CN107563645A - A kind of Financial Risk Analysis method based on big data - Google Patents
A kind of Financial Risk Analysis method based on big data Download PDFInfo
- Publication number
- CN107563645A CN107563645A CN201710785884.XA CN201710785884A CN107563645A CN 107563645 A CN107563645 A CN 107563645A CN 201710785884 A CN201710785884 A CN 201710785884A CN 107563645 A CN107563645 A CN 107563645A
- Authority
- CN
- China
- Prior art keywords
- index
- data
- samples
- sample
- fields
- 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
Abstract
The invention discloses a kind of Financial Risk Analysis method based on big data, including：Build multi-level data warehouse；Carry out data prediction；The shop of storage is tracked, builds positive negative sample；Business growth ability, operation ability, the profitability risk control model in shop are built by random forests algorithm；Business growth ability index, operation ability index, Profitability Index are taken out according to random forests algorithm, index is standardized, using analytic hierarchy process (AHP) agriculture products weight, calculates the index of each index, index is summed, the risk index using the value as shop.The relevant issues occurred in result and actual conditions that risk control model is calculated are analyzed, and optimize, the model optimized is put into actual production.The present invention takes the Distributed Calculation of multi-data source convergence and big data, completes the live effect that risk model calculates after realizing modeling, and reduce financial risks by the convergence analysis of multi-data source.
Description
Technical Field
The invention belongs to the field of financial analysis, and particularly relates to a financial risk analysis method based on big data.
Background
The main defects of the current financial risk analysis method (mainly aiming at Taobao Tianmao E-commerce platform) are as follows: and (1) the calculation speed is slow. Each calculation needs several minutes even hours of calculation time, and during the period, a user can not operate the software and only waits for the completion of the calculation; (2) The multi-core characteristic of the modern CPU cannot be fully utilized, only one of the processing cores can be utilized no matter how many processing cores the computer of a user has, and the resource utilization rate is low; and (3) the data source channel is single.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a financial risk analysis method based on big data, which improves the calculation speed of a risk model, realizes the real-time calculation effect of the risk model, summarizes data of multiple data sources, and improves the accuracy of finding enterprise risks.
The purpose of the invention is realized by the following technical scheme: a financial risk analysis method based on big data comprises the following steps:
(1) Constructing a multi-level data warehouse: the method comprises the steps of obtaining data related to credit mutation from the Internet, electronic merchant data disclosed on the Internet, self-describing data and related certification materials from e-commerce customers, and data from a third-party data platform;
(2) Data preprocessing, which mainly adopts the following modes to filter:
subjective filtering: dividing data into character fields and digital fields, and eliminating useless fields according to business experience;
deletion filtration: the actual field and the processed field are not completely complete, if the missing rate is low, the field is supplemented through a missing value supplementing algorithm, and if the missing rate of the field is greater than a threshold value, the field is directly discarded;
and (3) filtering variance: eliminating fields with extremely small variance;
and (3) correlation filtering: finding out all fields with high correlation, and removing the fields with small variance;
(3) Construction of positive and negative samples: tracking warehoused shops, periodically counting tables of all fields of the shops, selecting the shops with good performances in the history record as positive samples, taking the shops close to the shops as negative samples along with the time, and periodically cleaning out a batch of sample data by the data preprocessing method in the step (2);
(4) Constructing a risk control model: and constructing a risk control model of the growth ability, the operation ability and the profitability of the shop through a random forest algorithm.
(5) The scoring mechanism is as follows: performing k-means classification on the positive and negative samples, and performing corresponding evaluation according to indexes of a certain field or certain fields; and taking out a growth capacity index, an operation capacity index and a profitability index according to a random forest algorithm, carrying out standardization processing on the indexes, determining the weight of the indexes by using an analytic hierarchy process, calculating the index of each index, summing the indexes, and taking the value as the risk index of the shop.
(6) And comparing and analyzing the calculation result of the risk control model and related problems appearing in the actual situation, optimizing, and putting the optimized model into the actual production.
Further, in the step (4), the calculation of the kini coefficients in the random forest model and the determination method of the split nodes are as follows:
let T be one sample, T = s i I = 1.. K, k is the number of samples, and the sample T contains a positive sample (a) and a negative sample (B), where the number of training samples is N (T), the number of positive samples is N (a), and the number of negative samples is N (B);
a. calculating training sample Gini coefficient Gini (T)
Gini(T)＝1-p A (T) 2 -p B (T) 2
Wherein the content of the first and second substances,representing the probability of a positive sample in the training sample T;representing the probability of a negative sample in the training sample T;
b. determining split nodes
With the set X = { X 1 ,...,X n Representing original variables and context variables of the training samples, n is the number of characteristic variables selected by the samples, and the value of each variable is X i ＝{c i1 ,...,c im Let us assume variable X i ＝c，c∈{c i1 ,...,c im Divide the samples T into two subsets T (X) i ≤c)、T(X i C), calculating Gini (T) of the classification Xi＝c )；
Wherein N (T) (Xi≤c) )、N(T (Xi＞c) ) The number of samples of subsets T (Xi ≦ c), T (Xi > c), respectively, and
Gini(T (Xi≤c) )＝1-p A (T (Xi≤c) ) 2 -p B (T (Xi≤c) ) 2 ，
Gini(T (Xi>c) )＝1-p A (T (Xi>c) ) 2 -p B (T (Xi>c) ) 2 ，
and calculating the Gini coefficients of all the variables divided on all the values, and dividing the node with the minimum Gini coefficient into the optimal split nodes.
Further, in the step (5), the weight determination method by the analytic hierarchy process is as follows:
a. subjectively determining the importance of indexes, and constructing an n-order judgment matrix;
b. and (3) solving the weight of each factor by using a root method: firstly, solving the product of each row of elements of the judgment matrix, and then solving the square root of the product of each row for n times;
c. solving and judging a characteristic vector of the matrix: (1) solving the sum of all the n-th square roots; (2) then, the quotient of the n-th power root of each line and the sum of all the n-th power roots is used for forming n elements of the characteristic vector W; multiplying the judgment matrix by the eigenvector, and performing matrix operation to obtain a matrix AW;
d. finding the maximum eigenvalue λ max ：
e. And (3) carrying out consistency check:
and the RI is an average random consistency index, when CR is less than 0.1, the consistency check is passed, and if the consistency check is not passed, the original judgment matrix is corrected until the consistency check is passed.
f. And d, obtaining an index weight, namely the weight of the feature vector in the step c.
The invention has the beneficial effects that: the invention adopts multi-data source convergence and distributed calculation of big data, namely, the distributed calculation of the risk model is carried out while modeling, the real-time effect of the risk model calculation is completed after modeling, and the financial risk is reduced through the convergence analysis of the multi-data source.
Drawings
FIG. 1 is a conventional financial risk control calculation method;
FIG. 2 is a block diagram of a financial risk control calculation method according to the present invention.
Detailed Description
The invention is described in further detail below with reference to the figures and the specific embodiments.
As shown in fig. 2, the financial risk analysis method based on big data provided by the present invention includes the following steps:
(1) And constructing a multi-level data warehouse. The method mainly comprises the steps of obtaining data related to credit mutation from the Internet, electronic merchant data disclosed on the Internet, self-describing data (mainly account data of each electronic merchant platform) and related certification materials from electronic merchant customers, and data from a third-party data platform.
(2) And (4) preprocessing data. The processed data are used for a component risk control model, and the data preprocessing mainly adopts the following modes for filtering:
subjective filtration: and dividing the data into character fields and digital fields, and removing useless fields according to business experience.
Deletion filtration: the actual field and the processed field are not completely complete, if the missing rate is low, the field can be supplemented through a missing value supplementing algorithm, but if the missing value is too large, the field is directly failed, the result is only deviated from the real result if the field is forcibly supplemented, and if the missing rate of the field is more than the value, the field is directly discarded.
And (3) filtering the variance: the core of the dimension reduction is to reduce the data columns and simultaneously ensure that the lost information quantity is as small as possible, the size of the data variance can represent the size of the information quantity, and therefore fields with extremely small variances can be eliminated.
And (3) correlation filtering: some overlap must exist between fields, so that the fields can be mutually represented, the property is called the correlation of the fields, all fields with high correlation are found, and the fields with small variance (because the information content is less) are removed.
(3) And (5) constructing positive and negative samples. The warehoused stores are tracked, tables of all fields of the stores are counted regularly (possibly every day), stores which perform well in the history are selected as positive samples, and stores which go to stores with the passage of time are selected as negative samples. And (3) cleaning a batch of sample data periodically (every day) by the data preprocessing method in the step (2).
(4) Screening an algorithm and constructing a risk control model. Firstly, considering in a classical data mining classification algorithm, because fields obtained after primary screening are too many and are difficult to calculate by classification algorithms such as KNN and Bayes, and an ID3 decision tree algorithm and a C5.0 decision tree, a random forest algorithm and a boosted decision tree (GDBT) based on the ID3 decision tree have good effects on classification and screening of important indexes, the test is tried by the decision tree, the random forest and the boosted decision tree algorithm, compared with the C5.0 algorithm, the final result is that the accuracy of a test set of the random forest and the GDBT is much higher, and the GDBT consumes much more time than the random forest algorithm, so the random forest algorithm is adopted. And constructing a risk control model of the growth capacity, the operation capacity and the profit capacity of the shop through a random forest algorithm. And performing parallel calculation on the risk control models through Spark, and obtaining a calculation result and an analysis result of each risk control model.
Calculating a kini coefficient in the random forest model and determining split nodes:
let T be one sample, T = s i I = 1.. K, k is the number of samples, and the sample T contains a positive sample (a) and a negative sample (B), where the number of training samples is N (T), the number of positive samples is N (a), and the number of negative samples is N (B).
a. Calculating training sample Gini coefficient (T)
Gini(T)＝1-p A (T) 2 -p B (T) 2
WhereinRepresenting the probability of a positive sample in the training sample T;
representing the probability of a negative sample in the training sample T.
b. Determining split nodes
With set X = { X 1 ,...,X n Representing original variables and context variables of the training samples, n is the number of characteristic variables selected by the samples, and the value of each variable is X i ＝{c i1 ,...,c im Let us assume the variable X i ＝c，c∈{c i1 ,...,c im Divide the samples T into two subsets T (X) i ≤c)、T(X i C), calculating Gini (T) of the Gini coefficient of the division Xi＝c )。
Wherein N (T) (Xi≤c) )、N(T (Xi＞c) ) The number of samples of subsets T (Xi ≦ c), T (Xi > c), respectively, and
Gini(T (Xi≤c) )＝1-p A (T (Xi≤c) ) 2 -p B (T (Xi≤c) ) 2 ，
Gini(T (Xi>c) )＝1-p A (T (Xi>c) ) 2 -p B (T (Xi>c) ) 2 。
and calculating the Gini coefficients of all the variables divided on all the values, and dividing the node with the minimum Gini coefficient into the optimal split nodes.
(5) A scoring mechanism. Firstly, performing k-means classification on positive and negative samples, and performing corresponding evaluation according to indexes of a certain field or certain fields; first, considering the evaluation of one index, tests have found that there are some special cases, and we have not considered as thorough as possible in terms of one index. The risk indexes of the shops need to be considered together from a plurality of indexes, the importance degrees of different indexes are different, the growth capacity index, the operation capacity index and the profitability index are extracted according to the random forest algorithm, the indexes are subjected to corresponding standardization processing, the index weight is determined by using an analytic hierarchy process, the index of each index can be calculated, the indexes are summed, and the value is used as the risk index of the shops.
Determining the weight by an analytic hierarchy process:
a. subjectively determining the importance of the indexes, and constructing a judgment matrix as shown in the following table:
| A | C1 | C2 | C3 |
| C1 | 1 | 1/5 | 1/3 |
| C2 | 5 | 1 | 3 |
| C3 | 3 | 1/3 | 1 |
b. and solving the weight value of each factor by using a square root method. The product of each row of elements is first found, and after finding the product of each row of elements, the product of each row is then found to be the square root of the order n, for example, for a 3-order matrix, the square root of the order 3 of each product is required. The following table shows the product and the value of the n-th root.
| Product of products | Root of cubic boron | |
| First row | 0.066667 | 0.4054801 |
| Second row | 15 | 2.4662121 |
| Third row | 1 | 1 |
c. And solving the characteristic vector of the judgment matrix. (1) Obtaining the sum of 3 n-th-order square roots as 3.8716922; (2) the quotient of the sum of the n-th root and the 3 n-th roots of each row is then used, the 3 quotients constituting the 3 elements of the feature vector. The feature vector calculated in this example is (W) 1 ＝0.10472943,W 2 ＝0.636986,W 3 ＝0.258285)t。
Judging the matrix to multiply the eigenvector, and performing matrix operation to obtain a matrix AW = ((AW) 1 ,(AW) 2 ,(AW) 3 ). As follows:
d. maximum eigenvalue is found: the maximum characteristic value lambda is determined by the following formula max Is 3.038511.
e. And (3) carrying out consistency check: CI is the quotient of (maximum eigenvalue-n)/n-1, and in this example, CI is 0.019256, and when n is 3, RI is 0.58 by table lookup. CR is the ratio of CI to RI. In this example, CR is 0.033199 is less than 0.1, so the consistency check passes in this example, and if the consistency check does not pass, the original decision matrix must be modified until the consistency check passes.
CI =0.19256, and RI =0.58 when n =3
CR＝0.33199<0.1
c. Obtaining the index weight in this example: PC1=0.10472943, pc2=0.63698, pc3=0.258285, which is the weight of the feature vector in step c.
(6) And comparing and analyzing the calculation result of the risk control model and related problems appearing in the actual situation, optimizing, and putting the optimized model into the actual production.
Claims (3)
1. A financial risk analysis method based on big data is characterized by comprising the following steps:
(1) Constructing a multi-level data warehouse: the method comprises the steps of obtaining data related to credit mutation from the Internet, electronic merchant data disclosed on the Internet, self-describing data and relevant certification materials from an e-commerce customer, and data from a third-party data platform;
(2) Data preprocessing, which mainly adopts the following modes to filter:
subjective filtration: dividing data into character fields and digital fields, and eliminating useless fields according to business experience;
deletion filtration: the actual field and the processed field are not completely complete, if the missing rate is low, the field is supplemented through a missing value supplementing algorithm, and if the missing rate of the field is greater than a threshold value, the field is directly discarded;
and (3) filtering the variance: eliminating fields with extremely small variance;
and (3) correlation filtering: finding out all fields with high correlation, and removing the fields with small variance;
(3) Construction of positive and negative samples: tracking warehoused shops, periodically counting tables of all fields of the shops, selecting the shops with good performances in the history record as positive samples, taking the shops close to the shops as negative samples along with the time, and periodically cleaning out a batch of sample data by the data preprocessing method in the step (2);
(4) Constructing a risk control model: and constructing a risk control model of the growth ability, the operation ability and the profitability of the shop through a random forest algorithm.
(5) The scoring mechanism is as follows: performing k-means classification on the positive and negative samples, and performing corresponding evaluation according to the indexes of a certain field or certain fields; and taking out a growth capacity index, an operation capacity index and a profitability index according to a random forest algorithm, carrying out standardization processing on the indexes, determining the weight of the indexes by using an analytic hierarchy process, calculating the index of each index, summing the indexes, and taking the value as the risk index of the shop.
(6) And comparing and analyzing the calculation result of the risk control model and related problems appearing in the actual situation, optimizing, and putting the optimized model into the actual production.
2. The big data-based financial risk analysis method according to claim 1, wherein in the step (4), the calculation of the kini coefficients in the random forest model and the determination of the split nodes are as follows:
let T be one sample, T = s i I = 1.. K, k is the number of samples, and the sample T contains a positive sample (a) and a negative sample (B), where the number of training samples is N (T), the number of positive samples is N (a), and the number of negative samples is N (B);
a. calculating training sample Gini coefficient (T)
Gini(T)＝1-p A (T) 2 -p B (T) 2
Wherein the content of the first and second substances,representing the probability of a positive sample in the training sample T;representing the probability of a negative sample in the training sample T;
b. determining split nodes
With the set X = { X 1 ,...,X n Representing original variables and context variables of the training samples, n is the number of characteristic variables selected by the samples, and the value of each variable is X i ＝{c i1 ,...,c im Let us assume variable X i ＝c，c∈{c i1 ,...,c im Divide the sample T into two subsets T (X) i ≤c)、T(X i C), calculating Gini (T) of the classification Xi＝c )；
Wherein N (T) (Xi≤c) )、N(T (Xi＞c) ) The number of samples of subsets T (Xi ≦ c), T (Xi > c), respectively, and
Gini(T (Xi≤c) )＝1-p A (T (Xi≤c) ) 2 -p B (T (Xi≤c) ) 2 ，
Gini(T (Xi>c) )＝1-p A (T (Xi>c) ) 2 -p B (T (Xi>c) ) 2 ，
and calculating the Gini coefficients of all the variables divided on all the values, and dividing the node with the minimum Gini coefficient into the optimal split nodes.
3. The big data-based financial risk analysis method according to claim 1, wherein in the step (5), the analytic hierarchy process determines the weight by the following method:
a. subjectively determining the importance of indexes, and constructing an n-order judgment matrix;
b. and (3) solving the weight of each factor by using a root method: firstly, solving the product of each row of elements of the judgment matrix, and then solving the square root of the product of each row for n times;
c. solving and judging the characteristic vector of the matrix: (1) solving the sum of all the n-th-order square roots; (2) then, the quotient of the n-th power root of each line and the sum of all the n-th power roots is used for forming n elements of the characteristic vector W; multiplying the judgment matrix by the eigenvector, and performing matrix operation to obtain a matrix AW;
d. finding the maximum eigenvalue λ max ：
e. And (3) carrying out consistency check:
and the RI is an average random consistency index, when CR is less than 0.1, the consistency check is passed, and if the consistency check is not passed, the original judgment matrix is corrected until the consistency check is passed.
f. And d, obtaining an index weight, namely the weight of the feature vector in the step c.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710785884.XA CN107563645A (en) | 2017-09-04 | 2017-09-04 | A kind of Financial Risk Analysis method based on big data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710785884.XA CN107563645A (en) | 2017-09-04 | 2017-09-04 | A kind of Financial Risk Analysis method based on big data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107563645A true CN107563645A (en) | 2018-01-09 |
Family
ID=60979043
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710785884.XA Pending CN107563645A (en) | 2017-09-04 | 2017-09-04 | A kind of Financial Risk Analysis method based on big data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107563645A (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108198075A (en) * | 2018-03-28 | 2018-06-22 | 东莞市波动赢机器人科技有限公司 | Robotic training method based on autonomous learning |
| CN108492135A (en) * | 2018-03-08 | 2018-09-04 | 深圳萨摩耶互联网金融服务有限公司 | The tracking optimization method and tracking optimization system of channel port cost |
| CN108846532A (en) * | 2018-03-21 | 2018-11-20 | 宁波工程学院 | Business risk appraisal procedure and device applied to logistics supply platform chain |
| CN109359745A (en) * | 2018-09-13 | 2019-02-19 | 佛山储钱罐信息咨询服务有限公司 | A kind of system based on intelligent algorithm financial product management backstage intelligence O&M |
| CN109783537A (en) * | 2018-12-26 | 2019-05-21 | 阚研佳 | A kind of financial assessment system |
| CN109951859A (en) * | 2019-03-25 | 2019-06-28 | 腾讯科技（深圳）有限公司 | Wireless network connection recommended method, device, electronic equipment and readable medium |
| CN110634060A (en) * | 2018-06-21 | 2019-12-31 | 马上消费金融股份有限公司 | User credit risk assessment method, system, device and storage medium |
| CN111507649A (en) * | 2020-06-30 | 2020-08-07 | 南昌木本医疗科技有限公司 | Financial big data wind control platform based on block chain |
| CN111986026A (en) * | 2020-09-07 | 2020-11-24 | 中国银行股份有限公司 | Personal credit business risk assessment method and device based on analytic hierarchy process |
| CN112016770A (en) * | 2020-10-21 | 2020-12-01 | 平安科技（深圳）有限公司 | Medical insurance expense prediction method, device, equipment and storage medium |
| CN112232962A (en) * | 2020-11-19 | 2021-01-15 | 深圳前海微众银行股份有限公司 | Transaction index processing method, device and equipment |
| WO2021174693A1 (en) * | 2020-03-05 | 2021-09-10 | 平安科技（深圳）有限公司 | Data analysis method and apparatus, and computer system and readable storage medium |
| WO2021232588A1 (en) * | 2020-05-21 | 2021-11-25 | 平安国际智慧城市科技股份有限公司 | Food safety risk assessment method, apparatus, device, and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102402732A (en) * | 2010-09-14 | 2012-04-04 | 中国船舶工业综合技术经济研究院 | Method and system for evaluating scientific research projects |
| CN102436622A (en) * | 2011-12-28 | 2012-05-02 | 浙江汇信科技有限公司 | Method for evaluating network market operator credit status |
| CN105279691A (en) * | 2014-07-25 | 2016-01-27 | 中国银联股份有限公司 | Financial transaction detection method and equipment based on random forest model |
| CN105447634A (en) * | 2015-11-17 | 2016-03-30 | 苏州蜗牛数字科技股份有限公司 | Real-time risk control method based on big data platform and system thereof |
| CN105913195A (en) * | 2016-04-29 | 2016-08-31 | 浙江汇信科技有限公司 | All-industry data based enterprise's financial risk scoring method |
-
2017
- 2017-09-04 CN CN201710785884.XA patent/CN107563645A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102402732A (en) * | 2010-09-14 | 2012-04-04 | 中国船舶工业综合技术经济研究院 | Method and system for evaluating scientific research projects |
| CN102436622A (en) * | 2011-12-28 | 2012-05-02 | 浙江汇信科技有限公司 | Method for evaluating network market operator credit status |
| CN105279691A (en) * | 2014-07-25 | 2016-01-27 | 中国银联股份有限公司 | Financial transaction detection method and equipment based on random forest model |
| CN105447634A (en) * | 2015-11-17 | 2016-03-30 | 苏州蜗牛数字科技股份有限公司 | Real-time risk control method based on big data platform and system thereof |
| CN105913195A (en) * | 2016-04-29 | 2016-08-31 | 浙江汇信科技有限公司 | All-industry data based enterprise's financial risk scoring method |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108492135A (en) * | 2018-03-08 | 2018-09-04 | 深圳萨摩耶互联网金融服务有限公司 | The tracking optimization method and tracking optimization system of channel port cost |
| CN108492135B (en) * | 2018-03-08 | 2020-03-10 | 深圳萨摩耶互联网金融服务有限公司 | Tracking optimization method and tracking optimization system for channel port cost |
| CN108846532A (en) * | 2018-03-21 | 2018-11-20 | 宁波工程学院 | Business risk appraisal procedure and device applied to logistics supply platform chain |
| CN108198075A (en) * | 2018-03-28 | 2018-06-22 | 东莞市波动赢机器人科技有限公司 | Robotic training method based on autonomous learning |
| CN110634060A (en) * | 2018-06-21 | 2019-12-31 | 马上消费金融股份有限公司 | User credit risk assessment method, system, device and storage medium |
| CN109359745A (en) * | 2018-09-13 | 2019-02-19 | 佛山储钱罐信息咨询服务有限公司 | A kind of system based on intelligent algorithm financial product management backstage intelligence O&M |
| CN109783537A (en) * | 2018-12-26 | 2019-05-21 | 阚研佳 | A kind of financial assessment system |
| CN109951859A (en) * | 2019-03-25 | 2019-06-28 | 腾讯科技（深圳）有限公司 | Wireless network connection recommended method, device, electronic equipment and readable medium |
| CN109951859B (en) * | 2019-03-25 | 2021-08-20 | 腾讯科技（深圳）有限公司 | Wireless network connection recommendation method and device, electronic equipment and readable medium |
| WO2021174693A1 (en) * | 2020-03-05 | 2021-09-10 | 平安科技（深圳）有限公司 | Data analysis method and apparatus, and computer system and readable storage medium |
| WO2021232588A1 (en) * | 2020-05-21 | 2021-11-25 | 平安国际智慧城市科技股份有限公司 | Food safety risk assessment method, apparatus, device, and storage medium |
| CN111507649A (en) * | 2020-06-30 | 2020-08-07 | 南昌木本医疗科技有限公司 | Financial big data wind control platform based on block chain |
| CN111986026A (en) * | 2020-09-07 | 2020-11-24 | 中国银行股份有限公司 | Personal credit business risk assessment method and device based on analytic hierarchy process |
| CN112016770A (en) * | 2020-10-21 | 2020-12-01 | 平安科技（深圳）有限公司 | Medical insurance expense prediction method, device, equipment and storage medium |
| CN112232962A (en) * | 2020-11-19 | 2021-01-15 | 深圳前海微众银行股份有限公司 | Transaction index processing method, device and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107563645A (en) | A kind of Financial Risk Analysis method based on big data | |
| CN108564286B (en) | Artificial intelligent financial wind-control credit assessment method and system based on big data credit investigation | |
| CN112070125A (en) | Prediction method of unbalanced data set based on isolated forest learning | |
| CN110956273A (en) | Credit scoring method and system integrating multiple machine learning models | |
| CN108898476A (en) | A kind of loan customer credit-graded approach and device | |
| CN112232930A (en) | E-commerce platform customer segmentation method based on weighted RFM model | |
| CN112561568A (en) | Target customer prediction method, device and storage medium | |
| CN113111924A (en) | Electric power customer classification method and device | |
| CN115115265A (en) | RFM model-based consumer evaluation method, device and medium | |
| Casarin et al. | Dynamic predictive density combinations for large data sets in economics and finance | |
| CN116468536A (en) | Automatic risk control rule generation method | |
| CN115983622A (en) | Risk early warning method of internal control cooperative management system | |
| CN114298659A (en) | Data processing method and device for evaluation object index and computer equipment | |
| CN116091206B (en) | Credit evaluation method, credit evaluation device, electronic equipment and storage medium | |
| CN116468273A (en) | Customer risk identification method and device | |
| CN116167629A (en) | Distributed policy automatic generation method, system and storage medium | |
| CN115860924A (en) | Supply chain financial credit risk early warning method and related equipment | |
| CN115660720A (en) | Cigarette sales prediction method and equipment | |
| CN110955811B (en) | Power data classification method and system based on naive Bayes algorithm | |
| CN114820074A (en) | Target user group prediction model construction method based on machine learning | |
| CN113077271A (en) | Enterprise credit rating method and device based on BP neural network | |
| CN112070336A (en) | Manufacturing industry information quantitative analysis method and device based on analytic hierarchy process | |
| CN111339294A (en) | Client data classification method and device and electronic equipment | |
| CN104794235B (en) | Financial time series are segmented distribution characteristics computational methods and system | |
| CN113516388A (en) | Innovative development situation evaluation method of weighted principal component distance clustering based on entropy weight |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180109 |
|
| RJ01 | Rejection of invention patent application after publication |