CN105590156B - Detection method of high-risk bank card and data processing device - Google Patents
Detection method of high-risk bank card and data processing device Download PDFInfo
- Publication number
- CN105590156B CN105590156B CN201410686072.6A CN201410686072A CN105590156B CN 105590156 B CN105590156 B CN 105590156B CN 201410686072 A CN201410686072 A CN 201410686072A CN 105590156 B CN105590156 B CN 105590156B
- Authority
- CN
- China
- Prior art keywords
- data
- card
- dimension
- bank card
- end card
- 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.)
- Active
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses a method for detecting a high risk of a bank card, comprising the following steps: clustering historical data of various bank card transactions by using a K-means method so as to obtain a risk model, wherein the historical data is divided into two types of high-end cards and non-high-end cards as training data of the risk model, and each type of information data is represented by n dimensions; preprocessing transaction data of a bank card to be detected into data with the same dimensionality as training data of the risk model; determining, from the risk model, whether the preprocessed data complies with rules and characteristics that vary from the non-high-end card to the high-end card; and if not, determining that the bank card has a high risk. The application also discloses a data processing device.
Description
Technical Field
The invention relates to a data mining technology, in particular to a detection method of a high-risk bank card and a data processing device.
Background
At present, various types of bank cards are released by various large commercial banks. While obtaining intermediate business benefits and binding high-end customers through the release of various products, banks gradually realize that bank cards are an industry with intensive funds, intensive technology and intensive labor force, and the risks existing in the business are not negligible.
The general bank card business risks include two types of credit risks and fraud risks. Credit risk is created by lack of warranty, circulating credit limits, approval of applicant limits loose, and the like. And the fraud risk which accounts for more than 90% of the risk loss of the whole bank card business generally refers to counterfeit transactions, counterfeit card fraud, loss/theft, mail order/e-purchase fraud and the like. Currently, the worldwide annual losses due to fraud risk exceed two billion dollars (this figure is still growing). In addition, the fraudulent transactions are hidden easily in a large amount of valid transaction data, and statistically, 1 fraudulent transaction may occur in every 13000 transactions on average. Therefore, in order to combat fraud, the construction and implementation of fraud risk prevention systems are the core of risk control for bankcard transactions.
Under the existing risk control system of the bank card industry, the control of the bank card risk is generally realized by monitoring and detecting the bank card transaction, namely, for each transaction, whether the bank card transaction is a transaction with fraud suspicion or potential risk is discovered through a series of technologies and means. However, on one hand, controlling the risk through the transaction itself needs to face the large-scale bank card transaction amount, tens of millions of bank card transactions occur every day, and for the tens of millions of transactions, each transaction judges the risk and fraud attributes, and the risk and fraud attributes are huge burdens for both background systems and foreground business personnel. On the other hand, the risks related to the bank card transactions are various, and the methods for detecting each risk behavior are different, so that misjudgment and misjudgment are easily caused aiming at the risk detection of the bank card transactions.
Based on the practical difficult problem of bank card transaction risk control, some methods are assisted by transaction risk control, for example, some high-risk bank cards are focused during risk control, or some high-risk merchants are focused to reduce the risk control range and reduce erroneous judgment and missed judgment. However, the high-risk bank card and the high-risk merchant concerned by the current bank card transaction management are usually based on a blacklist system, that is, the bank card and the merchant who have the high-risk behavior are taken as key observation objects, but obviously many high-risk bank cards and merchants which should be focused on are omitted.
Disclosure of Invention
In order to solve the above problems, the present application provides a method of detecting that a bank card has a high risk, the method comprising: clustering historical data of various bank card transactions by using a K-means method so as to obtain a risk model, wherein the historical data is divided into two types of high-end cards and non-high-end cards as training data of the risk model, and each type of information data is represented by n dimensions; preprocessing transaction data of a bank card to be detected into data with the same dimensionality as training data of the risk model; determining, from the risk model, whether the preprocessed data complies with rules and characteristics that vary from the non-high-end card to the high-end card; and if not, determining that the bank card has a high risk.
In the above method, determining whether the preprocessed data complies with rules and characteristics that change from the non-high end card to the high end card based on the risk model comprises: for the preprocessed data N, the angle of inclusion β is calculated according to the following equation:
determining whether the included angle beta is within a first threshold range; and C is the central point of the non-high-end card cluster determined by the risk model, D is the central point of the high-end card cluster determined by the risk model, and the included angle beta represents the included angle between the vector extending from the center of the non-high-end card to the bank card represented by the preprocessed data and the vector extending from the center of the non-high-end card to the center of the high-end card.
In the above method, the clustering the historical data of various bank card transactions by using the K-means method to obtain the risk model includes: (a) collecting historical data of various bank card transactions, and preprocessing the historical data into n-dimensional data according to item classification; (b) performing two types of processing on the historical data expressed by the n dimension by using a K-means algorithm so as to obtain central points C and D of the two types of clusters, wherein C is the central point of a non-high-end card cluster, and D is the central point of a high-end card cluster, and seed nodes of the two types of clusters are set as a geometric center A point of high-end card information data and a geometric center B point of the non-high-end card information data; (c) evaluating the clustering result according to the included angle alpha between the vector AB and the vector CD; and (d) if the included angle α is greater than a second threshold, re-executing steps (b) and (c) and adjusting the weight of each dimension until α is less than or equal to the second threshold.
In the above method, the angle α is calculated according to the following formula:
in the above method, the adjustment is carried outDetermining the value v of each dimension of the vector AB during the weighting of each dimensioniValue u of the corresponding dimension to the vector CDiRatio v ofi/uiAnd calculating the mean of the ratio in each dimension according to the following formula:
In the above process, at vi/uiIf the value of (d) is greater than the average ave, the data weight of the ith dimension is increased, otherwise, the data weight of the ith dimension is reduced.
According to another aspect of the present application, there is provided a data processing apparatus, the apparatus comprising: a first unit configured to cluster historical data of various bank card transactions using a K-means method so as to obtain a risk model, the historical data being classified into two types of high-end cards and non-high-end cards as training data of the risk model, and each type of information data being represented in n-dimensions; a second unit configured to preprocess transaction data of a bank card to be detected into data having the same dimension as the training data of the risk model; a third unit configured to determine, according to the risk model, whether the preprocessed data complies with rules and characteristics that change from the non-high-end card to the high-end card; and a fourth unit configured to identify the bank card as having a high risk when it is determined that the preprocessed data do not comply with rules and characteristics that change from the non-high-end card to the high-end card.
In the above apparatus, the third unit is configured to calculate, for the preprocessed data N, an included angle β according to:
and the third unit is further configured to determine whether the included angle β is within a first threshold range; wherein C is the windAnd D is the central point of the high-end card cluster determined by the risk model, and the included angle beta represents the included angle between the vector extending from the center of the non-high-end card to the bank card represented by the preprocessed data and the vector extending from the center of the non-high-end card to the center of the high-end card.
In the above apparatus, the first unit is configured to perform the steps of: (a) collecting historical data of various bank card transactions, and preprocessing the historical data into n-dimensional data according to item classification; (b) performing two types of processing on the historical data expressed by the n dimension by using a K-means algorithm so as to obtain central points C and D of the two types of clusters, wherein C is the central point of a non-high-end card cluster, and D is the central point of a high-end card cluster, and seed nodes of the two types of clusters are set as a geometric center A point of high-end card information data and a geometric center B point of the non-high-end card information data; (c) evaluating the clustering result according to the included angle alpha between the vector AB and the vector CD; and (d) if the included angle α is greater than a second threshold, re-executing steps (b) and (c) and adjusting the weight of each dimension until α is less than or equal to the second threshold.
In the above apparatus, the first unit is configured to calculate the included angle α according to the following formula:
in the above apparatus, the first unit is configured to determine the value v of each dimension of the vector AB in adjusting the weight of each dimensioniValue u of the corresponding dimension to the vector CDiRatio v ofi/uiAnd calculating the mean of the ratio in each dimension according to the following formula:
In the above apparatus, the first unit is configured to be at vi/uiIs large in valueAnd increasing the data weight of the ith dimension when the average value ave is obtained, and otherwise, reducing the data weight of the ith dimension.
The technical scheme of the invention provides a technology for mining a bank card with high risk by utilizing bank card transaction data, which can comprehensively process a plurality of semantic data describing the behavior of the bank card, adopts a data mining method taking a clustering method as a core, but finds out the trend of the change and development of the bank card data by clustering instead of simply dividing the data into a plurality of classes, thereby finding out the bank card deviating from the trend, namely the bank card with risk (such as fraud and cash register behaviors). The comprehensive data mining method enables the investigation of the behavior of the bank card to be more comprehensive and the obtained result to be more accurate.
Drawings
The various aspects of the present invention will become more apparent to those of ordinary skill in the art after reading the detailed description of the invention in light of the accompanying drawings. Those skilled in the art will understand that: these drawings are only for the purpose of illustrating the technical solutions of the present invention in connection with the embodiments and are not intended to limit the scope of the present invention.
Fig. 1 is a schematic diagram of a detection method of a high-risk bank card based on data mining according to an embodiment of the present application.
Detailed Description
The following description is of some of the many possible embodiments of the invention and is intended to provide a basic understanding of the invention and is not intended to identify key or critical elements of the invention or to delineate the scope of the invention. It is easily understood that according to the technical solution of the present invention, other implementations that can be substituted with each other can be suggested by those skilled in the art without changing the spirit of the present invention. Therefore, the following detailed description and the accompanying drawings are merely illustrative of the technical aspects of the present invention, and should not be construed as all of the present invention or as limitations or limitations on the technical aspects of the present invention.
Fig. 1 is a schematic diagram of a detection method of a high-risk bank card based on data mining according to an embodiment of the present application. As shown in fig. 1, a large amount of bank card data needs to be input first for training the risk model. Then, the input data is preprocessed, for example, redundant information and missing information in the input data are removed, and the data are represented by n dimensions defined according to the number of semantics describing the behavior of the bank card. And then, carrying out clustering analysis on the preprocessed data, and evaluating the clustering result. And if the evaluation result meets the preset requirement, then carrying out risk analysis on the bank card to be detected. Otherwise, adjusting the weight in the risk model, and performing clustering analysis again until the evaluation result meets the preset requirement. Then, a risk determination is made. In one specific implementation, whether the bank card has the risk is determined by determining whether the data of the bank card to be detected conforms to the rules and characteristics of the change from the non-high-end card to the high-end card according to the obtained risk model. If the rule of changing from a non-high-end card to a high-end card is not followed, it can be determined that the bank card has a high risk.
In one particular embodiment, the training of the risk model may consist of five steps: data preparation, data clustering, clustering result evaluation, weight adjustment and risk judgment. The following description is made for these five steps.
1. And (4) preparing data. The data preparation is divided into two parts of data collection and data preprocessing, card information is required to be obtained from consumption records in the data collection part, the card information is represented by a vector, each digit of the vector represents different meanings capable of describing consumption behaviors of the bank card and is represented by numerical values such as consumption amount, consumption times, consumption time and the like of the bank card, the dimensionality of the card information vector can be customized and expanded at will, namely, consumption behaviors of the bank card can be described by adopting any number of indexes, and the bank card described by the card information must comprise two categories of a high-end card (such as a platinum card and a diamond card) and a non-high-end card (such as a common card); in the 'data preprocessing' part, card information data are subjected to normalization processing, and redundant information and missing information are removed.
2. And (6) clustering data. Performing class-2 processing on the prepared card information data by using a Kmeans algorithm, wherein seed nodes of the class-2 processing are respectively a geometric center (point A) of high-end card information data and a geometric center (point B) of non-high-end card information data, the distance used by the clustering is a weighted Euclidean distance, and the weight of each dimension of the bank card information is the same under the initial condition.
3. And evaluating a clustering result. For two classes in the class 2 results, their class centers are denoted by C and D, respectively, and the angle α between the vector AB and the CD is calculated:
a threshold value (e.g., pi/6, pi/8, 0, etc.) is set for alpha, if alpha is greater than the threshold value, the "weight adjustment" step is entered, otherwise the "risk determination" step is entered, and the coordinates and the respective dimensional weights of the alpha are saved C, D.
4. And (5) adjusting the weight. Value v for each dimension of vector ABiThen, the value u of the dimension corresponding to the vector CD is calculatediThe ratio of (A) to (B): v. ofi/uiAnd the mean of the ratio in each dimension:
wherein D is the dimension number of the bank card data, if vi/uiIf the value of (d) is greater than ave, increasing the data weight of the ith dimension, otherwise, reducing the data weight of the ith dimension, and returning to the step of data clustering to perform clustering again.
5. And (6) risk judgment. Calculating an included angle beta between CN and CD for each clustered bank card data point N:
a threshold value (e.g., pi/3, 2 pi/5, etc.) is set for beta, and if beta is greater than the threshold value, the bank card described by the piece of data can be determined to be a high-risk bank card.
When the obtained risk model is used for risk judgment, any bank card data is processed into data with the same dimension as that of model training data, each dimension has the same meaning, normalization processing is carried out on the data, the stored coordinates of C, D points and the weights of all dimensions are adjusted according to the weight in the fourth step of the risk model training stage, the operation which is the same as that of the risk judgment in the fifth step of the model training stage is carried out, and therefore whether the bank card data belongs to a high-risk bank card or not can be judged.
According to the technical scheme, the used bank card data is set to be composed of a plurality of dimensions when the model is trained according to the types of data used for describing transaction behavior characteristics in the bank card transaction data, and each dimension can be regarded as the reflection of different types of bank card risk behaviors on a data layer. Model training using multidimensional data allows as many types of risky bank cards as possible to be detected.
In addition, the clustering method is not used for simply clustering data into a plurality of classes, but is used for searching the change characteristics reflected by the data. The invention clusters aiming at the semantic characteristics of ' high-end cards ' and ' low-end cards ', and continuously adjusts the weight of each data dimension in the repeated clustering process based on the recognition that the change of the bank cards from the non-high-end cards to the high-end cards follows a certain rule and characteristic ', so that the characteristic of the important dimension is highlighted, and the characteristic of the non-important dimension is shielded, so that the rule of the data change development of the bank cards is found out.
Due to the recognition that the bank card changes from the non-high-end card to the high-end card according to certain rules and characteristics, the card which does not comply with the change characteristics can be considered to have potential risks or certain suspected fraud, and therefore the technical scheme of the application judges the bank card deviating from the change characteristics of the card data as the bank card at risk in the risk judgment stage.
Compared with the traditional blacklist method, the high-risk bank card found by the technical scheme is more comprehensive, the risk of the bank card can be judged by calculating the vector included angle when the model is applied, the calculation cost is low, and the application efficiency is high.
Hereinbefore, specific embodiments of the present invention are described with reference to the drawings. However, those skilled in the art will appreciate that various modifications and substitutions can be made to the specific embodiments of the present invention without departing from the spirit and scope of the invention. Such modifications and substitutions are intended to be included within the scope of the present invention as defined by the appended claims.
Claims (10)
1. A computer-implemented based method of detecting that a bank card has a high risk, the method comprising:
clustering historical data of various bank card transactions by using a K-means method so as to obtain a risk model, wherein the historical data is divided into two types of high-end cards and non-high-end cards as training data of the risk model, and each type of information data is represented by n dimensions;
preprocessing transaction data of a bank card to be detected into data with the same dimensionality as training data of the risk model;
determining, from the risk model, whether the preprocessed data complies with rules and characteristics that vary from the non-high-end card to the high-end card; and
if not, determining that the bank card has a high risk,
wherein determining whether the preprocessed data complies with rules and characteristics that change from the non-high-end card to the high-end card based on the risk model comprises:
for the preprocessed data N, the angle of inclusion β is calculated according to the following equation:
determining whether the included angle beta is within a first threshold range;
and C is the central point of the non-high-end card cluster, D is the central point of the high-end card cluster, and the included angle beta represents the included angle between the vector extending from the center of the non-high-end card to the bank card represented by the preprocessed data and the vector extending from the center of the non-high-end card to the center of the high-end card.
2. The method of claim 1, wherein the clustering historical data of various bank card transactions using the K-means method to obtain a risk model comprises:
(a) collecting historical data of various bank card transactions, and preprocessing the historical data into n-dimensional data according to item classification;
(b) performing two types of processing on the historical data expressed by the n dimension by using a K-means algorithm so as to obtain central points C and D of the two types of clusters, wherein C is the central point of a non-high-end card cluster, and D is the central point of a high-end card cluster, and seed nodes of the two types of clusters are set as a geometric center A point of high-end card information data and a geometric center B point of the non-high-end card information data;
(c) evaluating the clustering result according to the included angle alpha between the vector AB and the vector CD; and
(d) if the included angle α is greater than a second threshold, then steps (b) and (c) are re-executed and the weight of each dimension is adjusted until α is less than or equal to the second threshold.
4. the method of claim 2, wherein the value v of each dimension of the vector AB is determined during the adjustment of the weight of each dimensioniValue u of the corresponding dimension to the vector CDiRatio v ofi/uiAnd according toThe average of this ratio in each dimension is calculated by:
where d is the number of dimensions of the bank card data.
5. The method of claim 4, wherein at v, v isi/uiIf the value of (d) is greater than the average ave, the data weight of the ith dimension is increased, otherwise, the data weight of the ith dimension is reduced.
6. A computer-implemented based data processing apparatus, the apparatus comprising:
a first unit configured to cluster historical data of various bank card transactions using a K-means method so as to obtain a risk model, the historical data being classified into two types of high-end cards and non-high-end cards as training data of the risk model, and each type of information data being represented in n-dimensions;
a second unit configured to preprocess transaction data of a bank card to be detected into data having the same dimension as the training data of the risk model;
a third unit configured to determine, according to the risk model, whether the preprocessed data complies with rules and characteristics that change from the non-high-end card to the high-end card; and
a fourth unit configured to identify the bank card as having a high risk when it is determined that the pre-processed data does not comply with laws and characteristics that change from the non-high end card to the high end card,
wherein the third unit is configured to calculate, for the preprocessed data N, an angle β according to:
the third unit is further configured to determine whether the included angle β is within a first threshold range;
and C is the central point of the non-high-end card cluster, D is the central point of the high-end card cluster, and the included angle beta represents the included angle between the vector extending from the center of the non-high-end card to the bank card represented by the preprocessed data and the vector extending from the center of the non-high-end card to the center of the high-end card.
7. The apparatus of claim 6, wherein the first unit is configured to perform the steps of:
(a) collecting historical data of various bank card transactions, and preprocessing the historical data into n-dimensional data according to item classification;
(b) performing two types of processing on the historical data expressed by the n dimension by using a K-means algorithm so as to obtain central points C and D of the two types of clusters, wherein C is the central point of a non-high-end card cluster, and D is the central point of a high-end card cluster, and seed nodes of the two types of clusters are set as a geometric center A point of high-end card information data and a geometric center B point of the non-high-end card information data;
(c) evaluating the clustering result according to the included angle alpha between the vector AB and the vector CD; and
(d) if the included angle α is greater than a second threshold, then steps (b) and (c) are re-executed and the weight of each dimension is adjusted until α is less than or equal to the second threshold.
9. the apparatus of claim 7, wherein the first unit is configured to determine each of the vectors AB during the adjustment of the weights for each dimensionValue v of the dimensioniValue u of the corresponding dimension to the vector CDiRatio v ofi/uiAnd calculating the mean of the ratio in each dimension according to the following formula:
where d is the number of dimensions of the bank card data.
10. The apparatus of claim 9, wherein the first unit is configured to be at vi/uiIf the value of (d) is greater than the average ave, the data weight of the ith dimension is increased, otherwise, the data weight of the ith dimension is reduced.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410686072.6A CN105590156B (en) | 2014-11-25 | 2014-11-25 | Detection method of high-risk bank card and data processing device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410686072.6A CN105590156B (en) | 2014-11-25 | 2014-11-25 | Detection method of high-risk bank card and data processing device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105590156A CN105590156A (en) | 2016-05-18 |
CN105590156B true CN105590156B (en) | 2022-02-15 |
Family
ID=55929722
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410686072.6A Active CN105590156B (en) | 2014-11-25 | 2014-11-25 | Detection method of high-risk bank card and data processing device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105590156B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107563757B (en) | 2016-07-01 | 2020-09-22 | 阿里巴巴集团控股有限公司 | Data risk identification method and device |
CN108629680A (en) * | 2018-04-03 | 2018-10-09 | 中国农业银行股份有限公司 | A kind of Risk Identification Method and system |
CN109118053B (en) * | 2018-07-17 | 2022-04-05 | 创新先进技术有限公司 | Method and device for identifying card stealing risk transaction |
CN109919626B (en) * | 2019-03-11 | 2023-04-07 | ***股份有限公司 | High-risk bank card identification method and device |
CN110084468B (en) * | 2019-03-14 | 2020-09-01 | 阿里巴巴集团控股有限公司 | Risk identification method and device |
CN110290522B (en) * | 2019-07-17 | 2023-02-21 | 中国工商银行股份有限公司 | Risk identification method and device for mobile equipment and computer system |
CN111476664A (en) * | 2020-05-04 | 2020-07-31 | 武汉众邦银行股份有限公司 | Risk perception and risk prediction method based on real-time data flow |
CN112926989B (en) * | 2021-03-22 | 2023-09-05 | 华南理工大学 | Bank loan risk assessment method and equipment based on multi-view integrated learning |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8281990B2 (en) * | 2006-12-07 | 2012-10-09 | Smart Systems Innovations, Llc | Public transit system fare processor for transfers |
CN101236638A (en) * | 2008-02-20 | 2008-08-06 | 中国工商银行股份有限公司 | Web based bank card risk monitoring method and system |
CN101625740A (en) * | 2009-07-28 | 2010-01-13 | 交通银行股份有限公司 | Application business monitoring method, device and system |
CN103092931A (en) * | 2012-12-31 | 2013-05-08 | 武汉传神信息技术有限公司 | Multi-strategy combined document automatic classification method |
CN103559252A (en) * | 2013-11-01 | 2014-02-05 | 桂林电子科技大学 | Method for recommending scenery spots probably browsed by tourists |
-
2014
- 2014-11-25 CN CN201410686072.6A patent/CN105590156B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN105590156A (en) | 2016-05-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105590156B (en) | Detection method of high-risk bank card and data processing device | |
Awoyemi et al. | Credit card fraud detection using machine learning techniques: A comparative analysis | |
Zhang et al. | Machine learning and sampling scheme: An empirical study of money laundering detection | |
Dhankhad et al. | Supervised machine learning algorithms for credit card fraudulent transaction detection: a comparative study | |
CN108960833B (en) | Abnormal transaction identification method, equipment and storage medium based on heterogeneous financial characteristics | |
US20220383322A1 (en) | Clustering-based data selection for optimization of risk predictive machine learning models | |
US8145585B2 (en) | Automated methods and systems for the detection and identification of money service business transactions | |
Khan et al. | Implement credit card fraudulent detection system using observation probabilistic in hidden markov model | |
Arora et al. | Prediction of credit card defaults through data analysis and machine learning techniques | |
Abdelhamid et al. | Automatic bank fraud detection using support vector machines | |
Kumar et al. | Anti money laundering detection using Naïve Bayes classifier | |
Gao et al. | Research on Default Prediction for Credit Card Users Based on XGBoost‐LSTM Model | |
CN111861486A (en) | Abnormal account identification method, device, equipment and medium | |
Wang et al. | Credit Card Fraud Detection using Logistic Regression | |
CN112329862A (en) | Decision tree-based anti-money laundering method and system | |
Ahmed et al. | A Survey on Detection of Fraudulent Credit Card Transactions Using Machine Learning Algorithms | |
Bian et al. | Financial fraud detection: a new ensemble learning approach for imbalanced data | |
Jaiswal et al. | Credit card fraud detection using isolation forest and local outlier factor | |
Budianto et al. | Machine learning-based approach on dealing with binary classification problem in imbalanced financial data | |
Wang | Fraud detection based on FS-SMOTE model for credit card | |
Makatjane et al. | Detecting Financial Fraud in South Africa: A Comparison of Logistic Model Tree and Gradient Boosting Decision Tree | |
Anjum et al. | Cheat Detection for Credit Cards Using Artificial Intelligence | |
Thar et al. | Machine Learning Based Predictive Modelling for Fraud Detection in Digital Banking | |
Danaa et al. | Detecting electronic banking fraud on highly imbalanced data using hidden Markov models | |
Ekizoglu et al. | Fuzzy rule-based analysis of spatio-temporal ATM usage data for fraud detection and prevention |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |