CN114154696A - Method, system, computer device and storage medium for predicting fund flow - Google Patents
Method, system, computer device and storage medium for predicting fund flow Download PDFInfo
- Publication number
- CN114154696A CN114154696A CN202111400988.7A CN202111400988A CN114154696A CN 114154696 A CN114154696 A CN 114154696A CN 202111400988 A CN202111400988 A CN 202111400988A CN 114154696 A CN114154696 A CN 114154696A
- Authority
- CN
- China
- Prior art keywords
- fund
- data
- model
- prediction
- daily
- 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
- 238000000034 method Methods 0.000 title claims abstract description 72
- 238000012549 training Methods 0.000 claims abstract description 79
- 230000004927 fusion Effects 0.000 claims abstract description 45
- 238000012545 processing Methods 0.000 claims abstract description 25
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 23
- 238000006073 displacement reaction Methods 0.000 claims abstract description 22
- 238000006243 chemical reaction Methods 0.000 claims abstract description 18
- 230000002354 daily effect Effects 0.000 claims description 105
- 238000012360 testing method Methods 0.000 claims description 27
- 238000004422 calculation algorithm Methods 0.000 claims description 17
- 238000004364 calculation method Methods 0.000 claims description 17
- 238000004458 analytical method Methods 0.000 claims description 15
- 230000000694 effects Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 12
- 238000011156 evaluation Methods 0.000 claims description 10
- 230000003203 everyday effect Effects 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 8
- 230000006399 behavior Effects 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 6
- 238000007405 data analysis Methods 0.000 claims description 4
- 230000001131 transforming effect Effects 0.000 claims description 2
- 238000010801 machine learning Methods 0.000 abstract description 9
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 8
- 230000008859 change Effects 0.000 description 6
- 238000007726 management method Methods 0.000 description 6
- 238000003066 decision tree Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000036962 time dependent Effects 0.000 description 4
- 238000012300 Sequence Analysis Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 230000000750 progressive effect Effects 0.000 description 2
- 238000000611 regression analysis Methods 0.000 description 2
- 238000000638 solvent extraction Methods 0.000 description 2
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000004941 influx Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000036961 partial effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 238000012795 verification 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06N20/00—Machine learning
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Technology Law (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of big data processing, and discloses a fund flow prediction method, a system, computer equipment and a storage medium, wherein the fund flow prediction method comprises the steps of carrying out time sequence decomposition on historical fund flow data to obtain time sequence data; performing characteristic displacement conversion on the time series data to obtain training data; constructing a fund inflow prediction model and a fund outflow prediction model; training the fund inflow prediction model and the fund outflow prediction model by using the training data respectively; and carrying out weighted fusion on the output of the fund inflow prediction model and the output of the fund outflow prediction model to obtain a prediction result. The fund flow prediction method integrates the time series model and the machine learning regression model, not only fully utilizes the time series model to mine important time-related characteristics, but also explores and summarizes important influence factors through multi-party data. The fund flow prediction method provided by the disclosure has richer input characteristics and knowledge learned by a model, and greatly improves the accuracy of prediction.
Description
Technical Field
The invention relates to the technical field of big data processing, in particular to a fund flow prediction method, a system, computer equipment and a storage medium.
Background
With the popularization and growth of small and micro businesses, a large amount of capital expenditure and repayment are involved every day, and the consumer loan products which are released and returned along with borrowing for small and micro enterprises have a large user group, so that the capital management of the consumer loan products becomes a very troublesome problem. The method not only ensures that the total fund is positive but also obtains the maximum profit as far as possible, so that the fund mobility risk is reduced to the minimum under the condition of ensuring normal operation of daily business, and the accurate prediction of the fund inflow and fund outflow condition becomes important.
Disclosure of Invention
Based on this, it is necessary to provide a fund flow prediction method, system, computer device and storage medium for accurately predicting the inflow and outflow of funds.
A fund flow prediction method comprises the steps of carrying out time sequence decomposition on historical fund flow data to obtain time sequence data; performing characteristic displacement transformation on the time sequence data to obtain training data, wherein the characteristic displacement transformation comprises the step of transforming the time sequence data into data meeting the format requirement of input data of a selected prediction model; constructing a capital inflow forecasting model with supervised learning and a capital outflow forecasting model with supervised learning; training the fund inflow prediction model and the fund outflow prediction model respectively by using the training data; and performing weighted fusion on the output of the fund inflow prediction model and the output of the fund outflow prediction model to obtain a prediction result.
In one embodiment, the time-series decomposing the historical fund flow data, and the obtaining the time-series data comprises performing time-series decomposing on the historical fund flow data by using a Prophet time-series tool to obtain the time-series data, wherein the time-series data comprises a trend item, a period item, an activity effect item and an error item.
In one embodiment, before time-series decomposition is performed on the historical fund flow data to obtain time-series data, the method further comprises performing mobility analysis and special date behavior analysis on the historical fund flow data to determine time-related variables.
In one embodiment, after performing feature displacement transformation on the time series data and obtaining training data, the method further includes dividing the training data into training set data and test set data.
In one embodiment, the training set data is used to train the fund-inflow prediction model and the fund-outflow prediction model using a LightGBM regression algorithm, respectively.
In one embodiment, after the training of the funds-in and funds-out prediction models using the training data, respectively, the method further comprises evaluating prediction accuracy of the funds-in and funds-out prediction models using a fused model error value of the funds-in and funds-out prediction models; the fusion model error value is determined in the following manner: the fund inflow prediction model predicts a user daily fund inflow prediction value according to the test set data, and the fund outflow prediction model predicts a user daily fund outflow prediction value according to the test set data; calculating a daily loan repayment relative error according to the daily fund inflow predicted value and the actual fund inflow value of the user, and calculating a daily loan payment relative error according to the daily fund outflow predicted value and the actual fund outflow value of the user; calculating the error value of the fusion model according to the relative error of the repayment of the daily loan and the relative error of the daily loan payment;
the calculation mode of the relative error for the daily loan payment comprises the following steps:
wherein useamt is the relative error of loan branch per day, ziFor the user's daily fundsThe actual out-flow value is,
a daily fund flow prediction value for the user;
the calculation mode of the relative error of the daily loan repayment comprises the following steps:
wherein repayamt is the relative error of loan repayment every day, yiThe actual daily inflow of funds for the user,
a daily fund inflow prediction value for the user;
the calculation method of the fusion model error value comprises the following steps:
totalscore＝score(useamt)*α+score(repayamt)*(1-α)
wherein totalscore is the error value of the fusion model, score () is a mapping function of the relative error and the final score, α is a weight coefficient for adjusting the two models, the value range of α is [0,1], useamt is the relative error for daily loan payment, and repayamt is the relative error for daily loan repayment.
A fund flow prediction system comprises a data processing module, wherein the data processing module comprises a time sequence decomposition unit and a characteristic displacement conversion unit, and the time sequence decomposition unit is used for performing time sequence decomposition on historical fund flow data to obtain time sequence data; the characteristic displacement conversion unit is used for performing characteristic displacement conversion on the time sequence data to acquire training data; the model building module comprises a model processing unit and a model training unit, wherein the model processing unit is used for building a capital inflow prediction model with supervised learning and a capital outflow prediction model with supervised learning; the model training unit is used for respectively training the fund inflow prediction model and the fund outflow prediction model by utilizing the training data; and the result prediction module is used for performing weighted fusion on the output of the fund inflow prediction model and the output of the fund outflow prediction model to obtain a prediction result.
In one embodiment, the data processing module further includes a data analysis unit, configured to perform mobility analysis and special date behavior analysis on the historical fund flow data, and determine a time-dependent influence factor.
In one embodiment, the data processing module further comprises a data set dividing unit for dividing the training data into training set data and test set data.
In one embodiment, the model training unit utilizes the training set data to train the fund-inflow prediction model and the fund-outflow prediction model, respectively, using a LightGBM regression algorithm.
In one embodiment, the model building module further comprises a model evaluation unit for evaluating the prediction accuracy of the fund inflow prediction model and the fund outflow prediction model by using a fused model error value of the fund inflow prediction model and the fund outflow prediction model; the fusion model error value is determined in the following manner: the fund inflow prediction model predicts a user daily fund inflow prediction value according to the test set data, and the fund outflow prediction model predicts a user daily fund outflow prediction value according to the test set data; the model evaluation unit calculates a daily loan repayment relative error according to the daily fund inflow predicted value and the actual fund inflow value of the user, and calculates a daily loan payment relative error according to the daily fund outflow predicted value and the actual fund outflow value of the user; the model evaluation unit calculates the fusion model error value according to the relative error of the repayment of the daily loan and the relative error of the daily loan payment;
the calculation mode of the relative error for the daily loan payment comprises the following steps:
wherein useamt is the relative error of loan branch per day, ziThe actual daily value of the fund flow for the user,
a daily fund flow prediction value for the user;
the calculation mode of the relative error of the daily loan repayment comprises the following steps:
wherein repayamt is the relative error of loan repayment every day, yiThe actual daily inflow of funds for the user,
a daily fund inflow prediction value for the user;
the calculation method of the fusion model error value comprises the following steps:
totalscore＝score(useamt)*α+score(repayamt)*(1-α)
wherein totalscore is the error value of the fusion model, score () is a mapping function of the relative error and the final score, α is a weight coefficient for adjusting the two models, the value range of α is [0,1], useamt is the relative error for daily loan payment, and repayamt is the relative error for daily loan repayment.
A computer device comprising a memory storing a computer program and a processor implementing the steps of the method of fund flow prediction according to any one of the above embodiments when the processor executes the computer program.
A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of fund flow prediction according to any one of the preceding embodiments.
A computer program product comprising a computer program which, when executed by a processor, carries out the steps of the method of fund flow prediction according to any one of the preceding embodiments.
According to the fund flow prediction method, time sequence decomposition is carried out on historical fund flow data, and the time sequence data are converted into data meeting the requirement of a selected prediction model input data format and used as training data. And training the fund inflow prediction model and the fund outflow prediction model by utilizing the training data, and performing weighted fusion on the output of the fund inflow prediction model and the output of the fund outflow prediction model to obtain a final prediction result. The fund flow prediction method provided by the disclosure adopts a fusion mode of combining a time series model and a machine learning regression model, not only fully utilizes the time series model to mine important time-related characteristics, but also explores and summarizes important influence factors through multi-party data. Compared with the traditional time series algorithm which only considers a single time variable, the input characteristics and the knowledge learned by the model are richer, and therefore the accuracy of prediction is greatly improved.
Drawings
In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the specification, and other drawings can be obtained by those skilled in the art without inventive labor.
Fig. 1 is a schematic flow chart of a method for predicting a fund flow in a first embodiment of the present disclosure;
FIG. 2 is a schematic flow chart illustrating a method for evaluating the fund inflow prediction model and the fund outflow prediction model according to a second embodiment of the present disclosure;
fig. 3 is a block diagram of a fund flow prediction system in a third embodiment of the present disclosure;
fig. 4 is a block diagram of a fund flow prediction apparatus or system according to a fourth embodiment of the present disclosure.
Detailed Description
To facilitate an understanding of the invention, the invention will now be described more fully with reference to the accompanying drawings. Preferred embodiments of the present invention are shown in the drawings. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.
It will be understood that when an element is referred to as being "secured to" another element, it can be directly on the other element or intervening elements may also be present. When an element is referred to as being "connected" to another element, it can be directly connected to the other element or intervening elements may also be present. As used herein, the terms "vertical," "horizontal," "left," "right," "upper," "lower," "front," "rear," "circumferential," and the like are based on the orientation or positional relationship shown in the drawings for ease of description and simplicity of description, and do not indicate or imply that the referenced device or element must have a particular orientation, be constructed and operated in a particular orientation, and are therefore not to be considered limiting of the present invention.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
With the popularization and growth of small and micro businesses, a large amount of capital expenditure and repayment are involved every day, and the consumer loan products which are released and returned along with borrowing for small and micro enterprises have a large user group, so that the capital management of the consumer loan products becomes a very troublesome problem. The fund management of consumer loan products is to ensure that the total fund is positive and the maximum profit is obtained as much as possible, so that the fund management work needs to reduce the fund mobility risk to the minimum under the condition of ensuring the normal operation of daily business, and the accurate prediction of the fund inflow and outflow conditions becomes important. For a loan scenario, the influx of funds means the act of repayment and the efflux of funds means the act of loan payment. If daily fund flow conditions can be predicted relatively accurately, planning investment and accurate management can be carried out on available funds under the condition of ensuring safety, so that greater benefits are obtained.
Fig. 1 is a schematic flow chart of a method for predicting a fund flow in a first embodiment of the present disclosure, wherein in one embodiment, the method for predicting a fund flow may include the following steps S100 to S500.
Step S100: and performing time series decomposition on the historical fund flow data to obtain time series data.
And selecting proper data from the database as historical fund flow data for exploration and analysis according to actual model requirements and data conditions. The traditional machine learning regression prediction ignores the influence of time and trend on data development, and cannot analyze the correlation between time series of different dependent variable values and adjacent values, so that a part of information is lost. The method combines two methods of regression analysis and time sequence analysis, time sequence analysis is carried out on historical fund flow data to obtain time sequence data, hidden correlation behind data development can be carried out based on time sequence deep mining time, model fusion is carried out, and the defect that a regression prediction algorithm cannot capture important features in a time dimension is overcome.
Step S200: and performing characteristic displacement conversion on the time series data to obtain training data, wherein the characteristic displacement conversion comprises the step of converting the time series data into data meeting the requirement of the input data format of the selected prediction model.
In some embodiments of the present disclosure, machine learning regression methods may be used to predict the flow of funds, requiring accurate assurance of important features as model inputs. The present disclosure utilizes supervised learning prediction models to build both the funds-in and funds-out prediction models. The time series data is formatted to ensure that the time series data can be input to the selected predictive model. Since the fund flow prediction method is based on time series prediction and the machine learning method is a supervised learning method, the time series is converted into a format which can be used for supervised learning. By performing feature displacement conversion on the event sequence data, features can be converted into training data in a supervised learning format.
Step S300: and constructing a capital inflow forecasting model with supervised learning and a capital outflow forecasting model with supervised learning.
Since for the loan scenario, the flow of funds includes both the inflow of funds, meaning the repayment act, and the outflow of funds, meaning the loan payment act. Therefore, it is necessary to respectively establish prediction models for the situations of capital inflow and capital outflow, respectively predict the situations of capital inflow and capital outflow by using the capital inflow prediction model with supervised learning and the capital outflow prediction model with supervised learning, and synthesize the two prediction results to obtain the final capital flow prediction result.
Step S400: and training the fund inflow prediction model and the fund outflow prediction model respectively by using the training data.
And respectively carrying out training optimization on the fund inflow prediction model and the fund outflow prediction model by using the training data. And adjusting parameters of the fund inflow prediction model and the fund outflow prediction model according to the model prediction result by converting the characteristics into a supervised learning format and then using the supervised learning format as a model input variable. In practical application, the parameter adjustment can focus on the number of the base learners, the learning rate, the maximum depth of the tree, the regularization parameter and the like.
Step S500: and carrying out weighted fusion on the output of the fund inflow prediction model and the output of the fund outflow prediction model to obtain a prediction result.
And the fund inflow prediction model and the fund outflow prediction model respectively predict the fund inflow condition and the fund outflow condition, and perform weighted fusion on the two prediction outputs to obtain a final fund flow prediction result. In some embodiments of the present disclosure, the effect evaluation may be performed on the predicted outputs of the two models using a single point relative error addition method.
The fund flow prediction method provided by the disclosure combines two methods of regression analysis and time sequence analysis, deeply excavates hidden associations behind historical fund flow data, constructs a fund inflow prediction model and a fund outflow prediction model, and performs model fusion on the fund inflow prediction model and the fund outflow prediction model, thereby making up the defect that a regression prediction algorithm cannot capture important features in a time dimension. The method for predicting the fund flow considers the influence of the time sequence on the result during prediction, can analyze the correlation between the time sequence with different dependent variable values and the adjacent values, and can consider the influence of mutation factors such as holidays, marketing days or other important dates on the fund flow, so that the strain capacity and the robustness of the model are stronger. Compared with the traditional time series algorithm only considering a single time variable, the fund flow prediction method provided by the disclosure has richer input characteristics and learned knowledge of a model, and the prediction accuracy is greatly improved.
In one embodiment, before time-series decomposition is performed on the historical fund flow data to obtain time-series data, the fund flow prediction method further comprises performing mobility analysis and special date behavior analysis on the historical fund flow data to determine time-related influence factors. During data processing, time-series decomposition can be carried out on the historical fund flow data according to the determined time-related influence factors.
In the embodiment, the fund flow data of a certain consumption loan product in 3 years of 2018-2020 is selected for data exploration and analysis according to actual model requirements and data conditions. Since the prediction methods of the fund inflow and the fund outflow are similar, the fund outflow prediction scheme is taken as an example below.
In 2018 and 2020, the overall support of the consumer loan product is wholly represented by the trend that the fluctuation range at the end of the month is enhanced and the trend that the consumption load is greatly increased at the end of the quarter is increased. The overall supporting and using rule in 2018 and 2019 is consistent with the fluctuation rule along with time, the supporting and using rule is influenced by the ineffectiveness factor in 2020, the supporting and using rule is changed, but the trend of fluctuation according to the week is still followed, and the supporting and using rule in working days is more.
At present, seven legal holidays including New year, spring festival, Qingming, Wuyi labor festival, Dragon boat festival, mid autumn and national celebration are mainly arranged in China, and the total loan amount of 15 working days before the seven legal holidays is analyzed as follows: in 2019 and 2020, the total sum of loan payment before vacation of Yuan Dan is gradually increased, the consumption of loan payment one day before Yuan Dan in 2020 is greatly increased compared with the consumption of loan payment in the same period of the last year, and the increase of the total sum of loan payment in 2021 is smaller due to the influence of force-inequality factors. The total amount of money for consumption continuously floats 4-15 working days before the spring festival, and gradually decreases in the first 1-3 working days. The total amount of money spent presents a small peak 3-6 working days before the Qingming festival is vacated. The total sum of the money paid in the first 5 working days of the five-day work section is in an ascending trend. The total amount of money for payment is more in 3-7 working days before the holiday in mid-autumn festival, and the peak value is reached in the first 4-5 working days. The closer to the national festival in 2018 and 2019, the larger the total sum of the money is, but the total sum of the money before the holiday in the national festival in 2020 has no obvious change.
In one embodiment, time-series decomposing the historical fund flow data, and obtaining the time-series data comprises performing time-series decomposing on the historical fund flow data according to the influence factors by using a Prophet time-series tool to obtain the time-series data. The time series data may include a trend term, a period term, an activity effect term, and an error term.
Prophet is a facebook open-source time series prediction tool, is a time prediction tool based on an additive model, and can decompose a time series into a trend term, a period term, an activity effect term and an error term. In some embodiments of the present disclosure, the trend term may be used to represent aperiodic changes in the time series, the period term may be used to represent periodic changes in the time series, the campaign effect term may express some abnormal campaigns in the time series, such as holidays, marketing days, or other important dates, etc., and the error term may be used to represent abnormal errors that cannot be described by the model. Prophet can fit non-linear trends of year, week, season, and holiday. Prophet is very robust against missing values, definite transitions and a large number of outliers.
The model formula for Prophet is: y (t) ═ g (t) + s (t) + h (t) + e;
wherein g (t) represents a trend term, s (t) represents a period term, h (t) represents an activity effect term, and e represents an error term.
When time-series decomposition is performed on historical fund flow data by using Prophet, the dimension of decomposition can be selected by referring to the determined influence factors. In the embodiment, the fund flow data of the consumer loan products in the 2 years of 2018 and 2019 is selected for exploration analysis according to actual model requirements and data conditions. The following also takes the fund flow prediction scheme as an example.
The data sequence of the total daily expenditure in 2018 and 2019 is decomposed mainly from several dimensions of overall trend, change of holidays and festivals, week and month. From the overall trend, the total expenditure and utilization is in an upward trend as a whole; from the change of holidays, the total expenditure has a plurality of mutation points on the holidays; from the change of week, the total amount of the branch is less on weekends and is more stable on workdays; from the change of the month, the total payment amount has a sudden change trend at the end of each month, and the payment amount is more.
By combining the data analysis and exploration results and the time series decomposition results, the following conclusion can be drawn that the loan expenditure of the consumer loan products shows obvious periodic trends of weeks, months, seasons and the like, and important periods such as holiday marketing days and the like also have important influence on the loan expenditure.
And respectively processing according to detailed information of dates, and excavating potential influence factors. For example, whether the date on which the total amount of money is significantly changed is the beginning of the month, the working day, the holiday, the number of days from the previous holiday, etc., the time-dependent variables broken down may be as shown in table 1.
TABLE 1 example table of time-dependent variables
In one embodiment, the time series data is subjected to feature displacement conversion to obtain training data. Because the fund flow prediction method provided by the disclosure is a prediction based on a time series, and the machine learning method is a supervised learning method, the time series is required to be converted into a format which can be used for the supervised learning. For example, assume that there are 3 features, var1, var2, and var3, respectively, and the predicted value is result. Assuming that the current value is predicted using the signature of the previous N days, assuming N is 2, var1(t) represents the value of signature var1 at time t.
The original time series-based data format is shown in table 2.
TABLE 2 data Format schematic Table based on time series
| Date | var1 | var2 | var3 | result |
| 20210701 | 11 | 21 | 31 | 1000 |
| 20210702 | 12 | 22 | 32 | 2000 |
| 20210703 | 13 | 23 | 33 | 3000 |
| 20210704 | 14 | 24 | 34 | 4000 |
The data shown in table 3 will be obtained after converting the data shown in table 2 into a format with supervised learning.
TABLE 3 data schematic of supervised learning format
In one embodiment, after the feature displacement conversion is performed on the time series data to obtain the training data, the method further includes dividing the training data into training set data and test set data. After the time sequence data is converted into a format with supervised learning, training data which can be used for machine learning can be obtained. The training data is randomly divided into two parts, wherein one part of the data is used as training set data, and the other part of the data is used as test set data. The training set data may be used as input to a LightGBM model for training the funds-in and funds-out prediction models. The test set data may be used as verification data for testing generalization ability of the trained fund inflow prediction model and fund outflow prediction model on the test set data, and determining prediction results of the trained fund inflow prediction model and fund outflow prediction model.
Aiming at the problem of fund flow prediction of financial data, a traditional time series algorithm ARIMA model is commonly used for solving at present. The ARIMA Model (Autoregressive Integrated Moving Average Model), which is called Autoregressive Integrated Moving Average Model, is an important prediction method based on time series in statistics. Firstly, the conversion from a non-stationary time sequence to a stationary time sequence needs to be completed, and then the regression is carried out on the current value and the lag value of the random error term, so as to establish an ARIMA model. The central idea of the ARIMA model is to treat the data sequence formed by the predicted object based on the time sequence as a random sequence and use a certain mathematical model to try to describe and fit the sequence. Once established, this model can be used to predict future values using past values of the sequence.
However, the ARIMA model has high requirements for data stationarity, and in an actual application scenario, real data cannot directly meet the requirements and can be obtained only through complex processing. Meanwhile, due to the fact that data are not stable, the parameters need to be selected through manual observation and calculation, and compared with a method obtained through a computer, the prediction result is poor, and even the situation that a solution cannot be obtained may occur. Meanwhile, the ARIMA model can capture only linear relation, cannot capture nonlinear relation and cannot capture characteristics such as seasonality, periodicity and the like. Finally, the prediction of the ARIMA model is only related to time, other abundant features except time cannot be added for learning prediction, and the input features are single.
In order to solve the possible problems of the ARIMA model of the conventional time series prediction algorithm, in some embodiments of the present disclosure, the LightGBM algorithm is adopted as a regression learning model framework. The fund-in prediction model and the fund-out prediction model are trained separately using a LightGBM regression algorithm with training set data.
The core of the LightGBM framework is a nonlinear model gbdt (gradient Boosting Decision tree) algorithm, an integrated learning model using a Decision tree as a base classifier. The model is based on Boosting thought, and through iteration of a plurality of decision trees, each new training is to improve the last training result. The Boosting basic learning mechanism is as follows: s1, training a base learner by using an initial training set, S2 adjusting the distribution and weight of training samples according to the performance of the base learner, so that the samples with wrong classification of the base learner get more attention in the iterative process, S3 training the next base learner based on the adjusted sample distribution, repeating the training optimization steps of the base learner from S2 to S3 until the number of the base learners reaches a preset value, and performing weighted combination on the prediction results of the base learner.
In addition, compared with the traditional integrated tree algorithm, the LightGBM is deeply optimized in engineering and supports mass data training. Unlike other decision tree-based boosting algorithms that grow trees in layers, LightGBM grows trees in leaf nodes, which can improve the accuracy of the model.
The fund flow prediction method provided by the disclosure uses the integrated learning model LightGBM regression of machine learning, converts time series characteristics into a supervised learning format, and uses data divided into a training set as input variables of the LightGBM model. When adjusting the parameters of the LightGBM model, the number of the base learners, the learning rate, the maximum depth of the tree, the regularization parameters, and other parameters may be focused. The fusion mode of combining the time series model and the machine learning regression model not only can fully utilize the time series model to mine the time-related characteristics important to loan behaviors, but also summarizes the important influence factors through data exploration of multiple parties. Compared with the traditional time series algorithm which only considers a single time variable, the input characteristics and the knowledge learned by the model of the fund flow prediction method are richer, and the prediction accuracy is greatly improved.
Fig. 2 is a flowchart illustrating a method for evaluating a fund inflow prediction model and a fund outflow prediction model in a second embodiment of the present disclosure, wherein in one embodiment, after training the fund inflow prediction model and the fund outflow prediction model respectively by using training data, the method further comprises evaluating prediction accuracy of the fund inflow prediction model and the fund outflow prediction model by using a fusion model error value of the fund inflow prediction model and the fund outflow prediction model. In some embodiments of the present disclosure, the test set data may be utilized to test the generalization ability of the trained funds-in and funds-out predictive models across the test set. In some embodiments of the present disclosure, for a trained model, evaluation indexes commonly used for regression problems such as MAE (Mean Absolute Error), MAPE (Mean Absolute Percentage Error), MSE (Mean Squared Error), RMSE (Root Mean Squared Error), R2 (R-Square), and the like may be used to evaluate training effects of the fund inflow prediction model and the fund outflow prediction model. In the present embodiment, the R2 evaluation mode is selected to evaluate the fund inflow prediction model and the fund outflow prediction model. And calculating a fusion model error value of the fund inflow prediction model and the fund outflow prediction model, and evaluating the prediction accuracy of the fund inflow prediction model and the fund outflow prediction model by using the obtained fusion model error value.
The fusion model error value may be determined in the following manner, including steps S410 to S430 as follows.
Step S410: the fund inflow prediction model predicts a user daily fund inflow prediction value according to the test set data, and the fund outflow prediction model predicts a user daily fund outflow prediction value according to the test set data.
Step S420: and calculating a daily loan repayment relative error according to the daily fund inflow predicted value and the actual fund inflow value of the user, and calculating a daily loan payment relative error according to the daily fund outflow predicted value and the actual fund outflow value of the user.
Step S430: and calculating a fusion model error value according to the relative error of the repayment of the daily loan and the relative error of the payment of the daily loan.
The final prediction result of the fund flow prediction method is obtained by fusing the two models of the fund inflow and the fund outflow, and in some embodiments of the disclosure, the final model effect is evaluated by using a single-point relative error addition method.
And calculating errors of the daily fund inflow and outflow and the actual gross sum of all the users on the test set, obtaining the relative error of daily loan repayment and the relative error of daily loan payment of the users, and further calculating a fusion model error value according to the relative error of the daily loan repayment and the relative error of the daily loan payment.
The daily loan budget is calculated in the following manner using relative errors:
wherein useamt is the relative error of loan branch per day, ziThe actual daily value of the fund flow for the user,
a daily fund flow prediction value for the user.
The relative error of daily loan repayment is calculated as follows:
wherein repayamt is the relative error of loan repayment every day, yiThe actual daily inflow of funds for the user,
daily fund inflow prediction values for the customers.
From the calculation results of the branch relative error and the repayment relative error, a score for estimating the effect of the prediction result, i.e., a fusion model error value can be obtained, and the score of the fusion model error value is calculated as follows:
totalscore＝score(useamt)*α+score(repayamt)*(1-α)
wherein totalscore is the error value of the fusion model, score () is a mapping function of the relative error and the final score, α is a weight coefficient for adjusting the two models, the value range of α is [0,1], useamt is the relative error for daily loan payment, and repayamt is the relative error for daily loan repayment. The mapping function score () of the relative error and the final score can be set reasonably according to the actual application.
When the error value of the fusion model calculated by the fund inflow prediction model and the fund outflow prediction model on the test set is lower than the preset value, or the accuracy of the prediction results of the fund inflow prediction model and the fund outflow prediction model on the training set is reduced a lot, it is indicated that the two models are possibly over-fitted, the fund inflow prediction model and the fund outflow prediction model need to be retrained, and the model parameters need to be readjusted. Similarly, when adjusting the model parameters, the number of basis learners, the learning rate, the maximum depth of the tree, and the regularization parameters may be of great interest.
It should be understood that although the various steps in the flowcharts of fig. 1-2 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 1-2 may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed in turn or alternately with other steps or at least some of the other steps or stages.
Based on the description of the embodiment of the fund flow prediction method, the disclosure also provides a fund flow prediction system. The system may include systems (including distributed systems), software (applications), modules, components, servers, clients, etc. that use the methods described in embodiments of the present specification in conjunction with any necessary hardware-implemented devices. Based on the same innovative concept, the embodiments of the present disclosure provide an apparatus in one or more embodiments as described in the following embodiments. Since the implementation scheme of the apparatus for solving the problem is similar to that of the method, the specific implementation of the apparatus in the embodiment of the present specification may refer to the implementation of the foregoing method, and repeated details are not repeated. As used hereinafter, the term "unit" or "module" may be a combination of software and/or hardware that implements a predetermined function. Although the means described in the embodiments below are preferably implemented in software, an implementation in hardware, or a combination of software and hardware is also possible and contemplated.
Fig. 3 is a block diagram of a fund flow prediction system in a third embodiment of the present disclosure, where in one embodiment, the fund flow prediction system may be a terminal, a server, or a module, a component, a device, a unit, etc. integrated in the terminal. The fund flow prediction system may include a data processing module 100, a model building module 200, and an outcome prediction module 300.
The data processing module 100 comprises a time series decomposition unit and a characteristic displacement conversion unit. The time sequence decomposition unit is used for performing time sequence decomposition on the historical fund flow data to acquire time sequence data. The characteristic displacement conversion unit is used for performing characteristic displacement conversion on the time series data to obtain training data, and the characteristic displacement conversion comprises the step of converting the time series data into data meeting the format requirement of input data of the selected prediction model. The model building module 200 comprises a model processing unit and a model training unit. The model processing unit is used for constructing a supervised learning fund inflow prediction model and a supervised learning fund outflow prediction model. The model training unit is used for respectively training the fund inflow prediction model and the fund outflow prediction model by utilizing the training data. And the result prediction module 300 is configured to perform weighted fusion on the output of the fund inflow prediction model and the output of the fund outflow prediction model to obtain a prediction result.
In one embodiment, the data processing module 100 further includes a data analysis unit for performing mobility analysis and special date behavior analysis on the historical fund flow data to determine a time-dependent influence factor.
In one embodiment, the data processing module further comprises a data set partitioning unit for partitioning the training data into training set data and test set data.
In one embodiment, the model training unit utilizes the training set data to train the fund-in prediction model and the fund-out prediction model, respectively, using the LightGBM regression algorithm.
In one embodiment, the model building module further comprises a model evaluation unit for evaluating the prediction accuracy of the fund-in prediction model and the fund-out prediction model using the fused model error values of the fund-in prediction model and the fund-out prediction model.
The fusion model error value is determined in the following manner:
the fund inflow prediction model predicts a user daily fund inflow prediction value according to the test set data, and the fund outflow prediction model predicts a user daily fund outflow prediction value according to the test set data. Calculating a daily loan repayment relative error according to the daily fund inflow predicted value and the actual fund inflow value of the user, and calculating a daily loan payment relative error according to the daily fund outflow predicted value and the actual fund outflow value of the user;
calculating a fusion model error value according to the relative error of the repayment of the daily loan and the relative error of the payment of the daily loan;
the calculation mode of the relative error for daily loan payment comprises the following steps:
wherein useamt is the relative error of loan branch per day, ziThe actual daily value of the fund flow for the user,
a daily fund flow prediction value for the user;
the calculation mode of the relative error of the daily loan repayment comprises the following steps:
wherein repayamt is the relative error of loan repayment every day, yiThe actual daily inflow of funds for the user,
a daily fund inflow prediction value for the user;
the calculation method of the fusion model error value comprises the following steps:
totalscore＝score(useamt)*α+score(repayamt)*(1-α)
wherein totalscore is a fusion model error value, score () is a mapping function of a relative error and a final score, α is a weight coefficient for adjusting the two models, the value range of α is [0,1], useamt is a relative error for daily loan payment, and repayamt is a relative error for daily loan repayment.
With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
It is to be understood that the various embodiments of the methods, apparatus, etc. described above are described in a progressive manner, and like/similar elements may be referred to one another, with each embodiment focusing on differences from the other embodiments. Reference may be made to the description of other method embodiments for relevant points.
Fig. 4 is a block diagram of a fund flow prediction apparatus or system according to a fourth embodiment of the present disclosure. Referring to fig. 4, a fund flow prediction apparatus or system S00 includes a processing component S20 that further includes one or more processors and memory resources, represented by memory S22, for storing instructions, such as an application, executable by the processing component S20. The application program stored in the memory S22 may include one or more modules each corresponding to a set of instructions. Further, the processing component S20 is configured to execute instructions to perform the above-described method.
The fund flow prediction apparatus or system S00 may further include: the power component S24 is configured to perform power management of the fund flow forecasting device or system S00, the wired or wireless network interface S26 is configured to connect the fund flow forecasting device or system S00 to a network, and the input output (I/O) interface S28. The fund flow predicting device or system S00 may be operable based on an operating system stored in the memory S22, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, or the like.
In an exemplary embodiment, a computer-readable storage medium comprising instructions, such as the memory S22 comprising instructions, executable by a processor of the fund flow prediction apparatus or system S00 to perform the above method is also provided. The storage medium may be a computer-readable storage medium, which may be, for example, a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.
In an exemplary embodiment, there is also provided a computer program product including instructions executable by a processor of the asset flow prediction device or system S00 to perform the method described above.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the hardware + program class embodiment, since it is substantially similar to the method embodiment, the description is simple, and the relevant points can be referred to the partial description of the method embodiment.
It should be noted that, the descriptions of the apparatus, the electronic device, the server, and the like according to the method embodiments may also include other embodiments, and specific implementations may refer to the descriptions of the related method embodiments. Meanwhile, the new embodiment formed by the mutual combination of the features of the methods, the devices, the equipment and the server embodiments still belongs to the implementation range covered by the present disclosure, and the details are not repeated herein.
In the description herein, references to the description of "some embodiments," "other embodiments," "desired embodiments," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, a schematic description of the above terminology may not necessarily refer to the same embodiment or example.
The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (14)
1. A method for predicting fund flow is characterized by comprising the following steps:
performing time series decomposition on the historical fund flow data to obtain time series data;
performing characteristic displacement transformation on the time sequence data to obtain training data, wherein the characteristic displacement transformation comprises the step of transforming the time sequence data into data meeting the format requirement of input data of a selected prediction model;
constructing a capital inflow forecasting model with supervised learning and a capital outflow forecasting model with supervised learning;
training the fund inflow prediction model and the fund outflow prediction model respectively by using the training data;
and performing weighted fusion on the output of the fund inflow prediction model and the output of the fund outflow prediction model to obtain a prediction result.
2. The fund flow prediction method according to claim 1, wherein before time-series decomposition of historical fund flow data to obtain time-series data, the method further comprises:
and carrying out mobility analysis and special date behavior analysis on the historical fund flow data to determine time-related influence factors.
3. The fund flow prediction method according to claim 2, wherein the time-series decomposition of the historical fund flow data and the obtaining of the time-series data comprises:
and performing time series decomposition on the historical fund flow data according to the influence factors by utilizing a Prophet time series tool to obtain time series data, wherein the time series data comprises a trend item, a period item, an activity effect item and an error item.
4. The method of claim 1, wherein after performing feature-shift transformation on the time-series data to obtain training data, the method further comprises:
the training data is divided into training set data and test set data.
5. The fund flow prediction method according to claim 4, wherein the fund inflow prediction model and the fund outflow prediction model are trained using a LightGBM regression algorithm using the training set data, respectively.
6. The fund flow prediction method according to claim 4, wherein after the training data is used to train the fund flow-in prediction model and the fund flow-out prediction model, respectively, the method further comprises:
evaluating the prediction accuracy of the fund inflow prediction model and the fund outflow prediction model by using a fusion model error value of the fund inflow prediction model and the fund outflow prediction model;
the fusion model error value is determined in the following manner:
the fund inflow prediction model predicts a user daily fund inflow prediction value according to the test set data, and the fund outflow prediction model predicts a user daily fund outflow prediction value according to the test set data;
calculating a daily loan repayment relative error according to the daily fund inflow predicted value and the actual fund inflow value of the user, and calculating a daily loan payment relative error according to the daily fund outflow predicted value and the actual fund outflow value of the user;
calculating the error value of the fusion model according to the relative error of the repayment of the daily loan and the relative error of the daily loan payment;
the calculation mode of the relative error for the daily loan payment comprises the following steps:
wherein useamt is the relative error of loan branch per day, ziThe actual daily value of the fund flow for the user,
a daily fund flow prediction value for the user;
the calculation mode of the relative error of the daily loan repayment comprises the following steps:
wherein repayamt is the relative error of loan repayment every day, yiThe actual daily inflow of funds for the user,
a daily fund inflow prediction value for the user;
the calculation method of the fusion model error value comprises the following steps:
totalscore＝score(useamt)*α+score(repayamt)*(1-α)
wherein totalscore is the error value of the fusion model, score () is a mapping function of the relative error and the final score, α is a weight coefficient for adjusting the two models, the value range of α is [0,1], useamt is the relative error for daily loan payment, and repayamt is the relative error for daily loan repayment.
7. A system for predicting a flow of funds, comprising:
the data processing module comprises a time sequence decomposition unit and a characteristic displacement conversion unit, wherein the time sequence decomposition unit is used for performing time sequence decomposition on historical fund flow data to obtain time sequence data; the characteristic displacement conversion unit is used for performing characteristic displacement conversion on the time sequence data to obtain training data, and the characteristic displacement conversion comprises the step of converting the time sequence data into data meeting the format requirement of input data of a selected prediction model;
the model building module comprises a model processing unit and a model training unit, wherein the model processing unit is used for building a capital inflow prediction model with supervised learning and a capital outflow prediction model with supervised learning; the model training unit is used for respectively training the fund inflow prediction model and the fund outflow prediction model by utilizing the training data;
and the result prediction module is used for performing weighted fusion on the output of the fund inflow prediction model and the output of the fund outflow prediction model to obtain a prediction result.
8. The system of claim 7, wherein the data processing module further comprises:
and the data analysis unit is used for carrying out mobility analysis and special date behavior analysis on the historical fund flow data and determining time-related influence factors.
9. The system of claim 7, wherein the data processing module further comprises:
and the data set dividing unit is used for dividing the training data into training set data and test set data.
10. The fund flow prediction system of claim 9, wherein the model training unit utilizes the training set data to train the fund flow-in prediction model and the fund flow-out prediction model, respectively, using a LightGBM regression algorithm.
11. The system of claim 9, wherein the model building module further comprises:
a model evaluation unit for evaluating prediction accuracy of the fund inflow prediction model and the fund outflow prediction model using a fusion model error value of the fund inflow prediction model and the fund outflow prediction model;
the fusion model error value is determined in the following manner:
the fund inflow prediction model predicts a user daily fund inflow prediction value according to the test set data, and the fund outflow prediction model predicts a user daily fund outflow prediction value according to the test set data;
the model evaluation unit calculates a daily loan repayment relative error according to the daily fund inflow predicted value and the actual fund inflow value of the user, and calculates a daily loan payment relative error according to the daily fund outflow predicted value and the actual fund outflow value of the user;
the model evaluation unit calculates the fusion model error value according to the relative error of the repayment of the daily loan and the relative error of the daily loan payment;
the calculation mode of the relative error for the daily loan payment comprises the following steps:
wherein useamt is the relative error of loan branch per day, ziThe actual daily value of the fund flow for the user,
a daily fund flow prediction value for the user;
the calculation mode of the relative error of the daily loan repayment comprises the following steps:
wherein repayamt is the relative error of loan repayment every day, yiThe actual daily inflow of funds for the user,
a daily fund inflow prediction value for the user;
the calculation method of the fusion model error value comprises the following steps:
totalscore＝score(useamt)*α+score(repayamt)*(1-α)
wherein totalscore is the error value of the fusion model, score () is a mapping function of the relative error and the final score, α is a weight coefficient for adjusting the two models, the value range of α is [0,1], useamt is the relative error for daily loan payment, and repayamt is the relative error for daily loan repayment.
12. A computer arrangement comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, carries out the steps of the method of fund flow prediction according to claims 1-6.
13. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of fund flow prediction according to any one of claims 1 to 6.
14. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, carries out the steps of the method of fund flow prediction according to claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111400988.7A CN114154696A (en) | 2021-11-19 | 2021-11-19 | Method, system, computer device and storage medium for predicting fund flow |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111400988.7A CN114154696A (en) | 2021-11-19 | 2021-11-19 | Method, system, computer device and storage medium for predicting fund flow |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114154696A true CN114154696A (en) | 2022-03-08 |
Family
ID=80457288
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111400988.7A Pending CN114154696A (en) | 2021-11-19 | 2021-11-19 | Method, system, computer device and storage medium for predicting fund flow |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114154696A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117520441A (en) * | 2024-01-03 | 2024-02-06 | 国网浙江省电力有限公司金华供电公司 | Method, device and equipment for detecting abnormity of fund flow data |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3188093A1 (en) * | 2015-12-29 | 2017-07-05 | Tata Consultancy Services Limited | System and method for predicting response time of an enterprise system |
| CN108388852A (en) * | 2018-02-09 | 2018-08-10 | 北京天元创新科技有限公司 | A kind of region crowd density prediction technique and device based on deep learning |
| CN110210644A (en) * | 2019-04-17 | 2019-09-06 | 浙江大学 | The traffic flow forecasting method integrated based on deep neural network |
| WO2019218751A1 (en) * | 2018-05-16 | 2019-11-21 | 阿里巴巴集团控股有限公司 | Processing method, apparatus and device for risk prediction of insurance service |
| CN110852476A (en) * | 2019-09-29 | 2020-02-28 | 华东理工大学 | Passenger flow prediction method and device, computer equipment and storage medium |
| CN111260156A (en) * | 2020-02-18 | 2020-06-09 | 中国农业银行股份有限公司 | Construction method of cash flow prediction model and cash flow prediction method and device |
| CN111401609A (en) * | 2020-03-04 | 2020-07-10 | 平安科技（深圳）有限公司 | Prediction method and prediction device for traffic flow time series |
| CN111597971A (en) * | 2020-05-14 | 2020-08-28 | 北京交通大学 | Method for predicting short-term arrival passenger flow of urban rail transit |
| US20200279198A1 (en) * | 2016-12-01 | 2020-09-03 | Trovata, Inc. | Cash forecast system, apparatus, and method |
| CN111738507A (en) * | 2020-06-19 | 2020-10-02 | 中国工商银行股份有限公司 | Bank clearing position fund payment amount prediction method, device, equipment and medium |
| CN111915024A (en) * | 2020-09-25 | 2020-11-10 | 点内（上海）生物科技有限公司 | Sequence prediction model training method, prediction system, prediction method and medium |
| WO2020258994A1 (en) * | 2019-06-24 | 2020-12-30 | 深圳前海微众银行股份有限公司 | Node data prediction method and device |
| WO2021082809A1 (en) * | 2019-10-29 | 2021-05-06 | 山东科技大学 | Training optimization method for foreign exchange time series prediction |
| CN113011666A (en) * | 2021-03-29 | 2021-06-22 | 上海寻梦信息技术有限公司 | Aging prediction model training method, aging prediction method and related equipment |
| CN113129127A (en) * | 2021-04-21 | 2021-07-16 | 建信金融科技有限责任公司 | Early warning method and device |
| CN113553540A (en) * | 2020-04-24 | 2021-10-26 | 株式会社日立制作所 | Commodity sales prediction method |
| CN113657656A (en) * | 2021-08-10 | 2021-11-16 | 建信金融科技有限责任公司 | Loan data analysis and prediction method and device |
-
2021
- 2021-11-19 CN CN202111400988.7A patent/CN114154696A/en active Pending
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3188093A1 (en) * | 2015-12-29 | 2017-07-05 | Tata Consultancy Services Limited | System and method for predicting response time of an enterprise system |
| US20200279198A1 (en) * | 2016-12-01 | 2020-09-03 | Trovata, Inc. | Cash forecast system, apparatus, and method |
| CN108388852A (en) * | 2018-02-09 | 2018-08-10 | 北京天元创新科技有限公司 | A kind of region crowd density prediction technique and device based on deep learning |
| WO2019218751A1 (en) * | 2018-05-16 | 2019-11-21 | 阿里巴巴集团控股有限公司 | Processing method, apparatus and device for risk prediction of insurance service |
| CN110210644A (en) * | 2019-04-17 | 2019-09-06 | 浙江大学 | The traffic flow forecasting method integrated based on deep neural network |
| WO2020258994A1 (en) * | 2019-06-24 | 2020-12-30 | 深圳前海微众银行股份有限公司 | Node data prediction method and device |
| CN110852476A (en) * | 2019-09-29 | 2020-02-28 | 华东理工大学 | Passenger flow prediction method and device, computer equipment and storage medium |
| WO2021082809A1 (en) * | 2019-10-29 | 2021-05-06 | 山东科技大学 | Training optimization method for foreign exchange time series prediction |
| CN111260156A (en) * | 2020-02-18 | 2020-06-09 | 中国农业银行股份有限公司 | Construction method of cash flow prediction model and cash flow prediction method and device |
| CN111401609A (en) * | 2020-03-04 | 2020-07-10 | 平安科技（深圳）有限公司 | Prediction method and prediction device for traffic flow time series |
| CN113553540A (en) * | 2020-04-24 | 2021-10-26 | 株式会社日立制作所 | Commodity sales prediction method |
| CN111597971A (en) * | 2020-05-14 | 2020-08-28 | 北京交通大学 | Method for predicting short-term arrival passenger flow of urban rail transit |
| CN111738507A (en) * | 2020-06-19 | 2020-10-02 | 中国工商银行股份有限公司 | Bank clearing position fund payment amount prediction method, device, equipment and medium |
| CN111915024A (en) * | 2020-09-25 | 2020-11-10 | 点内（上海）生物科技有限公司 | Sequence prediction model training method, prediction system, prediction method and medium |
| CN113011666A (en) * | 2021-03-29 | 2021-06-22 | 上海寻梦信息技术有限公司 | Aging prediction model training method, aging prediction method and related equipment |
| CN113129127A (en) * | 2021-04-21 | 2021-07-16 | 建信金融科技有限责任公司 | Early warning method and device |
| CN113657656A (en) * | 2021-08-10 | 2021-11-16 | 建信金融科技有限责任公司 | Loan data analysis and prediction method and device |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117520441A (en) * | 2024-01-03 | 2024-02-06 | 国网浙江省电力有限公司金华供电公司 | Method, device and equipment for detecting abnormity of fund flow data |
| CN117520441B (en) * | 2024-01-03 | 2024-03-15 | 国网浙江省电力有限公司金华供电公司 | Method, device and equipment for detecting abnormity of fund flow data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8010324B1 (en) | Computer-implemented system and method for storing data analysis models | |
| US7251589B1 (en) | Computer-implemented system and method for generating forecasts | |
| Celik et al. | Evaluating and forecasting banking crises through neural network models: An application for Turkish banking sector | |
| Moody | Economic forecasting: Challenges and neural network solutions | |
| CN116090839B (en) | Multiple risk analysis and evaluation method and system for water resource coupling system | |
| Assaad et al. | Predicting the price of crude oil and its fluctuations using computational econometrics: deep learning, LSTM, and convolutional neural networks | |
| Hamerle et al. | Parameterizing credit risk models | |
| Liu et al. | LSTM based reserve prediction for bank outlets | |
| CN112700324A (en) | User loan default prediction method based on combination of Catboost and restricted Boltzmann machine | |
| Panda et al. | Long term electrical load forecasting: an empirical study across techniques and domains | |
| Zhemkov | Nowcasting Russian GDP using forecast combination approach | |
| Temür et al. | Comparison of forecasting performance of ARIMA LSTM and HYBRID models for the sales volume budget of a manufacturing enterprise | |
| CN114154696A (en) | Method, system, computer device and storage medium for predicting fund flow | |
| Abu-taieh et al. | Multi-criteria decision-making using an extended Gale and Shapley (E-GS) algorithm | |
| CN113537540B (en) | Heat supply gas consumption prediction model based on automatic characteristic engineering | |
| Ead | Using Monte Carlo simulation to evaluate performance of forecasting models in project control | |
| Cracan | Retail sales forecasting using LSTM and ARIMA-LSTM: A comparison with traditional econometric models and Artificial Neural Networks | |
| Derras et al. | Prediction of recovery time of infrastructure functionalities after an earthquake using machine learning | |
| Dendramis et al. | Measuring the Default Risk of Small Business Loans: Improved Credit Risk Prediction using Deep Learning | |
| Tang et al. | RFM user value tags and XGBoost algorithm for analyzing electricity customer demand data | |
| Chandrasekaran et al. | Uncertainty-Aware Functional Analysis for Electricity Consumption Prediction Using Multi-Task Optimization Learning Model | |
| Ogut et al. | PV Fleet Modeling via Smooth Periodic Gaussian Copula | |
| CN118410922B (en) | Data processing method and system based on product supply chain | |
| Supriyanto | Comparison of Grid Search and Evolutionary Parameter Optimization with Neural Networks on JCI Stock Price Movements during the Covid 19 | |
| DANIALI et al. | Periodizing management’s risks of construction projects with gray relational analysis and fmea approach |
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 |