CN113935824A - Loan intention estimation method and system based on deep learning model - Google Patents
Loan intention estimation method and system based on deep learning model Download PDFInfo
- Publication number
- CN113935824A CN113935824A CN202111193601.5A CN202111193601A CN113935824A CN 113935824 A CN113935824 A CN 113935824A CN 202111193601 A CN202111193601 A CN 202111193601A CN 113935824 A CN113935824 A CN 113935824A
- Authority
- CN
- China
- Prior art keywords
- data
- loan
- model
- information
- unit
- 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 81
- 238000013136 deep learning model Methods 0.000 title claims abstract description 39
- 238000012545 processing Methods 0.000 claims abstract description 122
- 238000012549 training Methods 0.000 claims abstract description 76
- 230000006399 behavior Effects 0.000 claims abstract description 52
- 238000011156 evaluation Methods 0.000 claims description 58
- 238000001514 detection method Methods 0.000 claims description 20
- 238000006243 chemical reaction Methods 0.000 claims description 18
- 238000000605 extraction Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 4
- 230000005484 gravity Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 description 21
- 239000013598 vector Substances 0.000 description 9
- 238000012795 verification Methods 0.000 description 9
- 230000000694 effects Effects 0.000 description 8
- 238000007781 pre-processing Methods 0.000 description 8
- 238000012360 testing method Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000002372 labelling Methods 0.000 description 5
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 238000003672 processing method Methods 0.000 description 5
- 238000004590 computer program Methods 0.000 description 4
- 238000010276 construction Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 239000012634 fragment Substances 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000012804 iterative process Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 2
- 230000003542 behavioural effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- General Business, Economics & Management (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention provides a loan intention estimation method and system based on a deep learning model, wherein the method comprises the following steps: acquiring data and dividing the ratio of the loan data to obtain first loan data; inputting the first loan data into a data processing module, wherein the data processing module comprises a numerical processing unit and a text processing unit; obtaining first output data according to the data processing module; inputting the first output data into a deep learning model to extract a first representative behavior feature of a user from a time sequence; taking the first representative behavior feature as first input information; performing model training by using the first input information as basic training data of a training loan intention estimation model to obtain first estimation information; obtaining a first effective loan variable by analyzing the effectiveness of the external loan variable of the user; and carrying out incremental learning on the loan intention estimation model by using the first effective loan variable to obtain second estimation information output by the loan intention estimation model.
Description
Technical Field
The invention relates to the technical field related to new-generation information technology, in particular to a loan intention estimation method and system based on a deep learning model.
Background
In the financial industry, conventional financial institutions rely on the loan cases of customers and professional experience to perform general demand assessment on the customers, so as to provide more accurate financial services for the customers. With the rapid development of artificial intelligence, methods for evaluating the needs of customers by collecting big data are gradually proposed.
The current technology mainly collects recent customer data and relies on expert experience to manually identify characteristic information of customers, so that a training model processes and evaluates customer requirements for the recent data, and service accuracy is improved.
However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the inventors of the present application find that the above-mentioned technology has at least the following technical problems:
in the prior art, data features need to be labeled by depending on expert experience, and due to lack of elements including time sequences, behavior trends in a future time window cannot be estimated, so that the technical problems of large workload of feature labeling and insufficient accuracy of prediction of customer requirements exist.
Disclosure of Invention
The embodiment of the application provides a loan intention estimation method and system based on a deep learning model, and solves the technical problems that in the prior art, data characteristics need to be labeled by depending on expert experience, and due to lack of elements including time sequences, behavior trends in a future time window cannot be estimated, so that the workload of labeling the characteristics is large, and the accuracy of predicting the requirements of customers is insufficient. By collecting loan data and carrying out basic processing on the loan data to obtain structured numerical data and unstructured text data, adding time series elements into the structured numerical data and the unstructured text data to extract behavior characteristic information of a client, training a model by using the characteristic information to obtain demand prediction information of the client, carrying out incremental learning on the model according to external effective loan data, adding time elements and automatically extracting characteristics, adding an effective loan data training model, and achieving the technical effects of improving generalization capability and prediction accuracy of the model.
In view of the foregoing problems, the embodiments of the present application provide a loan intention estimation method and system based on a deep learning model.
In a first aspect, an embodiment of the present application provides a loan intention estimation method based on a deep learning model, where the method includes: acquiring data and dividing the ratio of the loan data to obtain first loan data; inputting the first loan data into a data processing module, wherein the data processing module comprises a numerical processing unit and a text processing unit; obtaining first output data according to the data processing module; inputting the first output data into a deep learning model to extract a first representative behavior feature of a user from a time sequence; taking the first representative behavior feature as first input information; performing model training by using the first input information as basic training data of a training loan intention estimation model to obtain first estimation information; obtaining a first effective loan variable by analyzing the effectiveness of the external loan variable of the user; and carrying out incremental learning on the loan intention estimation model by using the first effective loan variable to obtain second estimation information output by the loan intention estimation model.
On the other hand, the embodiment of the application provides a loan intention estimation system based on a deep learning model, wherein the system includes: a first obtaining unit: the first obtaining unit is used for obtaining first loan data by carrying out data acquisition and proportion division on the loan data; a first input unit: the first input unit is used for inputting the first loan data into a data processing module, wherein the data processing module comprises a numerical processing unit and a text processing unit; a second obtaining unit: the second obtaining unit is used for obtaining first output data according to the data processing module; a first extraction unit: the first extraction unit is used for inputting the first output data into a deep learning model to extract a first representative behavior feature of a user from a time sequence; a first setting unit: the first setting unit is used for taking the first representative behavior characteristic as first input information; a first processing unit: the first processing unit is used for performing model training by taking the first input information as basic training data of a training loan intention estimation model to obtain first estimation information; a second processing unit: the second processing unit is used for obtaining a first effective loan variable by analyzing the effectiveness of the external loan variable of the user; a third processing unit: the third processing unit is used for carrying out increment learning on the loan intention estimation model by the first effective loan variable to obtain second estimation information output by the loan intention estimation model.
In a third aspect, an embodiment of the present application provides a loan intention estimation system based on a deep learning model, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the steps of the method according to any one of the first aspect when executing the program.
One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:
the method comprises the steps of acquiring data and dividing the ratio of the borrowing data to obtain first borrowing data; inputting the first loan data into a data processing module, wherein the data processing module comprises a numerical processing unit and a text processing unit; obtaining first output data according to the data processing module; inputting the first output data into a deep learning model to extract a first representative behavior feature of a user from a time sequence; taking the first representative behavior feature as first input information; performing model training by using the first input information as basic training data of a training loan intention estimation model to obtain first estimation information; obtaining a first effective loan variable by analyzing the effectiveness of the external loan variable of the user; the first effective loan variable is subjected to incremental learning on the loan intention estimation model to obtain second estimation information output by the loan intention estimation model, by collecting loan data and carrying out basic processing on the loan data to obtain structured numerical data and unstructured text data, adding time series elements into the structured numerical data and the unstructured text data to extract behavior characteristic information of a client, then the characteristic information is used for training the model to obtain the demand forecast information of the client, and the model is subjected to incremental learning according to the external effective loan data, and the borrowing and lending demand prediction information output by the model is the result according with the actual situation of the client and the financial institution, time elements are added, the characteristics are automatically extracted, and then an effective borrowing and lending data training model is added, so that the technical effect of improving the generalization capability and the prediction accuracy of the model is achieved.
The foregoing description is only an overview of the technical solutions of the present application, and the present application can be implemented according to the content of the description in order to make the technical means of the present application more clearly understood, and the following detailed description of the present application is given in order to make the above and other objects, features, and advantages of the present application more clearly understandable.
Drawings
Fig. 1 is a schematic flow chart of a loan intention estimation method based on a deep learning model according to an embodiment of the present application;
fig. 2 is a schematic flow chart of a model output uncertainty determination method of a loan intention estimation model based on a deep learning model according to an embodiment of the present application;
fig. 3 is a schematic flow chart of a method for determining confidence interval in loan intention estimation process based on a deep learning model according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of a loan intention estimation system based on a deep learning model according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of an exemplary electronic device according to an embodiment of the present application.
Description of reference numerals: the device comprises a first obtaining unit 11, a first input unit 12, a second obtaining unit 13, a first extracting unit 14, a first setting unit 15, a first processing unit 16, a second processing unit 17, a third processing unit 18, an electronic device 300, a memory 301, a processor 302, a communication interface 303 and a bus architecture 304.
Detailed Description
The embodiment of the application provides a loan intention estimation method and system based on a deep learning model, and solves the technical problems that in the prior art, data characteristics need to be labeled by depending on expert experience, and due to lack of elements including time sequences, behavior trends in a future time window cannot be estimated, so that the workload of labeling the characteristics is large, and the accuracy of predicting the requirements of customers is insufficient. By collecting loan data and carrying out basic processing on the loan data to obtain structured numerical data and unstructured text data, adding time series elements into the structured numerical data and the unstructured text data to extract behavior characteristic information of a client, training a model by using the characteristic information to obtain demand prediction information of the client, carrying out incremental learning on the model according to external effective loan data, adding time elements and automatically extracting characteristics, adding an effective loan data training model, and achieving the technical effects of improving generalization capability and prediction accuracy of the model.
Summary of the application
In the financial industry, conventional financial institutions rely on the loan cases of customers and professional experience to perform general demand assessment on the customers, so as to provide more accurate financial services for the customers. With the rapid development of artificial intelligence, methods for evaluating the needs of customers by collecting big data are gradually proposed. The current technology mainly collects recent customer data and relies on expert experience to manually identify characteristic information of customers, so that a training model processes and evaluates customer requirements for the recent data, and service accuracy is improved. However, in the prior art, data features need to be labeled by depending on expert experience, and due to lack of elements including time sequences, behavior trends in a future time window cannot be estimated, so that the technical problems of large workload of feature labeling and insufficient accuracy of prediction of customer requirements exist.
In view of the above technical problems, the technical solution provided by the present application has the following general idea:
the embodiment of the application provides a loan intention estimation method based on a deep learning model, wherein the method comprises the following steps: acquiring data and dividing the ratio of the loan data to obtain first loan data; inputting the first loan data into a data processing module, wherein the data processing module comprises a numerical processing unit and a text processing unit; obtaining first output data according to the data processing module; inputting the first output data into a deep learning model to extract a first representative behavior feature of a user from a time sequence; taking the first representative behavior feature as first input information; performing model training by using the first input information as basic training data of a training loan intention estimation model to obtain first estimation information; obtaining a first effective loan variable by analyzing the effectiveness of the external loan variable of the user; and carrying out incremental learning on the loan intention estimation model by using the first effective loan variable to obtain second estimation information output by the loan intention estimation model.
Having thus described the general principles of the present application, various non-limiting embodiments thereof will now be described in detail with reference to the accompanying drawings.
Example one
As shown in fig. 1, an embodiment of the present application provides a loan intention estimation method based on a deep learning model, where the method includes:
s100: acquiring data and dividing the ratio of the loan data to obtain first loan data;
specifically, the first lending data is a data set obtained by collecting lending client data according to big data and performing proportion division processing. The processing method is not limited: after borrowing and lending data of a client are collected and stored based on big data, the data are obtained according to the following steps of 8: 1: the proportion of 1 is divided into a training data set, a verification data set and a test data set respectively, wherein the training data set is used for training data of a client loan intention prediction model in the next step, the verification data set is used for verifying the performance stability degree of model iteration, and the test data set is used for detecting the output accuracy and the convergence degree of the model. The method comprises the steps of carrying out proportion division on loan data acquired based on big data and then storing the loan data to obtain first loan data, and providing a comprehensive data base for model training in the next step.
S200: inputting the first loan data into a data processing module, wherein the data processing module comprises a numerical processing unit and a text processing unit;
s300: obtaining first output data according to the data processing module;
specifically, the data processing model is a model for preprocessing data, and the preferred preprocessing methods are mainly divided into two types:
when the first loan data are structured numerical data, the first loan data are input into a numerical processing unit in the data processing model to be subjected to standardization (normalization), the first loan data are subjected to standard processing by using a min-max standardization processing method or a Z-score standardization processing method in an exemplary manner, after the first loan data are subjected to data standardization processing, indexes are in the same order of magnitude, the numerical data are subjected to standardization processing, and the result of the model can obtain convergence more quickly in the process of descending random gradients; the first loan data is detrended (ascending/descending) and detreriod to ensure that the most basic behavioral characteristics are extracted from the time series.
When the first loan data is unstructured text data, inputting a text processing unit in the data processing model to perform feature digital preprocessing, wherein an exemplary One-Hot-based coding uses integers for coding, and the coding process is to use an N-bit state register to code N states of loan clients under the same time node of the first loan data, wherein each state has an independent register bit, and only One bit is valid at any time, and the preprocessing of the unstructured text data is convenient to correspond to behavior feature vectors extracted based on time series at the later step.
After the first lending data are processed in the two modes, the processed data are decentralized, the importance degrees of all training data after the data are decentralized are the same, the influence of one group of training data is avoided being large, the generalization capability of the model obtained through training is further improved, and the accuracy of the model in processing data is improved. Through carrying out preprocessing on the first lending data, the obtained first output data are more suitable for a data set constructed by the model, so that the constructability of the model is improved, and the falling possibility of the model is increased.
S400: inputting the first output data into a deep learning model to extract a first representative behavior feature of a user from a time sequence;
specifically, in order to enhance the accuracy of the prediction of the future time demand of the customer based on the first output data, the first output data is input into a deep learning model to extract user behavior features corresponding to time series, and the user behavior features are recorded as the first representative behavior features. The following are exemplary: the deep learning model uses an encoder-decoder LSTM model, namely a time sequence prediction model, the encoder-decoder LSTM model is of a double-layer structure consisting of two submodels, and the first layer submodel is as follows: an encoder for reading and encoding an input sequence; the second layer of sub-models: a decoder for reading the coded input sequence and predicting the user behavior characteristics of each element in the output sequence under the time sequence is connected with the LSTM model; and obtaining user behavior characteristics representing a single element in the first output data under a time sequence and a characteristic vector corresponding to a user predicted behavior through multiple groups of processing, recording the characteristic as the first behavior characteristics, returning and storing the first behavior characteristics into a first layer of encoder, and stopping acquisition when the acquired characteristic vector reaches a preset coverage rate and a preset data volume, wherein the preset coverage rate is the proportion of the preset characteristic vector covering the type of the user characteristic behavior, and the preset data volume is the preset data volume enough for training a model, and can be set by self without limitation. By collecting enough and representative feature vectors of the first representative behavior features of the user, the prediction accuracy of the model for the future time requirements of the client is improved.
S500: taking the first representative behavior feature as first input information;
s600: performing model training by using the first input information as basic training data of a training loan intention estimation model to obtain first estimation information;
specifically, the first representative behavior feature is set as the first input information training loan intention prediction model, wherein the loan intention prediction model is an intelligent model based on neural network training, the preferred construction method is that on the basis of the deep learning model, when the feature vector stored in the encoder reaches the preset coverage rate and the preset data amount, the decoder is removed, a predictor with a full-connection structure is externally connected to the original encoder to construct the original model frame of the loan intention prediction model, then the loan intention prediction model is trained based on the first input information, the training data set is used to train the model in the training process, the iterative performance measurement of the model is verified by using the verification data set, the convergence degree of the model is verified by using the verification data set, and the first prediction information is the deviation parameter generated in the prediction model training process, the output result of the model can be adjusted in the training process by predicting the model parameters, so that a more accurate output result is obtained.
The loan willingness prediction model is trained through the first input information extracted based on the time sequence, so that the user behavior prediction capability of a client in a period of time in the future is improved, the corresponding financial service is provided in a targeted manner, and the technical effect of enhancing the prediction accuracy of the model for the client's future time requirement is achieved.
S700: obtaining a first effective loan variable by analyzing the effectiveness of the external loan variable of the user;
s800: and carrying out incremental learning on the loan intention estimation model by using the first effective loan variable to obtain second estimation information output by the loan intention estimation model.
Specifically, the first valid loan variable is a loan service function that is actually available for a future time period, the future time period being the same as the time period predicted for the customer behavior; furthermore, increment learning training is carried out on the loan intention estimation model based on the first effective loan variable, so that external variable factors influencing loan are added into the intention prediction model, and the second estimation information output by the loan intention estimation model after increment learning accurately represents the loan intention of the user according with actual complex conditions, wherein external factors include but are not limited to city, gender, holidays, applied loan categories and other factors. Incremental learning means that a learning system can continuously learn new knowledge from a new sample, most of the previously learned knowledge can be stored, model incremental learning is performed by using external variable data on the basis of the loan intention estimation model which is trained originally, applicability of the model and accuracy of a processing result are improved, time cost is reduced compared with a retraining model, and training efficiency is improved.
Further, as shown in fig. 2, based on the model training by using the first input information as basic training data for training the loan intention estimation model, the method step S600 further includes:
s610: constructing a model evaluation unit, wherein the model evaluation unit comprises a first evaluation unit and a second evaluation unit, the first evaluation unit is used for parameter deviation evaluation, and the second evaluation unit is used for influence deviation evaluation;
s620: inputting the loan intention estimation model into the model evaluation unit to obtain first evaluation information and second evaluation information output by the model evaluation unit, wherein the first evaluation information is estimation deviation information of the first evaluation unit, and the second evaluation information is estimation deviation information of the second evaluation unit;
s630: and generating the first estimation information according to the first estimation information and the second estimation information, wherein the first estimation information is prediction deviation information.
Specifically, the model evaluation unit is a module for evaluating deviation generated by the model, and comprises a first evaluation unit for evaluating parameter deviation generated during model training and a second evaluation unit for evaluating deviation caused by influence of a multi-training iterative process, wherein the deviation is divided into model parameter deviation and deviation caused by influence of the iterative process and is marked as noise influence deviation.
The determination method of the model parameter deviation is as follows: in order to improve the prediction accuracy of the model, the model is subjected to iterative training for many times, and a new sample point x is selected from the test set every time in the training process*And randomly removing hidden units in each hidden layer with a probability p in the process of forwarding the model. When the hidden unit is randomly discarded, a variational discarding method is used in an encoder, and a conventional discarding method is used in a predictor, wherein the variational discarding method refers to that when the discarding method is used, each element of a parameter matrix is randomly discarded, and the same discarding mask is used at all times; the conventional discarding method means that when a deep neural network is trained, a part of neurons can be discarded randomly (and corresponding connecting edges of the neurons are discarded at the same time) to obtain a network with fewer nodes; the probability p can be set by itself, preferably 0.5. Further, repeating the process for B times to obtain B different products
Value, recorded as
The output values obtained in the predictor after using the conventional dropping method are input for the samples after using the variational dropping method at the encoder. Still further, uncertainty of the model can be evaluated using uncertainty of all samples:
wherein
Representing the average output value of B samples obtained in an input predictor after a variational discarding method is used in an encoder, and B represents
To
When the number of samples B is large enough, the deviation of the samples can be used for representing the parameter deviation of the model,
and representing the parameter deviation of the model and recording the parameter deviation as the first evaluation information.
The model parameter deviation obtaining process of the B times can be regarded as B times of iteration process of the model, in the B times of iteration process, the performance of model iteration is verified by using a verification data set, the deviation generated in the verification iteration process is called noise influence deviation, wherein the uncertainty determination mode caused by noise is as follows: derived from the representation of the data on the verification set, X '═ { X'1,…,x′VY'1,…,y′VPoints in the validation set, uncertainty of the data on the validation set
Wherein X ' and Y ' are verification data set, X '1To x'VTo verify a feature vector, y ', in a data set characterizing a behavior feature of a user as a function of a time sequence'1To y'VTo verify feature vector identification information characterizing the predicted behavior of the user in the data set,
feature vector input prediction for X' after using a variational discarding method at an encoderThe result of the response of the detector is,
and representing the deviation generated by the verification process and recording the deviation as the second evaluation information.
Furthermore, deviation data of the whole model after B iterations can be obtained and used
And x represents the overall uncertainty of the model after B iterations and is recorded as the first estimated information, and when the first estimated information reaches a preset value for a certain time, the model is basically converged, wherein the preset value is a deviation value corresponding to the preset accuracy of the model. And defining a reference value for the convergence degree of the model by constructing the first estimation information, so that the realizability of the model is improved.
Further, as shown in fig. 3, based on the generating the first pre-estimated information according to the first evaluation information and the second evaluation information, the method step S630 further includes:
s631: constructing a first confidence interval according to the first pre-estimated information;
s632: determining a first confidence level coefficient by performing a confidence level analysis on the first confidence interval;
s633: and when the first confidence level coefficient meets a preset confidence level coefficient, realizing abnormal detection according to the first confidence interval.
Specifically, the first confidence interval refers to an error range between a sample statistic value and an overall parameter value under a certain confidence level, and the greater the confidence interval, the higher the confidence level; the first confidence level coefficient refers to the probability that the overall parameter value falls within an interval of sample statistics. By way of example and not limitation: b user predicted characteristic behavior results generated by the B iterations correspond to B model overall deviations, and the average value corresponding to the B user predicted characteristic behavior results is the average value
Recording the first pre-estimated information x corresponding to the model integral deviation at the B-th time as a preset deviation value, and constructing a preset confidence interval
When the deviation value is within a preset confidence interval, testing whether the model is converged or not by using the test data set, wherein in B iterations, the proportion of n user predicted characteristic behavior results falling into the preset confidence interval occupying B user predicted characteristic behavior results is the preset confidence level coefficient; the first confidence level coefficient is a confidence level coefficient corresponding to a confidence interval in the iterative process from 1 to B-1 times, when the first confidence level coefficient is the same as the preset confidence level coefficient, the convergence degree of the model can be detected by using the test data set, and when the detection result is unconverged, which indicates that the model is abnormal, the iterative training of the model needs to be continued; when the first confidence level coefficient is different from the preset confidence level coefficient, the iterative training is continued until the model reaches convergence, and the training is stopped. And carrying out abnormity detection on the model reaching the preset confidence interval by constructing the first confidence interval, and stopping training when the model reaches convergence, thereby improving the accuracy of model training.
Further, based on the inputting of the first loan data into a data processing module, where the data processing module includes a numerical processing unit and a text processing unit, the method step S200 further includes:
s210: classifying the first borrowing data to obtain first numerical value borrowing data and first text borrowing data;
s220: inputting the first numerical value lending data into the numerical value processing unit for normalization processing and dissimilarity processing to obtain second numerical value lending data;
s230: inputting the first text lending data into the text processing unit for coding to obtain second text lending data;
s240: and taking the second numerical value lending data and the second text lending data as output information of the data processing module.
Specifically, when the first borrowing data is structured numerical data, the first borrowing data is recorded as the first numerical value borrowing data; the second numerical value lending data is a data set obtained by inputting the first numerical value lending data into a numerical value processing unit in the data processing model for standardization (normalization), and deleting data which has a difference with other numerical values exceeding a preset threshold value, wherein the preset threshold value can be set according to expert experience. And performing standard processing on the first loan data by using a min-max standard processing method or a Z-score standard processing method, wherein after the first loan data is subjected to data standard processing, indexes of the first loan data are in the same order of magnitude, comparing the indexes, and deleting single data exceeding a preset threshold. The numerical data is subjected to standardization processing and dissimilarity processing, so that the convergence of the model result can be obtained more quickly in the process of random gradient descent; further, the second loan data is detrended (ascending/descending) and detreriod to ensure that the most basic behavioral characteristics are extracted from the time series.
When the first borrowing data is unstructured text data, recording as the first text borrowing data; the second text data is characterized in that the first text lending data is input into a text processing unit in the data processing model for feature digital preprocessing, an exemplary One-Hot-based code uses integers for coding, and the coding process is to use an N-bit state register to code N states of lending clients of the first lending data at the same time node, wherein each state has an independent register bit, and only One bit is effective at any time.
And setting the preprocessed second numerical value lending data and the preprocessed second text lending data as the first output data, and preprocessing the first lending data to obtain the first output data which is more suitable for a data set constructed by the model, so that the constructability of the model is improved, and the landing possibility of the model is increased.
Further, the method further includes step S900:
s910: constructing a first sliding rule, wherein the first sliding rule is a coverage sliding logic rule;
s920: the sliding window carries out time-based conversion on the first loan data according to the first sliding rule to obtain first loan conversion data corresponding to the first loan data;
s930: and inputting the first loan conversion data into a segment detection model to obtain a first detection result in the segment detection model.
Specifically, the first sliding rule refers to a predetermined decentralization rule for decentralizing loan data corresponding to a time series and avoiding that the influence of the first loan data corresponding to a certain time series on a model is different from the influence of the first loan data corresponding to other time series; further, in order to avoid that fragmented data without time elements are obtained while decentralizing, the first sliding rule is preset to be at least 50% coverage, the sliding window is a plug-in for decentralizing the first loan data, decentralizing is performed on the first loan data in the sliding window, which is defined as a timesharing removal conversion, and the coverage of the sliding window on the first loan data is at least 50% coverage preset by the first sliding rule. When the data obtained after the first loan data is subjected to the time conversion is marked as the first loan conversion data, in order to avoid the situation that the obtained data is fragmented data, the first loan conversion data is input into a fragment detection model for detection, and the fragment detection model is a model constructed based on a preset time sequence. When the first loan conversion data is input into the fragment detection model, if the time sequence corresponding to the first loan conversion data is smaller than a preset time sequence, the first loan conversion data is fragment data and needs to be acquired again; and if the time sequence corresponding to the first loan conversion data is greater than or equal to a preset time sequence, representing the data as preprocessed data, and then constructing a next step model. The de-temporal conversion is preferably used as a preprocessing mode after step S200, and the output accuracy of the model trained by the preprocessed data is higher.
Further, based on the incremental learning of the loan intention estimation model by using the first effective loan variable, second estimation information output by the loan intention estimation model is obtained, and step S800 of the method further includes:
further, the first valid loan variable comprises a first predetermined variable specific gravity.
S810: obtaining first loss data according to the first effective loan variable;
s820: inputting the first loss data into the loan intention estimation model for incremental learning to obtain an incremental intention estimation model, wherein the incremental intention estimation model is obtained by training a plurality of groups of data to convergence;
s830: and obtaining the second estimated information according to the incremental will estimation model.
Specifically, the first preset variable proportion is data representing the influence degree of different first valid loan variables on the loan intention forecast information of the user, and examples are as follows: the influence degrees of external variable information such as cities, sexes, holidays, loan types applied, loan averages of the past 90 days and the like on the loan intention estimation information of the client are differentiated, and different influence proportions are preferably obtained according to expert experience evaluation to obtain the first preset variable proportion; further, the first loss data is information representing the comprehensive influence degree of the first effective loan variable on loan intention estimation information; furthermore, the first loss data is input into the loan intention estimation model for incremental learning, a plurality of groups of first loss data are used for iterative training of the loan intention estimation model, when convergence is achieved, the increment intention estimation model is obtained, the second estimation information output by the increment intention estimation model can accurately represent the user behavior prediction characteristics of the user in a period of time in the future, the acquired data period is far longer than the predictable duration, and the method is exemplarily shown as follows: three months of historical data were processed to analyze the user behavior prediction characteristics for the next 15 days.
External variable factors are added into loan intention estimation factors through incremental learning, so that the incremental intention estimation model obtained through training can accurately represent user behavior prediction characteristics of a user in a future period of time, and the adaptability and generalization capability of the model are improved.
To sum up, the loan intention estimation method and system based on the deep learning model provided by the embodiment of the application have the following technical effects:
1. the embodiment of the application provides a borrowing willingness estimation method based on a deep learning model, borrowing data are collected and are subjected to basic processing, structured numerical data and unstructured text data are obtained, time series elements are added into the structured numerical data and the unstructured text data, behavior characteristic information of a client is extracted, then the characteristic information training model is used for obtaining demand prediction information of the client, incremental learning is carried out on the model according to external effective borrowing data, the demand prediction information of the client output by the model is a result according with the actual situation of the client and a financial institution, time elements are added, characteristics are automatically extracted, then an effective borrowing data training model is added, and the technical effect of improving the generalization ability and the prediction accuracy of the model is achieved.
2. External variable factors are added into loan intention estimation factors through incremental learning, so that the incremental intention estimation model obtained through training can accurately represent user behavior prediction characteristics of a user in a future period of time, and the adaptability and generalization capability of the model are improved.
Example two
Based on the same inventive concept as the deep learning model-based loan intention estimation method in the foregoing embodiment, as shown in fig. 4, an embodiment of the present application provides a deep learning model-based loan intention estimation system, where the system includes:
the first obtaining unit 11: the first obtaining unit 11 is configured to obtain first loan data by performing data acquisition and proportion division on the loan data;
first input unit 12: the first input unit 12 is configured to input the first loan data into a data processing module, where the data processing module includes a numerical processing unit and a text processing unit;
the second obtaining unit 13: the second obtaining unit 13 is configured to obtain first output data according to the data processing module;
the first extraction unit 14: the first extraction unit 14 is used for inputting the first output data into a deep learning model to extract a first representative behavior feature of a user from a time sequence;
first setting unit 15: the first setting unit 15 is configured to take the first representative behavior feature as first input information;
the first processing unit 16: the first processing unit 16 is configured to perform model training by using the first input information as basic training data of a training loan intention estimation model to obtain first estimation information;
the second processing unit 17: the second processing unit 17 is configured to obtain a first valid loan variable by performing validity analysis on a loan external variable of the user;
the third processing unit 18: the third processing unit 18 is configured to perform incremental learning on the loan intention prediction model by using the first effective loan variable, and obtain second prediction information output by the loan intention prediction model.
Further, the system further comprises:
a first building unit: the first construction unit is used for constructing a model evaluation unit, wherein the model evaluation unit comprises a first evaluation unit and a second evaluation unit, the first evaluation unit is used for parameter deviation evaluation, and the second evaluation unit is used for influence deviation evaluation;
a second input unit: the second input unit is used for inputting the loan intention estimation model into the model evaluation unit to obtain first evaluation information and second evaluation information output by the model evaluation unit, wherein the first evaluation information is estimation deviation information of the first evaluation unit, and the second evaluation information is estimation deviation information of the second evaluation unit;
a first generation unit: the first generating unit is configured to generate the first estimation information according to the first evaluation information and the second evaluation information, where the first estimation information is prediction deviation information.
Further, the system further comprises:
a second building element: the second construction unit is used for constructing a first confidence interval according to the first pre-estimated information;
a first determination unit: the first determining unit is used for determining a first confidence level coefficient by carrying out confidence level analysis on the first confidence interval;
a first detection unit: the first detection unit is used for realizing abnormal detection according to the first confidence interval when the first confidence level coefficient meets a preset confidence level coefficient.
Further, the system further comprises:
a first classification unit: the first classification unit is used for classifying the first borrowing data to obtain first numerical value borrowing data and first text borrowing data;
a fourth processing unit: the fourth processing unit is used for inputting the first numerical value lending data into the numerical value processing unit for normalization processing and dissimilarity processing to obtain second numerical value lending data;
a fifth processing unit: the fifth processing unit is used for inputting the first text lending data into the text processing unit for coding processing to obtain second text lending data;
a first setting unit: the first setting unit is used for taking the second numerical value lending data and the second text lending data as output information of the data processing module.
Further, the system further comprises:
a third building element: the third construction unit is used for constructing a first sliding rule, wherein the first sliding rule is a coverage sliding logic rule;
a sixth processing unit: the sixth processing unit is configured to perform timesharing conversion on the first loan data according to the first sliding rule by the sliding window to obtain first loan conversion data corresponding to the first loan data;
a third input unit: the third input unit is configured to input the first loan conversion data into a segment detection model, so as to obtain a first detection result in the segment detection model.
Further, the system wherein the first available loan variable comprises a first predetermined variable specific gravity.
Further, the system further comprises:
a third obtaining unit: the third obtaining unit is used for obtaining first loss data according to the first effective loan variable;
a fourth input unit: the fourth input unit is used for inputting the first loss data into the loan intention estimation model for incremental learning to obtain an incremental intention estimation model, wherein the incremental intention estimation model is obtained by training a plurality of groups of data to convergence;
a fourth obtaining unit: the fourth obtaining unit is configured to obtain the second estimation information according to the incremental will estimation model.
Exemplary electronic device
The electronic device of the embodiment of the present application is described below with reference to figure 5,
based on the same inventive concept as the deep learning model-based loan intention estimation method in the foregoing embodiment, the embodiment of the present application further provides a deep learning model-based loan intention estimation system, including: a processor coupled to a memory, the memory for storing a program that, when executed by the processor, causes a system to perform the method of any of the first aspects.
The electronic device 300 includes: processor 302, communication interface 303, memory 301. Optionally, the electronic device 300 may also include a bus architecture 304. Wherein, the communication interface 303, the processor 302 and the memory 301 may be connected to each other through a bus architecture 304; the bus architecture 304 may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus architecture 304 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in FIG. 5, but this is not intended to represent only one bus or type of bus.
The communication interface 303 is a system using any transceiver or the like, and is used for communicating with other devices or communication networks, such as ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), wired access network, and the like.
The memory 301 may be, but is not limited to, a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, an electrically erasable Programmable read-only memory (EEPROM), a compact disk read-only memory (CD-ROM) or other optical disk storage, optical disk storage (including compact disk, laser disk, optical disk, digital versatile disk, blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. The memory may be self-contained and coupled to the processor through a bus architecture 304. The memory may also be integral to the processor.
The memory 301 is used for storing computer-executable instructions for executing the present application, and is controlled by the processor 302 to execute. The processor 302 is configured to execute the computer-executable instructions stored in the memory 301, so as to implement a deep learning model-based loan intention estimation method provided by the above-described embodiment of the present application.
Optionally, the computer-executable instructions in the embodiments of the present application may also be referred to as application program codes, which are not specifically limited in the embodiments of the present application.
The embodiment of the application provides a loan intention estimation method and system based on a deep learning model, and solves the technical problems that in the prior art, data characteristics need to be labeled by depending on expert experience, and due to lack of elements including time sequences, behavior trends in a future time window cannot be estimated, so that the workload of labeling the characteristics is large, and the accuracy of predicting the requirements of customers is insufficient. By collecting loan data and carrying out basic processing on the loan data to obtain structured numerical data and unstructured text data, adding time series elements into the structured numerical data and the unstructured text data to extract behavior characteristic information of a client, training a model by using the characteristic information to obtain demand prediction information of the client, carrying out incremental learning on the model according to external effective loan data, adding time elements and automatically extracting characteristics, adding an effective loan data training model, and achieving the technical effects of improving generalization capability and prediction accuracy of the model.
Those of ordinary skill in the art will understand that: the various numbers of the first, second, etc. mentioned in this application are only used for the convenience of description and are not used to limit the scope of the embodiments of this application, nor to indicate the order of precedence. "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. "at least one" means one or more. At least two means two or more. "at least one," "any," or similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one (one ) of a, b, or c, may represent: a, b, c, a-b, a-c, b-c, or a-b-c, wherein a, b, c may be single or multiple.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the application to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable system. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device including one or more available media integrated servers, data centers, and the like. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
The various illustrative logical units and circuits described in this application may be implemented or operated upon by general purpose processors, digital signal processors, Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other programmable logic systems, discrete gate or transistor logic, discrete hardware components, or any combination thereof. A general-purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing systems, e.g., a digital signal processor and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a digital signal processor core, or any other similar configuration.
The steps of a method or algorithm described in the embodiments herein may be embodied directly in hardware, in a software element executed by a processor, or in a combination of the two. The software cells may be stored in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. For example, a storage medium may be coupled to the processor such the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an ASIC, which may be disposed in a terminal. In the alternative, the processor and the storage medium may reside in different components within the terminal. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Although the present application has been described in conjunction with specific features and embodiments thereof, it will be evident that various modifications and combinations can be made thereto without departing from the spirit and scope of the application. Accordingly, the specification and figures are merely exemplary of the present application as defined in the appended claims and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations.
Claims (9)
1. A loan intention estimation method based on a deep learning model is disclosed, wherein the method comprises the following steps:
acquiring data and dividing the ratio of the loan data to obtain first loan data;
inputting the first loan data into a data processing module, wherein the data processing module comprises a numerical processing unit and a text processing unit;
obtaining first output data according to the data processing module;
inputting the first output data into a deep learning model to extract a first representative behavior feature of a user from a time sequence;
taking the first representative behavior feature as first input information;
performing model training by using the first input information as basic training data of a training loan intention estimation model to obtain first estimation information;
obtaining a first effective loan variable by analyzing the effectiveness of the external loan variable of the user;
and carrying out incremental learning on the loan intention estimation model by using the first effective loan variable to obtain second estimation information output by the loan intention estimation model.
2. The method of claim 1, wherein the model training is performed using the first input information as basic training data for training a loan intention estimation model to obtain first estimation information, and the method further comprises:
constructing a model evaluation unit, wherein the model evaluation unit comprises a first evaluation unit and a second evaluation unit, the first evaluation unit is used for parameter deviation evaluation, and the second evaluation unit is used for influence deviation evaluation;
inputting the loan intention estimation model into the model evaluation unit to obtain first evaluation information and second evaluation information output by the model evaluation unit, wherein the first evaluation information is estimation deviation information of the first evaluation unit, and the second evaluation information is estimation deviation information of the second evaluation unit;
and generating the first estimation information according to the first estimation information and the second estimation information, wherein the first estimation information is prediction deviation information.
3. The method of claim 2, the generating the first pre-estimated information based on the first and second estimated information, the method further comprising:
constructing a first confidence interval according to the first pre-estimated information;
determining a first confidence level coefficient by performing a confidence level analysis on the first confidence interval;
and when the first confidence level coefficient meets a preset confidence level coefficient, realizing abnormal detection according to the first confidence interval.
4. The method of claim 1, said inputting said first loan data into a data processing module, wherein said data processing module includes a numerical processing unit and a text processing unit, said method further comprising:
classifying the first borrowing data to obtain first numerical value borrowing data and first text borrowing data;
inputting the first numerical value lending data into the numerical value processing unit for normalization processing and dissimilarity processing to obtain second numerical value lending data;
inputting the first text lending data into the text processing unit for coding to obtain second text lending data;
and taking the second numerical value lending data and the second text lending data as output information of the data processing module.
5. The method of claim 4, further comprising:
constructing a first sliding rule, wherein the first sliding rule is a coverage sliding logic rule;
the sliding window carries out time-based conversion on the first loan data according to the first sliding rule to obtain first loan conversion data corresponding to the first loan data;
and inputting the first loan conversion data into a segment detection model to obtain a first detection result in the segment detection model.
6. The method of claim 1 wherein said first valid loan variable comprises a first predetermined variable specific gravity.
7. The method of claim 1, wherein the incremental learning of the first valid loan variable to the loan intent prediction model is performed to obtain second prediction information output by the loan intent prediction model, and the method further comprises:
obtaining first loss data according to the first effective loan variable;
inputting the first loss data into the loan intention estimation model for incremental learning to obtain an incremental intention estimation model, wherein the incremental intention estimation model is obtained by training a plurality of groups of data to convergence;
and obtaining the second estimated information according to the incremental will estimation model.
8. A loan intention estimation system based on a deep learning model, wherein the system comprises:
a first obtaining unit: the first obtaining unit is used for obtaining first loan data by carrying out data acquisition and proportion division on the loan data;
a first input unit: the first input unit is used for inputting the first loan data into a data processing module, wherein the data processing module comprises a numerical processing unit and a text processing unit;
a second obtaining unit: the second obtaining unit is used for obtaining first output data according to the data processing module;
a first extraction unit: the first extraction unit is used for inputting the first output data into a deep learning model to extract a first representative behavior feature of a user from a time sequence;
a first setting unit: the first setting unit is used for taking the first representative behavior characteristic as first input information;
a first processing unit: the first processing unit is used for performing model training by taking the first input information as basic training data of a training loan intention estimation model to obtain first estimation information;
a second processing unit: the second processing unit is used for obtaining a first effective loan variable by analyzing the effectiveness of the external loan variable of the user;
a third processing unit: the third processing unit is used for carrying out increment learning on the loan intention estimation model by the first effective loan variable to obtain second estimation information output by the loan intention estimation model.
9. A loan intention estimation system based on a deep learning model comprises: a processor coupled with a memory for storing a program that, when executed by the processor, causes a system to perform the method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111193601.5A CN113935824A (en) | 2021-10-13 | 2021-10-13 | Loan intention estimation method and system based on deep learning model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111193601.5A CN113935824A (en) | 2021-10-13 | 2021-10-13 | Loan intention estimation method and system based on deep learning model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113935824A true CN113935824A (en) | 2022-01-14 |
Family
ID=79278806
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111193601.5A Pending CN113935824A (en) | 2021-10-13 | 2021-10-13 | Loan intention estimation method and system based on deep learning model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113935824A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108053314A (en) * | 2018-01-31 | 2018-05-18 | 国信优易数据有限公司 | A kind of Loan Demand Forecasting Methodology |
| CN108182634A (en) * | 2018-01-31 | 2018-06-19 | 国信优易数据有限公司 | A kind of training method for borrowing or lending money prediction model, debt-credit Forecasting Methodology and device |
| CN110009159A (en) * | 2019-04-11 | 2019-07-12 | 湖北风口网络科技有限公司 | Financial Loan Demand prediction technique and system based on network big data |
| CN112700325A (en) * | 2021-01-08 | 2021-04-23 | 北京工业大学 | Method for predicting online credit return customers based on Stacking ensemble learning |
-
2021
- 2021-10-13 CN CN202111193601.5A patent/CN113935824A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108053314A (en) * | 2018-01-31 | 2018-05-18 | 国信优易数据有限公司 | A kind of Loan Demand Forecasting Methodology |
| CN108182634A (en) * | 2018-01-31 | 2018-06-19 | 国信优易数据有限公司 | A kind of training method for borrowing or lending money prediction model, debt-credit Forecasting Methodology and device |
| CN110009159A (en) * | 2019-04-11 | 2019-07-12 | 湖北风口网络科技有限公司 | Financial Loan Demand prediction technique and system based on network big data |
| CN112700325A (en) * | 2021-01-08 | 2021-04-23 | 北京工业大学 | Method for predicting online credit return customers based on Stacking ensemble learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108364195B (en) | User retention probability prediction method and device, prediction server and storage medium | |
| CN109190808B (en) | User behavior prediction method, device, equipment and medium | |
| CN111898578B (en) | Crowd density acquisition method and device and electronic equipment | |
| CN107679626A (en) | Machine learning method, device, system, storage medium and equipment | |
| JP2001502831A (en) | A method for classifying the statistical dependence of measurable time series | |
| CN108197795B (en) | Malicious group account identification method, device, terminal and storage medium | |
| CN115801463B (en) | Industrial Internet platform intrusion detection method and device and electronic equipment | |
| CN114428748B (en) | Simulation test method and system for real service scene | |
| CN112988840A (en) | Time series prediction method, device, equipment and storage medium | |
| CN111861667A (en) | Vehicle recommendation method and device, electronic equipment and storage medium | |
| CN114564814A (en) | Dynamic threshold Gaussian kernel density estimation system and method for sparse data | |
| CN114584601A (en) | User loss identification and intervention method, system, terminal and medium | |
| CN113553577A (en) | Unknown user malicious behavior detection method and system based on hypersphere variational automatic encoder | |
| CN110751400B (en) | Risk assessment method and device | |
| CN111783883A (en) | Abnormal data detection method and device | |
| CN111859057A (en) | Data feature processing method and data feature processing device | |
| CN113935824A (en) | Loan intention estimation method and system based on deep learning model | |
| CN111654853B (en) | Data analysis method based on user information | |
| CN114238280A (en) | Method and device for constructing financial sensitive information standard library and electronic equipment | |
| CN109670593B (en) | Method for evaluating and predicting layer calculation time in deep learning model | |
| CN115600112B (en) | Method, device, equipment and medium for obtaining behavior prediction model training set | |
| CN109711535B (en) | Method for predicting layer calculation time in deep learning model by using similar layer | |
| CN115629715B (en) | Method and system for improving accuracy of judging block types in flash memory | |
| CN112882822B (en) | Method, apparatus, device and storage medium for generating load prediction model | |
| CN115329968B (en) | Method, system and electronic equipment for determining fairness of quantum machine learning algorithm |
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 |