CN113177831A - Financial early warning system and method constructed by applying public data - Google Patents
Financial early warning system and method constructed by applying public data Download PDFInfo
- Publication number
- CN113177831A CN113177831A CN202110271010.9A CN202110271010A CN113177831A CN 113177831 A CN113177831 A CN 113177831A CN 202110271010 A CN202110271010 A CN 202110271010A CN 113177831 A CN113177831 A CN 113177831A
- Authority
- CN
- China
- Prior art keywords
- early warning
- financial
- public data
- data
- applying
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000012544 monitoring process Methods 0.000 claims abstract description 9
- 230000007246 mechanism Effects 0.000 claims description 20
- 238000010606 normalization Methods 0.000 claims description 12
- 239000012535 impurity Substances 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 230000000977 initiatory effect Effects 0.000 claims description 3
- 238000003780 insertion Methods 0.000 claims description 3
- 230000037431 insertion Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
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
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
-
- 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
- G06N3/08—Learning methods
Abstract
The invention relates to a financial early warning system constructed by applying public data, which comprises: the searching and alarming module is used for monitoring internet information to obtain input data; the warning sign identifying and analyzing module is used for extracting document data characteristics in the process of converting input data into warning signs so as to clearly identify event information related to early warning enterprises; the pre-warning degree module is used for predicting the warning degree information of an enterprise, early warning the financial risk and dividing the risk level; the invention also discloses a financial early warning method constructed by applying the public data, which can efficiently, quickly and accurately obtain early warning information.
Description
Technical Field
The invention belongs to the technical field of data processing, and particularly relates to a financial early warning system constructed by using public data.
Background
The financial risk early warning system is an early warning system established by a financial supervision authority for better and effectively monitoring financial operation institutions and is used for early warning and forecasting the financial risk which may occur. It is a dual function of risk management and operation evaluation, especially for the prevention and warning of finance and business of listed companies. The meaning of the method is that according to the related regulatory law and the operation principle of the company financial business, a set of pre-warning function, index (Indicator), critical Value or reference Value (default Value), or discriminant model, etc. are selected and defined for a plurality of variables, and in the aspect of financial pre-warning operation, for the prediction of the financial crisis company, the supervisor mostly monitors the operation condition of the financial operation organization by the traditional financial index.
However, the existing early warning system only performs early warning on financial risks according to the volume price data directly related to finance, neglects the influence of the mass text data in the internet and has the problem of low early warning accuracy.
Disclosure of Invention
The invention aims to provide a financial early warning system constructed by applying public data, which can be used for early warning financial risks.
The second purpose of the invention is to provide a financial early warning method constructed by applying public data, which can provide an effective financial risk early warning mode.
The financial early warning system is constructed by applying public data, and comprises a searching alarm source module, a searching alarm source module and a monitoring module, wherein the searching alarm source module is used for monitoring internet information to obtain input data;
the warning sign identifying and analyzing module is used for extracting document data characteristics in the process of converting input data into warning signs so as to clearly identify event information related to early warning enterprises;
and the pre-alarming degree module is used for predicting the alarming degree information of the enterprise.
The present invention is also characterized in that,
the internet information comprises official business networks, financial and newspaper, transactions, news, impurities, media, networks and public opinion public data.
The alert level comprises two levels of alert and no alert.
The second technical scheme adopted by the invention is that the financial early warning method constructed by applying public data is implemented according to the following steps:
step 1, monitoring internet information in real time to obtain input data;
step 2, document data characteristics of a process of changing input data into warning signals are extracted by using a BERT model so as to clarify event information related to early warning enterprises;
and 3, predicting the police degree information of the enterprise by using the LSTM and dividing different police degree grades.
The present invention is also characterized in that,
in the step 1, the internet information comprises official business networks, financial and newspaper, transactions, news, impurities, media, networks and public opinion public data.
The specific process of the step 2 is as follows:
step 2.1, word embedding is carried out, and the expression of word embedding is as follows:
X＝Epos(onehot(S)+Eseg) (1)
in the formula (1), onehot represents word insertion, EsegRepresenting sentence embedding, EposIndicating position embedding;
step 2.2, defining a single attention machine mechanism:
in the formula (2), WQ、WK、WVIs a trainable parameter; softmax represents a softmax normalization function, mapping values to the range of 0-1; d represents the dimension of the vector;
step 2.3, combining a plurality of single attention mechanisms to form a multi-head attention mechanism layer, wherein the expression of the multi-head attention mechanism layer is as follows:
MultiHead(X)＝Concat(Att1,...,Atth)WO (3)
in the formula (3), Concate represents a link operation, h represents the number of attention heads, WORepresenting trainable parameters;
step 2.4, the output vector of the multi-head attention mechanism layer is added with the word in an embedding mode to form a residual error network, and the final output of the attention mechanism layer is obtained through batch normalization:
Matt＝BN(MultiHead(X)+X) (4)
in formula (4), BN represents batch normalization;
step 2.5, the final output obtained in step 2.4 passes through a full connection layer with a residual error network and batch normalization to obtain a document file characteristic B:
B＝BERT(X)＝BN(Dense(Matt)+Matt) (5)；
for a document S with n words, the character file characteristic B ═ B1,…,bn+1},
Where d represents the dimension of the vector.
The specific process of the step 3 is as follows:
step 3.1, assume the hidden layer output h of the last LSTMt-1And the tth document vector BtThen LSTM can be expressed as:
ft＝σ(Wf·Y+bf) (7)
it＝σ(Wi·Y+bi) (8)
ot＝σ(Wo·Y+bo) (9)
ct＝ft⊙ct-1+it⊙tanh(Wc·Y+bc) (10)
ht＝ot⊙tanh(ct) (11)
in the formula, Wi，Wf，Wo，Wc，bi，bf，bo，bcIs a trainable parameter, σ is a sigmod initiation function,. indicates a Hadamard product, ctRepresents the output of the t-th document;
step 3.2, vector h obtained by step 3.1tPredicting the alert level of the enterprise:
lt＝softmax(Dense(ht)) (12)。
when l istA warning is indicated by more than or equal to 0.5, namely abnormal; when l ist< 0.5 indicates no alarm, i.e., normal.
The invention has the beneficial effects that:
(1) the financial early warning system constructed by applying the public data avoids the problem of low accuracy caused by the early warning mode of financial risks according to the volume price data directly related to finance;
(2) according to the financial early warning method established by applying the public data, the police degree information of a prediction enterprise is generated by extracting the characteristics of mass text data in the Internet, and the accuracy of financial risk prediction is improved.
Drawings
FIG. 1 is a flow chart of a financial early warning method using public data construction according to the present invention;
fig. 2 is a flowchart of step 2 in the financial early warning method constructed by applying public data according to the present invention.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention relates to a financial early warning system constructed by applying public data, which comprises: the system comprises a searching and warning source module, a searching and warning module and a warning display module, wherein the searching and warning source module is used for monitoring internet information to obtain input data, and the internet information comprises public data such as official networks, financial reports, transactions, news, impurities, media, networks, public opinions and the like;
the warning sign identifying and analyzing module is used for extracting document data characteristics in the process of converting input data into warning signs so as to clearly identify event information related to early warning enterprises;
the pre-alarming degree module is used for predicting the alarming degree information of the enterprise, and the alarming degree grade is divided into two grades of alarming and non-alarming.
The invention provides a financial early warning method constructed by applying public data, which is realized by a financial early warning system constructed by applying the public data, and is specifically implemented according to the following steps as shown in figure 1:
step 1, monitoring internet information in real time to obtain input data, wherein the internet information comprises public data such as enterprise official networks, financial newspapers, transactions, news, impurities, media, networks, public opinions and the like;
step 2, extracting document data characteristics in the process of converting input data into warning signals by using a BERT model pre-trained by financial news data so as to clarify event information related to early warning enterprises;
BERT is a language model proposed by google in 2018, which improves the result on a specific task by pre-training on a large number of documents, and BERT network mainly consists of three parts, a word embedding layer, an attention mechanism layer and a full link layer. The word embedding layer converts the words into codes, and the codes pass through a multi-layer network module consisting of an attention mechanism layer and a full connection layer, and are finally obtained output. The word embedding layer includes three kinds of word embedding, sentence embedding, and position embedding. The word embedding is to code the words, the sentence embedding is to express different sentences by using codes, and the position embedding is to use the learned position embedding;
step 2.1, assume that the text file set S of n text files is ═ S1,s2,…,snAnd (4) sorting the text file set according to time, and embedding words, wherein the word embedding expression is as follows:
X＝Epos(onehot(S)+Eseg) (1)
in the formula (1), onehot represents word insertion, EsegRepresenting sentence embedding, EposIndicating position embedding;
step 2.2, defining a single attention machine mechanism:
in the formula (2), WQ、WK、WVIs a trainable parameter; softmax represents a softmax normalization function, mapping values to the range of 0-1; d represents the dimension of the vector;
step 2.3, combining a plurality of single attention mechanisms to form a multi-head attention mechanism layer, wherein the expression of the multi-head attention mechanism layer is as follows:
MultiHead(X)＝Concat(Att1(X),...,Atth(X))WO (3)
in the formula (3), Concate represents a link operation, h represents the number of attention heads, WORepresenting trainable parameters;
step 2.4, the output vector of the multi-head attention mechanism layer is added with the word in an embedding mode to form a residual error network, and the final output of the attention mechanism layer is obtained through batch normalization:
Matt＝BN(MultiHead(X)+X) (4)
in formula (4), BN represents batch normalization;
step 2.5, the final output obtained in step 2.4 passes through a full connection layer with a residual error network and batch normalization to obtain a document file characteristic B:
B＝BERT(X)＝BN(Dense(Matt)+Matt) (5)；
for a document S with n words, the character file characteristic B ═ B1,…,bn+1},
Wherein d represents the dimension of the vector, and Dense represents the fully-connected layer;
step 3, predicting the police degree information of the enterprise by using the LSTM, and dividing different police degree grades;
long-term memory (LSTM) is a neural network used to process column data. He can process the data that changes compared to a normal network. For example, the meaning of a certain document has different meanings due to different contents of the previous document, and the LSTM can well solve the problems;
step 3.1, since there are two inputs for each LSTM unit, assume that the hidden layer output h of the last LSTM ist-1And the tth document vector BtThen LSTM can be expressed as:
ft＝σ(Wf·Y+bf) (7)
it＝σ(Wi·Y+bi) (8)
ot＝σ(Wo·Y+bo) (9)
ct＝ft⊙ct-1+it⊙tanh(Wc·Y+bc) (10)
ht＝ot⊙tanh(ct) (11)
in the formula, Wi，Wf，Wo，Wc，bi，bf，bo，bcIs a trainable parameter, σ is a sigmod initiation function,. indicates a Hadamard product, ctRepresents the output of the t-th document;
step 3.2, vector h obtained by step 3.1tPredicting the alert level of the enterprise:
lt＝softmax(Dense(ht)) (12)
when l istA warning is indicated by more than or equal to 0.5, namely abnormal;
when l ist< 0.5 indicates no alarm, i.e., normal.
Claims (8)
1. A financial early warning system constructed by applying public data is characterized by comprising:
the searching and alarming module is used for monitoring internet information to obtain input data;
the warning sign identifying and analyzing module is used for extracting document data characteristics in the process of converting input data into warning signs so as to clearly identify event information related to early warning enterprises;
and the pre-alarming degree module is used for predicting the alarming degree information of the enterprise.
2. The financial early warning system constructed by public data as claimed in claim 1, wherein the internet information comprises corporate public data, financial reports, transactions, news, impurities, media, networks and public opinion public data.
3. The financial early warning system constructed by applying the public data as claimed in claim 1, wherein the alert level comprises two levels of alert and no alert.
4. A financial early warning method constructed by using public data is characterized in that the financial early warning system constructed by using the public data according to any one of claims 1 to 3 is implemented by the following steps:
step 1, monitoring internet information in real time to obtain input data;
step 2, document data characteristics of a process of changing input data into warning signals are extracted by using a BERT model so as to clarify event information related to early warning enterprises;
and 3, predicting the police degree information of the enterprise by using the LSTM and dividing different police degree grades.
5. The financial early warning method built by public data as claimed in claim 4, wherein in the step 1, the internet information comprises official business networks, financial and newspaper, transactions, news, impurities, media, network and public opinion public data.
6. The financial early warning method constructed by applying the public data as claimed in claim 4, wherein the specific process of the step 2 is as follows:
step 2.1, word embedding is carried out, and the expression of word embedding is as follows:
X＝Epos(onehot(S)+Eseg) (1)
in the formula (1), onehot represents word insertion, EsegRepresenting sentence embedding, EposIndicating position embedding;
step 2.2, defining a single attention machine mechanism:
in the formula (2), WQ、WK、WVIs a trainable parameter; softmax represents a softmax normalization function, mapping values to the range of 0-1; d represents the dimension of the vector;
step 2.3, combining a plurality of single attention mechanisms to form a multi-head attention mechanism layer, wherein the expression of the multi-head attention mechanism layer is as follows:
MultiHead(X)＝Concat(Att1，...，Atth)WO (3)
in the formula (3), Concate represents a link operation, h represents the number of attention heads, WORepresenting trainable parameters;
step 2.4, the output vector of the multi-head attention mechanism layer is added with the word in an embedding mode to form a residual error network, and the final output of the attention mechanism layer is obtained through batch normalization:
Matt＝BN(MultiHead(X)+X) (4)
in formula (4), BN represents batch normalization;
step 2.5, the final output obtained in step 2.4 passes through a full connection layer with a residual error network and batch normalization to obtain a document file characteristic B:
B＝BERT(X)＝BN(Dense(Matt)+Matt) (5)；
for a document S with n words, the character file characteristic B ═ B1，...，bn+1}，
Where d represents the dimension of the vector.
7. The financial early warning method constructed by applying the public data as claimed in claim 4, wherein the specific process of the step 3 is as follows:
step 3.1, assume the hidden layer output h of the last LSTMt-1And the tth document vector BtThen LSTM can be expressed as:
ft＝σ(Wf·Y+bf) (7)
it＝σ(Wi·Y+bi) (8)
ot＝σ(Wo·Y+bo) (9)
ct＝ft⊙ct-1+it⊙tanh(Wc·Y+bc) (10)
ht＝ot⊙tanh(ct) (11)
in the formula, Wi，Wf，Wo，Wc，bi，bf，bo，bcIs a trainable parameter, σ is a sigmod initiation function,. indicates a Hadamard product, ctRepresents the output of the t-th document;
step 3.2, vector h obtained by step 3.1tPredicting the alert level of the enterprise:
lt＝softmax(Dense(ht)) (12)。
8. the financial early warning method constructed by applying public data as claimed in claim 7, wherein when l is equal totA warning is indicated by more than or equal to 0.5, namely abnormal; when l ist< 0.5 indicates no alarm, i.e., normal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110271010.9A CN113177831B (en) | 2021-03-12 | Financial early warning system constructed by application of public data and early warning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110271010.9A CN113177831B (en) | 2021-03-12 | Financial early warning system constructed by application of public data and early warning method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113177831A true CN113177831A (en) | 2021-07-27 |
| CN113177831B CN113177831B (en) | 2024-05-17 |
Family
ID=
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111209401A (en) * | 2020-01-03 | 2020-05-29 | 西安电子科技大学 | System and method for classifying and processing sentiment polarity of online public opinion text information |
| CN111241300A (en) * | 2020-01-09 | 2020-06-05 | 中信银行股份有限公司 | Public opinion early warning and risk propagation analysis method, system, equipment and storage medium |
| CN111401061A (en) * | 2020-03-19 | 2020-07-10 | 昆明理工大学 | Method for identifying news opinion involved in case based on BERT and Bi L STM-Attention |
| CN111523301A (en) * | 2020-06-05 | 2020-08-11 | 泰康保险集团股份有限公司 | Contract document compliance checking method and device |
| CN111831895A (en) * | 2019-04-19 | 2020-10-27 | 天津科技大学 | Network public opinion early warning method based on LSTM model |
| CN112231472A (en) * | 2020-09-18 | 2021-01-15 | 昆明理工大学 | Judicial public opinion sensitive information identification method integrated with domain term dictionary |
| US20210034701A1 (en) * | 2019-07-30 | 2021-02-04 | Baidu Usa Llc | Coreference-aware representation learning for neural named entity recognition |
| CN112395482A (en) * | 2020-11-06 | 2021-02-23 | 北京工商大学 | Food safety incident early warning platform |
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111831895A (en) * | 2019-04-19 | 2020-10-27 | 天津科技大学 | Network public opinion early warning method based on LSTM model |
| US20210034701A1 (en) * | 2019-07-30 | 2021-02-04 | Baidu Usa Llc | Coreference-aware representation learning for neural named entity recognition |
| CN111209401A (en) * | 2020-01-03 | 2020-05-29 | 西安电子科技大学 | System and method for classifying and processing sentiment polarity of online public opinion text information |
| CN111241300A (en) * | 2020-01-09 | 2020-06-05 | 中信银行股份有限公司 | Public opinion early warning and risk propagation analysis method, system, equipment and storage medium |
| CN111401061A (en) * | 2020-03-19 | 2020-07-10 | 昆明理工大学 | Method for identifying news opinion involved in case based on BERT and Bi L STM-Attention |
| CN111523301A (en) * | 2020-06-05 | 2020-08-11 | 泰康保险集团股份有限公司 | Contract document compliance checking method and device |
| CN112231472A (en) * | 2020-09-18 | 2021-01-15 | 昆明理工大学 | Judicial public opinion sensitive information identification method integrated with domain term dictionary |
| CN112395482A (en) * | 2020-11-06 | 2021-02-23 | 北京工商大学 | Food safety incident early warning platform |
Non-Patent Citations (1)
| Title |
|---|
| 徐龙;: "基于自注意力和胶囊网络的短文本情感分析", 计算机与现代化, no. 07, 15 July 2020 (2020-07-15) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11556992B2 (en) | System and method for machine learning architecture for enterprise capitalization | |
| CN109657947B (en) | Enterprise industry classification-oriented anomaly detection method | |
| CN106611375A (en) | Text analysis-based credit risk assessment method and apparatus | |
| CN111950932A (en) | Multi-source information fusion-based comprehensive quality portrait method for small and medium-sized micro enterprises | |
| CN108108352A (en) | A kind of enterprise's complaint risk method for early warning based on machine learning Text Mining Technology | |
| Minhas et al. | From spin to swindle: Identifying falsification in financial text | |
| CN110502638B (en) | Enterprise news risk classification method based on target entity | |
| CN111597340A (en) | Text classification method and device and readable storage medium | |
| CN113095927A (en) | Method and device for identifying suspicious transactions of anti-money laundering | |
| CN116383399A (en) | Event public opinion risk prediction method and system | |
| CN110782349A (en) | Model training method and system | |
| Mishev et al. | Forecasting corporate revenue by using deep-learning methodologies | |
| CN117556225B (en) | Pedestrian credit data risk management system | |
| CA2607477A1 (en) | Test mining systems and methods for early detection and warning | |
| CN112632964B (en) | NLP-based industry policy information processing method, device, equipment and medium | |
| Fieberg et al. | Machine learning in accounting research | |
| Feng et al. | A representation learning framework for stock movement prediction | |
| Anese et al. | Impact of public news sentiment on stock market index return and volatility | |
| Ardyanta et al. | A prediction of stock price movements using support vector machines in Indonesia | |
| CN113177831A (en) | Financial early warning system and method constructed by applying public data | |
| CN113222471B (en) | Asset wind control method and device based on new media data | |
| CN113177831B (en) | Financial early warning system constructed by application of public data and early warning method | |
| CN114741501A (en) | Public opinion early warning method and device, readable storage medium and electronic equipment | |
| Tornés et al. | Detecting forged receipts with domain-specific ontology-based entities & relations | |
| Wang et al. | Preprocessing and feature extraction methods for microfinance overdue data |
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 | ||
| GR01 | Patent grant |