particularly obtain large-scale financial affairs news data by a web crawler algorithm according to the requirements of the financial vertical field, store the data into a MySQL database, and perform data cleaning and pretreatment, manual marking and screening. The method comprises the steps of obtaining vector representation of characters by utilizing a pretrained Bert model in the financial field, extracting local information characteristics in the texts by an attention mechanism in a natural language processing technology to obtain fusion vector representation which can be understood by a computer, analyzing public sentiments by a model, and returning relevant event description, event main bodies, event types and emotion analysis. By applying the technical scheme, the accuracy and the response rate of public opinion analysis are improved by using a supervised model for marking and screening data, the industrial requirements are met, and the uniform shared interactive information related to events is obtained.
Description
Technical Field
The invention relates to a technology for explaining consulting information and analyzing financial public sentiment by a computer, in particular to a public sentiment analysis method and system in the financial field based on supervised deep learning algorithm joint training.
Background
Public sentiments related to the financial industry present a surge situation, the appearance time is relatively centralized, the information interaction amount is large, and the interaction times are frequent. The generation, expansion and transmission of financial public opinions have important influence on investors, financial institutions, financial industry and even macroscopic economic operation, and the financial crisis events can be caused by small credit crisis. Therefore, effective monitoring and analysis of the financial public sentiment can grasp the rhythm of expected management, reduce and avoid outbreak of the crisis of the financial public sentiment, and have important reference values in the aspects of assisting netizens or financial institutions to invest, assisting governments to grasp public sentiment wind directions of stock markets, and enabling stocks or enterprises to grasp asset changes of certain financial companies.
Most of traditional public opinion analysis systems are based on specialized financial dictionaries, and text emotion is judged in a dictionary mode. However, the semantics in the financial public opinion text cannot be well understood based on the dictionary mode, the difficulty of sensing financial risks is increased, and the semantic understanding and reasoning capability does not achieve the expected effect. Meanwhile, the use of the unstructured public data still stays at the data extraction level, so that great promotion space is provided.
With the rise of deep learning, the deep neural network obtains advanced results in tasks such as emotion analysis and text classification. Therefore, many existing financial public opinion analysis systems adopt a neural network-based method, which is mainly divided into two types: one method is based on a convolutional neural network method, and the performance of local features of the text can be captured by using the convolutional neural network to extract local continuous phrase features in the text. The other method is based on a long and short memory neural network method, and short-term context information and long-distance dependence can be effectively captured by using the long and short memory neural network, so that sequence information in the text can be extracted. However, the two methods have the defects that long-distance dependence in the financial text, such as expression of some emotion turning appearing in the text, cannot be fully extracted based on the convolutional neural network method; the long and short type memory neural network method tends to information at the beginning and the end of the text, has an unobtrusive effect on extracting local information in the text, and cannot fuse multi-scale information in the text.
From the technical details in the field of financial public opinion analysis, there are four common tasks: one event description extraction: giving a text T, extracting the things happening at a specific time and a specific place in the T, and playing an important decision reference role in the public opinion monitoring field and the financial field; two event main bodies are extracted: giving a text T, and extracting the main bodies of all events in the T, wherein the main bodies comprise company names, organization names and the like; the three event types are classified: given a text T, classifying the events extracted from T, wherein the event types comprise: stock price change, high-pipe change, investment co-purchase and the like; four emotion analyses are carried out: given a text T, and combining specific events extracted from the T, predicting influence analysis on event bodies extracted from the T. Emotional analysis typically contains three categories of results, positive, negative, and neutral.
The event description extraction is the most difficult and the most critical task among the four tasks. The challenge of event extraction is represented by the complexity of text, where the input text may be a sentence, a paragraph or a chapter, and the event to be extracted may be a phrase, a sentence or a sentence group. The accurate extraction of events also has great influence on judgment of emotion analysis, and understanding of the model to natural language can be well tested. But most in the financial field are unstructured data and the construction of data sets for such tasks is relatively difficult. How to effectively evaluate the model effect is under intensive study.
Practice proves that the accuracy of the existing financial public opinion analysis model cannot meet the complex requirements of the financial field in the industry, and the analysis model has low response speed and cannot meet the requirements of real-time public opinion analysis.
Disclosure of Invention
In view of the above deficiencies of the prior art, the present invention aims to provide a public opinion analysis method and system in the financial field, which solves the problems of poor satisfaction and low efficiency of public opinion analysis via a computer in the financial field.
The invention achieves a technical solution of the above purpose: a public opinion analysis method in the financial field is characterized by comprising the following steps: data are formulated and collected, corresponding to the requirements of the financial vertical field, large-scale financial and economic news data are obtained through a web crawler algorithm and stored in a MySQL database, and data cleaning and preprocessing are carried out;
marking training data, namely manually marking and screening financial news data in a MySQL database;
constructing a deep learning model, namely acquiring vector representation of characters by utilizing a pretrained Bert model in the financial field, and extracting local information characteristics in a text by an attention mechanism in a natural language processing method to obtain fusion vector representation which can be understood by a computer;
and the public opinion analysis is used for analyzing public opinion news by utilizing a built public opinion database and a trained deep learning model and returning event description, event main bodies, event types and emotion analysis.
The other technical solution of the invention for realizing the above purpose is as follows: the utility model provides a public opinion analytic system in finance field which characterized in that includes:
the data formulating and collecting unit is used for acquiring large-scale financial and economic news data through a web crawler algorithm according to the requirements of the financial vertical field, storing the large-scale financial and economic news data into a MySQL database, and cleaning and preprocessing the data;
the training data marking unit is used for manually marking and screening financial news data in the MySQL database;
the deep learning model building unit is used for acquiring vector representation of characters by utilizing a pretrained Bert model in the financial field, and extracting local information characteristics in the text by an attention mechanism in a natural language processing method to obtain fusion vector representation which can be understood by a computer;
and the public opinion analyzing unit is used for analyzing the public opinion news by utilizing the built public opinion database and the trained deep learning model and returning event description, event main bodies, event types and emotion analysis.
The new technical solution of public opinion analysis in the financial field has obvious progress: the method and the system use high-quality labeled data to train to obtain a supervised model, thereby improving the accuracy rate and greatly improving the response rate to less than 1 second; through the model obtained by training, the requirement of uniformly sharing the interactive information can be met, and the event main body, the emotion analysis result and the reason for generating the emotion analysis result can be fed back at one time.
Drawings
Fig. 1 is a topological architecture diagram of a typical finance public opinion analysis system.
Fig. 2 is a flowchart illustrating major steps of the financial public opinion analysis according to the present invention.
Fig. 3 is a schematic structural diagram of a deep learning model constructed by the financial public opinion analysis method of the present invention.
Detailed Description
The following detailed description of the embodiments of the present invention is provided in conjunction with the accompanying drawings to make the technical solution of the present invention easier to understand and grasp, so as to define the protection scope of the present invention more clearly.
The invention provides a financial field public opinion analysis method and system based on deep learning algorithm joint training, aiming at the current situation that the development and the start of the traditional financial field public opinion analysis method are low and the related requirements of the financial field cannot be met, and solving the problems of poor requirement satisfaction and low efficiency.
As shown in fig. 1, a typical finance public opinion analysis system generally includes five modules, namely, data acquisition and cleaning, feature extraction, event description and event subject extraction, event type classification and emotion analysis. The respective components are further briefly described below.
Data acquisition and cleaning: the module performs data capture and cleaning through a crawler framework. The crawler framework is composed of a crawler task controller, a sub-crawler instance, a bloom filter, a database asynchronous queue and a MySQL database set group. And storing the data captured by the crawler framework into a MySQL database, and randomly extracting samples from the database and performing data cleaning to obtain a training data set. The cleaning rule comprises: eliminating invalid texts; removing repeated texts; deleting overlong and overlong short texts; carrying out reproduction and simplification; and removing meaningless symbols containing url and pinyin.
Feature extraction: this module converts the input, natural language form of public opinion news into fixed-dimension vectors for subsequent processing by the machine. Early commonly used methods were word representation methods such as one-hot representation and distributed word vectors; the context-based word representation method pre-trained by a large-scale corpus has also been widely used in recent years, such as ELMo, GPT, Bert, and the like. Meanwhile, in order to better represent information such as semantic syntax, the word vector may be combined with linguistic features such as part-of-speech tags, named entities, question types, and the like to represent the word vector at a finer granularity. The word vector of the text obtained through the embedded coding layer coding is represented by extracting more context information through a common neural network model (a Recurrent Neural Network (RNN), a Convolutional Neural Network (CNN) and a Transformer structure based on a multi-head self-attention mechanism), and the obtained word vector can effectively extract the features in the text by utilizing the mutual relation between words.
Event description and event subject extraction: the text features obtained by the feature extraction module adopt a bidirectional long and short memory neural network (Bi-LSTM) in combination with a Conditional Random Field (CRF) model to solve the extraction of event description and event subject. Bi-LSTM is to adopt forward and backward LSTM for each word sequence, and then combine the outputs at the same time. Thus for each instant, it corresponds to information in the forward and backward directions. The CRF is more accurate in word segmentation and classification information through transfer matrix constraint.
Event type classification: the module is based on the event description extracted from the event description and event subject extraction module and the representation of the input text, and adds an attention mechanism model to perform a multi-label classification task to obtain the event type.
Emotion analysis: the module is also based on the event description extracted from the event description and event subject extraction module, the characterization of the event subject and the input text, and adds an attention mechanism model to perform a text classification task to obtain the emotion classification.
As can be seen, the typical financial public opinion analysis system performs event extraction and emotion analysis by means of sub-task decomposition, and cannot meet the requirement of uniformly sharing interactive information by 'event subject-event description-event type-emotion analysis', and cannot give the reason why the event subject generates what emotion due to what event. Namely (What, How, Why).
With reference to fig. 2 and fig. 3, the invention innovatively provides a financial field public opinion analysis method and system based on deep learning algorithm joint training for the above-mentioned places which cannot meet the demand, so that the response efficiency and accuracy are improved, the event description in the text can be accurately extracted from the input public opinion news data, and the emotion analysis and event type are predicted according to the event description and the event subject. The public opinion analysis method mainly comprises four main steps of data formulation and collection, training data labeling, deep learning model construction and public opinion analysis.
In the steps, the data specification and collection refers to the fact that large-scale financial news data are acquired through a web crawler algorithm and stored in a MySQL database according to the requirements of the financial vertical field, and data cleaning and preprocessing are performed.
And marking training data, and manually marking and screening financial news data in the MySQL database. The content marked manually comprises event description, event main body, event type and emotion analysis of corresponding financial news data; because the implementation result of the manual labeling has a large influence on the deep learning model, in order to ensure the accuracy of the model analysis, the manual screening is needed to avoid errors. And the screened marked content accounts for more than 30% of the whole marked content.
The deep learning model is constructed by firstly utilizing a pre-trained Bert (bidirectional Encoder retrieval from transformations) model in the financial field to obtain vector representation of characters, and the model has the characteristics of good character processing effect, small model and high efficiency aiming at the financial field. And secondly, extracting local information features in the text through an Attention mechanism (Attention) in a natural language processing method to obtain fusion vector representation which can be understood by a computer.
And the public opinion analysis is to utilize the built public opinion database and the trained deep learning model to search the latest public opinion news containing the keywords in the database according to the keywords or directly input a piece of public opinion news, and then return Event description (eventdescription), Event subject (Eventsubject), emotion analysis (Sentiment analysis) and Event type (Event classification) in the news, namely, uniformly share the interactive information.
Corresponding to the public opinion analysis method in the financial field, the system is realized by the programmed modification of a computer. The system architecture body formed by the specific programming comprises the following four parts: the system comprises a data formulation and collection unit, a training data labeling unit, a deep learning model construction unit and a public opinion analysis unit. Each component inherits the elements of the traditional public opinion analysis system and carries out optimization and improvement, which is mainly embodied in the artificial marking of training data and the optimization of deep learning models based on the marked training data, and the function realization of the component can refer to the specific explanation of the public opinion analysis method.
Furthermore, the training data labeling unit also comprises a labeling module used for labeling incomplete sets of the screening results which are not completely labeled manually or need to be discussed further in emotion analysis, so that the training data in the deep learning model construction can be removed.
From a more intuitive and visualized example, a computer system applying the public opinion analysis method and system in the financial field can provide a series of implementation cases of forward reasoning.
1. In the public opinion news input program, for example, directly inputting a piece of news: "the company E has a good, the company A has a word-open and a word-stop, and the net profit of the company in the first half year is 33.53 hundred million yuan, which is increased by 3.07 percent on a par. "; the model will automatically extract the event body, event description, event type and emotion analysis of the news, namely: ("company a", "open one-word stop, company's net profit of 33.53 billion yuan in the first half year, increase by 3.07%", "finance-profit publication", "negative").
2. For public opinion news with complex semantics and messy company entities, the traditional public opinion analysis model can only identify the company entities in the news in sequence, and cannot provide corresponding relations. For example: "the power plate moves and rises in a different way, company B and company C rise and stop, company D rises by 6%, and company E and company F rise. ". The model of the invention can answer the mutual influence in one time: the "power board abnormal movement" means "B company", "power board abnormal movement pulling up, rising and stopping", "stock-stock price change", "front face", "C company", "power board abnormal movement pulling up, rising and stopping", "stock-stock price change", "front face", "D company", "power board abnormal movement pulling up, rising by 6%", "stock-stock price change", "front face"), ("E company", "power board abnormal movement pulling up, following rising", "stock-stock price change", "front face"), ("F company", "power board abnormal movement pulling up, following rising", "stock-stock price change", "front face").
3. For multi-perspective public sentiment news, the traditional public sentiment analysis model can only give the overall emotion of the news and cannot be connected with the analysis and judgment of specific events in the news. For example: "the G company absorbs and merges the H company according to the stockholders of the G company and the H company, the G company continues after the absorption and the mergence, and the H company releases and logs out. ". The model of the invention can be combined with specific events in news to analyze the emotion of the company entity corresponding to the event. ("local cargo shipping Limited", "absorb merge H", "invest and buy-and buy restructure", "positive"), ("H", "decommissioning logoff", "other events", "negative").
4. For chapter-level public opinion news and macro news, the traditional public opinion analysis model cannot process overlong public opinion news. For example: "company I1.18% stock right faces judicial change after two times of streaming, last month is less than 2200 ten thousand shares by the fourth capital east", news links:
. The model of the present invention can also analyze the specific event subject, event description, event type and emotion at one time, wherein the 4727.79 ten thousand shares held by the company "I" and "J" will face the change of judicial expertise and account for 1.18% of the total shares in the line, "share-share change" and "negative" respectively.
5. If the system searches by keywords, such as inputting "car", the system will search the database for the latest public opinion news related to "car" by matching algorithm. For example: the message of 11 months and 4 days reports in XX, only the top line is 3 days, and the first payment financing scheme of the T company 0 changes. Day 11 and 4, the official network of the company T displays that, in the online financing lease plan of day 11 and 1, the option 0 first payment has been cancelled, and the minimum first payment adjustment is 10%. For this purpose, the car in the financial network contacts the relevant person in charge of company T, but the other party does not respond. ". The model will also automatically extract the event body, event description, event type, and emotional analysis of the news. ("T corporation", "11 month 1 day on financing lease plan, 0 first payment option has been cancelled, lowest first payment adjusted to 10%", "other events", "neutral").
In summary, the public opinion analyzing method and system in the financial field according to the present invention can be seen in detail with the embodiments shown in the drawings, which have outstanding substantive features and significant progress. The method and the system use high-quality labeled data to train to obtain a supervised model, thereby improving the accuracy rate and greatly improving the response rate to less than 1 second; through the model obtained by training, the requirement of uniformly sharing the interactive information can be met, and the event main body, the emotion analysis result and the reason for generating the emotion analysis result can be fed back at one time.
Claims (8)
1. A public opinion analysis method in the financial field is characterized by comprising the following steps: data are formulated and collected, corresponding to the requirements of the financial vertical field, large-scale financial and economic news data are obtained through a web crawler algorithm and stored in a MySQL database, and data cleaning and preprocessing are carried out;
marking training data, namely manually marking and screening financial news data in a MySQL database;
constructing a deep learning model, namely acquiring vector representation of characters by utilizing a pretrained Bert model in the financial field, and extracting local information characteristics in a text by an attention mechanism in a natural language processing method to obtain fusion vector representation which can be understood by a computer;
and the public opinion analysis is used for analyzing public opinion news by utilizing a built public opinion database and a trained deep learning model and returning event description, event main bodies, event types and emotion analysis.
2. The public opinion analysis method in financial field as claimed in claim 1, wherein: in data formulation and collection, samples are randomly extracted from a MySQL database for data cleaning, and the rule of data cleaning comprises the steps of removing invalid texts, removing repeated texts, deleting too long and too short texts, changing from complex to simple and removing meaningless symbols containing url and pinyin.
3. The public opinion analysis method in financial field as claimed in claim 1, wherein: in the training data annotation, the content labeled manually comprises event description, event subject, event type and emotion analysis of the corresponding financial news data.
4. The public opinion analysis method in financial field as claimed in claim 1, wherein: in training data labeling, the labeled content obtained through manual screening accounts for more than 30% of the total labeled content.
5. The public opinion analysis method in financial field as claimed in claim 1, wherein: in the deep learning model, vector representation of characters obtained through a Bert model comprises event description and an event main body, fusion vector representation extracted through an attention mechanism comprises emotion analysis and an event type, and a model for feeding back unified shared interaction information is obtained through training based on training data screened after marking.
6. The utility model provides a public opinion analytic system in finance field which characterized in that includes:
the data formulating and collecting unit is used for acquiring large-scale financial and economic news data through a web crawler algorithm according to the requirements of the financial vertical field, storing the large-scale financial and economic news data into a MySQL database, and cleaning and preprocessing the data;
the training data marking unit is used for manually marking and screening financial news data in the MySQL database;
the deep learning model building unit is used for acquiring vector representation of characters by utilizing a pre-trained BERT model in the financial field, and extracting local information characteristics in the text by an attention mechanism in a natural language processing method to obtain fusion vector representation which can be understood by a computer;
and the public opinion analyzing unit is used for analyzing the public opinion news by utilizing the built public opinion database and the trained deep learning model and returning event description, event main bodies, event types and emotion analysis.
7. The public opinion analysis system in finance field according to claim 6, wherein: and a data cleaning rule is set in the data formulation and collection unit, and the data cleaning rule comprises the steps of eliminating invalid texts, removing repeated texts, deleting overlong and overlong texts, changing a complex text into a simple text and removing meaningless symbols containing url and pinyin.
8. The public opinion analysis system in finance field according to claim 6, wherein: in the training data labeling unit, the system also comprises a labeling module used for labeling incomplete sets of screening results which are not completely labeled manually or need to be discussed further in emotion analysis.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111383790.2A CN114168575A (en) | 2021-11-22 | 2021-11-22 | Public opinion analysis method and system in financial field |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111383790.2A CN114168575A (en) | 2021-11-22 | 2021-11-22 | Public opinion analysis method and system in financial field |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114168575A true CN114168575A (en) | 2022-03-11 |
Family
ID=80479832
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111383790.2A Pending CN114168575A (en) | 2021-11-22 | 2021-11-22 | Public opinion analysis method and system in financial field |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114168575A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116821489A (en) * | 2023-06-21 | 2023-09-29 | 易方达基金管理有限公司 | Stock screening method and system |
| CN117788136A (en) * | 2023-11-24 | 2024-03-29 | 浙江孚临科技有限公司 | Financial wind control system based on blockchain and public opinion |
| CN118014734A (en) * | 2024-01-31 | 2024-05-10 | 华南理工大学 | Training method, analysis method, device, system and medium of financial analysis model |
-
2021
- 2021-11-22 CN CN202111383790.2A patent/CN114168575A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116821489A (en) * | 2023-06-21 | 2023-09-29 | 易方达基金管理有限公司 | Stock screening method and system |
| CN116821489B (en) * | 2023-06-21 | 2024-05-10 | 易方达基金管理有限公司 | Stock screening method and system |
| CN117788136A (en) * | 2023-11-24 | 2024-03-29 | 浙江孚临科技有限公司 | Financial wind control system based on blockchain and public opinion |
| CN118014734A (en) * | 2024-01-31 | 2024-05-10 | 华南理工大学 | Training method, analysis method, device, system and medium of financial analysis model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114168575A (en) | Public opinion analysis method and system in financial field | |
| CN103778215A (en) | Stock market forecasting method based on sentiment analysis and hidden Markov fusion model | |
| CN111274814B (en) | Novel semi-supervised text entity information extraction method | |
| CN110442823A (en) | Website classification method, Type of website judgment method, storage medium and intelligent terminal | |
| Nagar et al. | Using text and data mining techniques to extract stock market sentiment from live news streams | |
| Weng et al. | Stock price prediction based on LSTM and BERT | |
| CN113255360A (en) | Document rating method and device based on hierarchical self-attention network | |
| CN114416969A (en) | LSTM-CNN online comment sentiment classification method and system based on background enhancement | |
| CN112527866A (en) | Stock trend prediction method and system based on text abstract emotion mining | |
| Yang et al. | Measuring digitalization capabilities using machine learning | |
| CN113505207B (en) | Machine reading understanding method and system for financial public opinion research report | |
| Gatchalee et al. | Thai text classification experiment using cnn and transformer models for timely-timeless content marketing | |
| Zhang et al. | A study of relationship between investor sentiment and stock price: Realization of investor sentiment classification based on bayesian model | |
| Zhang et al. | Stock trend forecasting method based on sentiment analysis and system similarity model | |
| Yang et al. | A general solution and practice for automatically constructing domain knowledge graph | |
| Ma et al. | A study on sentiment analysis for smart tourism | |
| CN113688233A (en) | Text understanding method for semantic search of knowledge graph | |
| Lai et al. | Chinese stock market prediction based on multifeature fusion and TextCNN | |
| Vaca et al. | Board of Directors' Profile: A Case for Deep Learning as a Valid Methodology to Finance Research | |
| Luo et al. | Research on credit risk assessment of listed companies in science and technology sector by introducing industry research report information | |
| Huang et al. | Digital Transformation Strategy for Financial Management of Entity Enterprises in the Information Age | |
| Tan et al. | Context-aware Embeddings for Stock Prediction with Visual Clues | |
| Brice et al. | Enhanced Text Classification Using DistilBERT with Low-Rank Adaptation: A Comparative Study | |
| CN118747216A (en) | Loan use text classification method and system based on convolutional neural network | |
| Shang et al. | Building public market opinion indices for electricity consumption prediction of Chinese nonferrous metal industry |
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 |