CN113487425A - Method and system for backtracking daytime liquidity condition based on historical data - Google Patents
Method and system for backtracking daytime liquidity condition based on historical data Download PDFInfo
- Publication number
- CN113487425A CN113487425A CN202110887690.7A CN202110887690A CN113487425A CN 113487425 A CN113487425 A CN 113487425A CN 202110887690 A CN202110887690 A CN 202110887690A CN 113487425 A CN113487425 A CN 113487425A
- Authority
- CN
- China
- Prior art keywords
- backtracking
- daytime
- liquidity
- historical
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- 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
- G06Q40/02—Banking, e.g. interest calculation or account maintenance
Abstract
The invention provides a daytime mobility condition backtracking method and system based on historical data. The method comprises the following steps: acquiring historical transaction data; preprocessing historical transaction data; extracting the characteristics of historical transaction data; training a backtracking prediction model by using the characteristics; and predicting the daytime liquidity condition by using the trained backtracking prediction model. The daytime liquidity condition backtracking method and system based on historical data provided by the invention realize backtracking analysis of daytime liquidity conditions of commercial banks, so that the daytime liquidity conditions are analyzed, dependence on manual experience is reduced, and liquidity analysis is carried out more reasonably.
Description
Technical Field
The invention relates to the technical field of financial science and technology, in particular to a daytime mobility condition backtracking method and system based on historical data.
Background
Requirements of the "commercial bank liquidity risk management method": commercial banks should enforce day liquidity risk management, ensure that there are sufficient day liquidity positions and related financing arrangements, meet day payment requirements in time under normal and stressful scenarios, and make new requirements for liquidity prediction: the effective amount is the predicted total amount of cash inflow and outflow per day, and the scale and gap of cash inflow and outflow at each time point during the day. And it is clear that commercial banks should perform backtracking analysis on daytime liquidity conditions by combining historical data. The patent is a backtracking analysis method for realizing day mobility conditions according to an algorithm based on historical data in order to meet the supervision requirements.
The method adopted by commercial banks before this is that the bank loan deposit change analysis method is adopted, the traditional analysis of the day liquidity condition of the bank is actually the analysis of the liquidity and the required amount of the fund, the practical condition of the fund of the bank is in change every day, the fund change depends on the change of the loan deposit fund amount of the bank, the deposit absorption can cause the increase of the bank deposit fund, the loan is added, and the shortage of the deposit fund can occur. Therefore, the liquidity condition is traditionally analyzed through the change trend of the bank deposit, the deposit is a passive debt of the bank, and the initiative of deposit change is more mastered in the client audience, but is still regularly followed. On the other hand, the change of loan demand is different from the fluctuation of deposit, the initiative of the loan is held in the bank, but the bank can only issue a new loan if the liquidity is sufficient. Usually, the balance of liquidity is calculated by calculating daily total deposit amount, deposit fluctuation amount, total loan amount, and loan fluctuation amount, and the liquidity is calculated by calculating the change in liquidity from the change in deposit amount to the change in loan amount, thereby determining whether the liquidity is sufficient.
The existing method for analyzing the liquidity condition based on loan-saving change has the following defects:
the uncertainty of advance payment of deposit, advance repayment of loan and demand deposit cannot be predicted, the demand deposit and the like mainly depend on customer behaviors, and although calculation can be performed through indexes such as demand deposit settlement rate and the like, the indexes are usually evaluated according to manual experience, and due to the difference of experience values, the uncertainty is large, and the prediction accuracy is low.
And secondly, the flow and fund amount of payment transactions of each payment channel can not be predicted in a refined mode, for example, the flow condition of specific time points every day is backtracked and analyzed.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a daytime liquidity condition backtracking method and system based on historical data, so that the daytime liquidity condition of a commercial bank can be analyzed by backtracking analysis, dependence on manual experience is reduced, and liquidity analysis is more reasonably performed.
In order to solve the technical problem, the invention provides a daytime mobility condition backtracking method based on historical data, which comprises the following steps: acquiring historical transaction data; preprocessing historical transaction data; extracting the characteristics of historical transaction data; training a backtracking prediction model by using the characteristics; and predicting the daytime liquidity condition by using the trained backtracking prediction model.
In some embodiments, the pre-processing of the historical transaction data includes: the historical transaction data is categorized according to payment channel, business type, institution, etc.
In some embodiments, pre-processing the historical transaction data further comprises: missing data in the historical transaction data is filled.
In some embodiments, the backtracking prediction model comprises: an ARIMA model, an XGboost model and an LSTM model which are respectively trained.
In some embodiments, training the backtracking prediction model with features includes: respectively training an ARIMA model, an XGboost model and an LSTM model by utilizing training data; performing linear fusion on the ARIMA model, the XGboost model and the LSTM model according to respective fitting effects; and other time series models.
In some embodiments, training the backtracking prediction model with features includes: and calculating the data mean value of the same moment within five days, seven days and fourteen days, dividing the time into five types in a month and a daytime type, and performing single hot coding to obtain the characteristics of prediction.
In some embodiments, five types in the month include: the first 6 days, 6-12 days, 12-18 days, 18-24 days, 24-30 days.
In some embodiments, the daytime types include: 6-8 points, 8-10 points, 10-12 points, 12-14 points, 14-16 points and 16-18 points.
In some embodiments, further comprising: and before the trained backtracking prediction model is used for predicting the day mobility condition, effect evaluation is carried out on the backtracking prediction model.
In addition, the invention also provides a daytime mobility condition backtracking system based on historical data, which comprises: one or more processors; storage means for storing one or more programs; when executed by the one or more processors, cause the one or more processors to implement a method for daytime liquidity condition backtracking based on historical data as described above.
After adopting such design, the invention has at least the following advantages:
the invention fully satisfies the 'commercial bank liquidity risk management method', the management method requires that the commercial bank should carry out backtracking analysis on the daytime liquidity condition by combining historical data, effectively measures the expected cash inflow total amount and outflow total amount every day, and scales and gaps of the cash inflow and outflow at each time point every day;
the method utilizes an analysis algorithm model and combines the historical transaction detail data analysis of the commercial bank, thereby effectively predicting the fund and gap condition at each time point in the day and analyzing the fund gap of each payment channel and each fund clearinghouse;
the method can promote the accuracy and refinement of liquidity analysis and prediction, and through the analysis and prediction of historical data, the liquidity scale can be predicted with high accuracy, the refined management of liquidity is promoted, and the defects of traditional loan storage change calculation and manual experience prediction are overcome.
Drawings
The foregoing is only an overview of the technical solutions of the present invention, and in order to make the technical solutions of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and the detailed description.
Fig. 1 is a flow chart of a daytime liquidity condition backtracking method based on historical data.
Detailed Description
The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, and it will be understood that they are described herein for the purpose of illustration and explanation and not limitation.
The method is characterized in that backtracking analysis of daytime liquidity conditions of commercial banks is realized by combining the basic transaction data and historical transaction data with algorithms such as ARIMA, XGboost, LSTM and the like, so that the daytime liquidity conditions are analyzed, dependence on manual experience is reduced, and liquidity analysis is carried out more reasonably.
The working process is as follows:
historical transaction data processing
Classifying data according to payment channel
Populating historical missing data
Performing characteristic engineering: and calculating the average value of data at the same moment in five days, seven days and fourteen days, dividing the time into five types in a month (the first 6 days, 6-12 days, … and the like), and using the daytime data (6-8 points, 8-10 points, … and the like) as the predicted characteristics after carrying out unique hot coding.
Three models of daily fund inflow and outflow training are predicted according to the history of each type of data respectively
ARIMA model: the traditional method for constructing the time sequence model has a good fitting effect on a sequence showing a dependency relationship on time, such as obvious trend periodicity and the like.
XGboost: an algorithm from a gradient boosting decision tree variation can add more features to the model than a conventional time series model, thereby improving the effect of the fit. A time step, denoted as k, needs to be set in the process of implementing this algorithm. The features of the first k days are all taken as the features of the (k + 1) th day to be predicted.
LSTM: an improvement of a Recurrent Neural Network (RNN) solves data with long-term memory dependence, and the selection characteristic is consistent with XGboost.
Model fusion and effect assessment: and calculating evaluation indexes (mse, mae and the like) of the three models on a training set, drawing a visual image to evaluate the effect of the models, performing linear fusion on the models according to the fitting effect of the models, and storing the models with good effect as a json format to the local.
Characterizing predictive data
Analysis was performed using the final model above and the resulting data was saved.
Historical data of a commercial bank is utilized, retrospective analysis is carried out on the historical data through an analysis algorithm model (XGboost, LSTM), and the characteristic of one-hot coding of mean value and date is added at the same time, so that the model can better capture the time dependence of the data, and the fitting of the model is more accurate.
The invention fully satisfies the liquidity risk management method of the commercial bank, and the management method requires that the commercial bank should carry out backtracking analysis on the liquidity condition in the day by combining historical data, effectively measures the expected inflow total amount and outflow total amount of cash in each day, and scales and gaps of the inflow and outflow of the cash at each time point in the day.
The invention uses the analysis algorithm model and combines the historical transaction detail data analysis of the commercial bank, thereby effectively predicting the fund and gap condition at each time point during the day and analyzing the fund gap of each payment channel and each clearing client.
The method can promote the accuracy and refinement of liquidity analysis and prediction, and through the analysis and prediction of historical data, the liquidity scale can be predicted with high accuracy, the refined management of liquidity is promoted, and the defects of traditional loan storage change calculation and manual experience prediction are overcome.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the present invention in any way, and it will be apparent to those skilled in the art that the above description of the present invention can be applied to various modifications, equivalent variations or modifications without departing from the spirit and scope of the present invention.
Claims (10)
1. A daytime liquidity condition backtracking method based on historical data is characterized by comprising the following steps:
acquiring historical transaction data;
preprocessing historical transaction data;
extracting the characteristics of historical transaction data;
training a backtracking prediction model by using the characteristics;
and predicting the daytime liquidity condition by using the trained backtracking prediction model.
2. The historical data-based daytime liquidity condition backtracking method according to claim 1, wherein the preprocessing of the historical transaction data comprises:
according to the payment channel, the historical transactions are divided according to the channel.
3. The method for backtracking daytime liquidity conditions based on historical data according to claim 2, wherein the preprocessing is performed on the historical transaction data, and further comprising:
missing data in the historical transaction data is filled.
4. The historical data-based daytime liquidity condition backtracking method according to claim 1, wherein the backtracking prediction model comprises: an ARIMA model, an XGboost model and an LSTM model which are respectively trained.
5. The historical data-based daytime mobility condition backtracking method according to claim 4, wherein the training of the backtracking prediction model by using features comprises:
respectively training an ARIMA model, an XGboost model and an LSTM model by utilizing training data;
and performing linear fusion on the ARIMA model, the XGboost model and the LSTM model according to respective fitting effects.
6. The method for backtracking daytime liquidity conditions based on historical data according to claim 1, wherein training a backtracking prediction model by using features comprises:
and calculating the data mean value of the same moment within five days, seven days and fourteen days, dividing the time into five types in a month and a daytime type, and performing single hot coding to obtain the characteristics of prediction.
7. The daytime liquidity condition backtracking method based on historical data according to claim 6, wherein five types in a month comprise: the first 6 days, 6-12 days, 12-18 days, 18-24 days, 24-30 days.
8. The daytime liquidity condition backtracking method based on historical data according to claim 6, wherein the daytime types comprise: 6-8 points, 8-10 points, 10-12 points, 12-14 points, 14-16 points and 16-18 points.
9. The method for backtracking daytime liquidity conditions based on historical data according to claim 1, further comprising:
and before the trained backtracking prediction model is used for predicting the day mobility condition, effect evaluation is carried out on the backtracking prediction model.
10. A system for backtracking daytime liquidity conditions based on historical data, comprising:
one or more processors;
storage means for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement the method for historical data-based daytime liquidity condition backtracking according to any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110887690.7A CN113487425A (en) | 2021-08-03 | 2021-08-03 | Method and system for backtracking daytime liquidity condition based on historical data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110887690.7A CN113487425A (en) | 2021-08-03 | 2021-08-03 | Method and system for backtracking daytime liquidity condition based on historical data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113487425A true CN113487425A (en) | 2021-10-08 |
Family
ID=77944480
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110887690.7A Pending CN113487425A (en) | 2021-08-03 | 2021-08-03 | Method and system for backtracking daytime liquidity condition based on historical data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113487425A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107977754A (en) * | 2017-12-18 | 2018-05-01 | 深圳前海微众银行股份有限公司 | Data predication method, system and computer-readable recording medium |
| CN110363661A (en) * | 2019-08-05 | 2019-10-22 | 中国工商银行股份有限公司 | Bank liquidity prediction technique and device |
| CN110390425A (en) * | 2019-06-20 | 2019-10-29 | 阿里巴巴集团控股有限公司 | Prediction technique and device |
| CN112787878A (en) * | 2019-11-08 | 2021-05-11 | 大唐移动通信设备有限公司 | Network index prediction method and electronic equipment |
-
2021
- 2021-08-03 CN CN202110887690.7A patent/CN113487425A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107977754A (en) * | 2017-12-18 | 2018-05-01 | 深圳前海微众银行股份有限公司 | Data predication method, system and computer-readable recording medium |
| CN110390425A (en) * | 2019-06-20 | 2019-10-29 | 阿里巴巴集团控股有限公司 | Prediction technique and device |
| CN110363661A (en) * | 2019-08-05 | 2019-10-22 | 中国工商银行股份有限公司 | Bank liquidity prediction technique and device |
| CN112787878A (en) * | 2019-11-08 | 2021-05-11 | 大唐移动通信设备有限公司 | Network index prediction method and electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 胡盼盼: "《自然语言处理从入门到实战》", 北京：中国铁道出版社, pages: 154 - 155 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110400022B (en) | Cash consumption prediction method and device for self-service teller machine | |
| Gilchrist et al. | Investment: fundamentals and finance | |
| CN110400021B (en) | Bank branch cash usage prediction method and device | |
| Blanco et al. | Credit scoring models for the microfinance industry using neural networks: Evidence from Peru | |
| Mare | Contribution of macroeconomic factors to the prediction of small bank failures | |
| KR102009309B1 (en) | Management automation system for financial products and management automation method using the same | |
| CN105719013A (en) | Method and system for calculating enterprise asset debts | |
| Li et al. | The stock–bond comovements and cross-market trading | |
| Li et al. | Take Bitcoin into your portfolio: a novel ensemble portfolio optimization framework for broad commodity assets | |
| Scheller et al. | How does the choice of Value-at-Risk estimator influence asset allocation decisions? | |
| Antunes et al. | Cost efficiency of Chinese banks: Evidence from DEA and MLP-SSRP analysis | |
| CN112766814A (en) | Training method, device and equipment for credit risk pressure test model | |
| CN112862182A (en) | Investment prediction method and device, electronic equipment and storage medium | |
| CN116542760A (en) | Method and device for evaluating data | |
| Hirk et al. | A corporate credit rating model with autoregressive errors | |
| CN111899093B (en) | Method and device for predicting default loss rate | |
| CN113487425A (en) | Method and system for backtracking daytime liquidity condition based on historical data | |
| Gogo et al. | Analysis of Factors Affecting Liquidity Risk in Indonesian Islamic Banking | |
| CN104636613A (en) | Management right value assessment system for tourist attraction | |
| Talebi et al. | The Evaluation of the Capability of the Regression & Neural Network Models in Predicting Future Cash Flows | |
| US20200051175A1 (en) | Method and System for Predicting and Indexing Probability of Financial Stress | |
| Theophilopoulou | The impact of macroeconomic uncertainty on inequality: An empirical study for the UK | |
| Stundziene | Prediction of Lithuanian GDP: Are regression models or time series models better? | |
| Nie | Computer Intelligent Value Evaluation Model through ARMA and Long Short-Term Memory Neural Network | |
| Affinito | What do almost 20 years of micro data and two crises say about the relationship between central bank and interbank market liquidity? Evidence from Italy |
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 |