CN114140248A - AI artificial intelligence technology-based abnormal transaction identification method - Google Patents
AI artificial intelligence technology-based abnormal transaction identification method Download PDFInfo
- Publication number
- CN114140248A CN114140248A CN202111523767.9A CN202111523767A CN114140248A CN 114140248 A CN114140248 A CN 114140248A CN 202111523767 A CN202111523767 A CN 202111523767A CN 114140248 A CN114140248 A CN 114140248A
- Authority
- CN
- China
- Prior art keywords
- transaction
- abnormal
- abnormal transaction
- data
- artificial intelligence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/259—Fusion by voting
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of abnormal transaction detection, in particular to an abnormal transaction identification method based on an AI artificial intelligence technology, which comprises the following steps: constructing an abnormal transaction detection model based on an AI artificial intelligence technology; acquiring trade order data, and establishing a corresponding trade sample set based on the trade order data; performing characteristic engineering processing on the transaction sample set, and constructing a corresponding training data set; training an abnormal transaction detection model based on a training data set; and completing abnormal transaction identification through the trained abnormal transaction detection model. The abnormal transaction identification method can improve the identification accuracy rate of the abnormal transaction and reduce the identification cost.
Description
Technical Field
The invention relates to the technical field of abnormal transaction detection, in particular to an abnormal transaction identification method based on an AI artificial intelligence technology.
Background
With the rapid development of the internet technology, more and more users trade through the e-commerce platform or the financial platform, which brings great convenience to the users, but also brings a plurality of abnormal trade situations. The abnormal transaction refers to a transaction in which the current operating condition is significantly different from the past. The occurrence of abnormal transactions is often associated with factors such as risk event outbreak, market fluctuation, customer base and operation environment change, and is an object of important attention in industries such as finance, retail and logistics.
To solve the problem of identification of the existing abnormal transactions, a chinese patent with publication number CN107918905A discloses "an abnormal transaction identification method", which includes: acquiring historical transaction data in a preset time period; determining at least one transaction group with a communication relation according to transaction main bodies and transaction behaviors in historical transaction data; extracting the characteristic data of the transaction group, and determining the transaction characteristic data of each transaction in the transaction group based on the characteristic data of the transaction group; and carrying out abnormity detection on the transaction characteristic data to determine abnormal transactions.
According to the abnormal transaction identification method in the existing scheme, the transaction group is determined, the range of risk transaction is narrowed, abnormal detection is carried out in the range of the transaction group through the transaction characteristic data, and then abnormal transaction can be efficiently and quickly detected. With the research and development of the AI artificial intelligence technology, the AI artificial intelligence model is widely applied due to the advantages of high recognition accuracy, low cost and the like. However, how to design an abnormal transaction identification method based on the AI artificial intelligence technology is a technical problem which needs to be solved urgently.
Disclosure of Invention
Aiming at the defects of the prior art, the technical problems to be solved by the invention are as follows: how to design an abnormal transaction identification method based on an AI artificial intelligence technology, thereby improving the identification accuracy of the abnormal transaction and reducing the identification cost.
In order to solve the technical problems, the invention adopts the following technical scheme:
an AI artificial intelligence technology-based abnormal transaction identification method comprises the following steps:
s1: constructing an abnormal transaction detection model based on an AI artificial intelligence technology;
s2: acquiring trade order data, and establishing a corresponding trade sample set based on the trade order data;
s3: performing characteristic engineering processing on the transaction sample set, and constructing a corresponding training data set;
s4: training an abnormal transaction detection model based on a training data set;
s5: and completing abnormal transaction identification through the trained abnormal transaction detection model.
Preferably, in step S1, an abnormal transaction detection model is constructed based on the isolated forest algorithm.
Preferably, in step S2, data preprocessing and data cleaning are performed on the transaction order data, so as to establish a corresponding transaction sample set.
Preferably, the data preprocessing and data cleansing include data deduplication, data completion, outlier screening, and data filtering.
Preferably, in step S3, the abnormality degree of each transaction sample in the transaction sample set is first determined; then, feature construction and feature selection are carried out based on the abnormal degree so as to obtain the category of each data in the transaction sample and abnormal labels for indicating whether the data are normal or not; and finally, taking the data with the abnormal labels as training data to construct a corresponding training data set.
Preferably, when determining the degree of abnormality of the transaction sample, the corresponding attribute of each transaction flow in the transaction sample is obtained, and then the degree of abnormality of the corresponding transaction sample is determined according to the attribute of the transaction flow.
Preferably, in step S4, the abnormal transaction detection model is trained by the following steps:
s401: sorting the training data in the training data set in an abnormal descending manner;
s402: in the model tuning stage, model parameters of the abnormal transaction detection model are adjusted based on the training data set;
s403: in the model fusion stage, model fusion is carried out in a simple fusion, weighted fusion and/or model fusion mode;
s404: and constructing a corresponding loss function according to the classification result of the abnormal transaction detection model and the class marking data error, and optimizing the model parameters of the abnormal transaction detection model based on the loss function until the model converges.
Preferably, after the abnormal transaction detection model is trained, the model is verified by a cross-validation method.
Preferably, the abnormal transaction detection model is packaged into a corresponding identification module by adopting a micro-service framework, and then the cloud deployment of the identification module is realized by adopting an algorithm scheduling service, an FTP service, a monitoring process and a heartbeat report mode, so that the opening on each transaction service line can be realized.
Preferably, after the identification module identifies the abnormal transaction, the interception gateway intercepts the abnormal transaction in a corresponding interception mode according to the transaction service type.
Compared with the prior art, the abnormal transaction identification method based on the AI artificial intelligence technology has the following beneficial effects:
according to the invention, the abnormal transaction detection model is constructed and trained based on the AI artificial intelligence technology, so that the abnormal transaction identification can be effectively completed based on the AI artificial intelligence technology, the identification accuracy of the abnormal transaction can be improved, and the identification cost can be reduced. Meanwhile, the training data set is constructed through characteristic engineering processing, and the abnormal transaction detection model is trained on the basis of the training data set, so that the training effect and the identification accuracy of the abnormal transaction detection model can be effectively ensured.
Drawings
For purposes of promoting a better understanding of the objects, aspects and advantages of the invention, reference will now be made in detail to the present invention as illustrated in the accompanying drawings, in which:
FIG. 1 is a logic diagram of an abnormal transaction identification method.
Detailed Description
The following is further detailed by the specific embodiments:
example (b):
the embodiment discloses an abnormal transaction identification method based on an AI artificial intelligence technology.
As shown in fig. 1, the abnormal transaction identification method based on the AI artificial intelligence technology includes the following steps:
s1: constructing an abnormal transaction detection model based on an AI artificial intelligence technology; and constructing an abnormal transaction detection model based on an isolated forest algorithm.
S2: acquiring trade order data, and establishing a corresponding trade sample set based on the trade order data; the acquisition of the transaction order data comprises transaction logs, background logs, consumption logs, application lists, user login logs, user transaction logs, activity logs and other scenes related to industrial APP transaction behaviors. The transaction order data includes order frequency, payment amount, payment account/collection account, and the like.
S3: performing characteristic engineering processing on the transaction sample set, and constructing a corresponding training data set;
s4: training an abnormal transaction detection model based on a training data set; training of the model based on the isolated forest algorithm is an existing mature means, and is not described herein any further.
S5: and completing abnormal transaction identification through the trained abnormal transaction detection model.
An isolated forest algorithm (iForest) is a rapid anomaly detection method based on Ensemble, has the advantages of linear time complexity and high precision, and is a state-of-the-art algorithm which meets the requirement of big data processing. iForest is applied to anomaly detection of continuous data, and an anomaly is defined as "outlier that is easily isolated", which can be understood as a point that is sparsely distributed and is far from a population with high density. Statistically, in the data space, the sparsely distributed regions indicate that the probability of data occurring in the regions is low, and thus the data falling in the regions can be considered abnormal.
According to the invention, the abnormal transaction detection model is constructed and trained based on the AI artificial intelligence technology, so that the abnormal transaction identification can be effectively completed based on the AI artificial intelligence technology, the identification accuracy of the abnormal transaction can be improved, and the identification cost can be reduced. Meanwhile, the training data set is constructed through characteristic engineering processing, and the abnormal transaction detection model is trained on the basis of the training data set, so that the training effect and the identification accuracy of the abnormal transaction detection model can be effectively ensured.
In the specific implementation process, data preprocessing and data cleaning are carried out on the transaction order data, and then a corresponding transaction sample set is established. The data preprocessing and data cleaning comprise data deduplication, data completion, outlier screening and data filtering. And filtering users through specific marks, such as low-activity users, high-frequency users and the like, and improving the data quality.
According to the method, the data quality of the transaction order data is improved through data preprocessing and data cleaning, and then the effect of establishing a training data set can be guaranteed, so that the training effect and the identification accuracy of an abnormal transaction detection model can be guaranteed.
In the specific implementation process, the abnormality degree of each transaction sample in the transaction sample set is determined; then, feature construction and feature selection are carried out based on the abnormal degree so as to obtain the category of each data in the transaction sample and abnormal labels for indicating whether the data are normal or not; and finally, taking the data with the abnormal labels as training data to construct a corresponding training data set. And when determining the abnormality degree of the transaction sample, acquiring the corresponding attribute of each transaction flow in the transaction sample, and then determining the abnormality degree of the corresponding transaction sample according to the attribute of the transaction flow. Attributes of the transaction stream include user attributes, economic attributes, social data, and transaction attributes.
The method comprises the following steps: for example, two attributes of date and time, whether the transaction time is the business time of a working day (the transaction event of the working day is 1, and the other is 0) is constructed, so that new characteristics with more expressive power and information quantity are obtained; assisted by feature splitting. And the external correlation characteristics are used for constructing the characteristics of easily distinguishing abnormal samples through model learning by using the data contents such as transaction logs, background logs, consumption logs, application lists, user login logs, user transaction logs, activity logs and the like through the characteristic fields. A piece of transaction flow data is a sample. Suitable samples are for example:
user attributes: login times, login time periods, login duration and the like;
the economic property is as follows: escrow amount, account balance, common transaction patterns, etc.;
transaction attributes: transaction amount, transaction type, transaction object, transaction frequency and the like;
social attributes: payer, payee identity, etc.;
when data are constructed based on the dimensions, statistical data can be constructed according to distribution characteristics besides the absolute values of the data.
The training data set for training the abnormal transaction detection model is constructed by determining the data abnormality degree, the characteristic construction, the characteristic selection and the abnormality marking, so that the construction effect of the training data set can be ensured, and the training effect and the identification accuracy of the abnormal transaction detection model can be ensured.
In the specific implementation process, an abnormal transaction detection model is trained through the following steps:
s401: sorting the training data in the training data set in an abnormal descending manner;
s402: in the model tuning stage, model parameters of the abnormal transaction detection model are adjusted based on the training data set; adjusting parameters such as trees, learning rate and max _ feature by means of random forest example and the like, and then fine-adjusting the parameters (such as minimum classification sample number and the like) of each tree; in adjusting the parameter loss rate, the parameter with the largest influence is the main parameter. Training of the model based on the isolated forest algorithm is an existing mature means, and is not described herein any further.
S403: in the model fusion stage, model fusion is carried out in a simple fusion, weighted fusion and/or model fusion mode;
s404: and constructing a corresponding loss function according to the classification result of the abnormal transaction detection model and the class marking data error, and optimizing the model parameters of the abnormal transaction detection model based on the loss function until the model converges. The construction of the loss function is an existing mature means, and is not described herein.
Simple fusion: 1) classifying problems, fusing voting methods, and not considering the scores of the single models; 2) and (4) regression problem, and calculating a fusion result on the assumption that the weight of each model is consistent.
And (3) weighted fusion: basically, the difference is that each model scores itself and the weight of the score is high.
Model fusion: the best effect can be achieved generally by inputting the outputs of a plurality of single models into a certain model as inputs and fusing the models, but the problem of overfitting exists.
According to the invention, the abnormal transaction detection model is trained through abnormal sequencing, model tuning and model fusion, so that the training effect and the identification accuracy of the abnormal transaction detection model can be effectively ensured.
In the specific implementation process, after the abnormal transaction detection model is trained, the model is verified through a cross verification method. Model verification by cross-validation is a mature means in the prior art and is not described herein.
The invention can effectively ensure the training effect of the abnormal transaction detection model through a cross verification method.
In the specific implementation process, a micro-service framework is adopted to package the abnormal transaction detection model into a corresponding identification module, and then cloud deployment of the identification module is realized in the modes of algorithm scheduling service, FTP service, monitoring process and heartbeat report, so that the method can be opened on each transaction service line. And after the identification module identifies the abnormal transaction, intercepting the abnormal transaction by adopting a corresponding intercepting mode based on the intercepting gateway according to the transaction service type.
The invention improves the transportability of the abnormal transaction identification service component by a micro-service encapsulation means, thereby improving the application effect of the abnormal transaction identification.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that, while the invention has been described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Meanwhile, the detailed structures, characteristics and the like of the common general knowledge in the embodiments are not described too much. Finally, the scope of the claims should be determined by the content of the claims, and the description of the embodiments and the like in the specification should be used for interpreting the content of the claims.
Claims (10)
1. An AI artificial intelligence technology-based abnormal transaction identification method is characterized by comprising the following steps:
s1: constructing an abnormal transaction detection model based on an AI artificial intelligence technology;
s2: acquiring trade order data, and establishing a corresponding trade sample set based on the trade order data;
s3: performing characteristic engineering processing on the transaction sample set, and constructing a corresponding training data set;
s4: training an abnormal transaction detection model based on a training data set;
s5: and completing abnormal transaction identification through the trained abnormal transaction detection model.
2. The AI artificial intelligence technology-based abnormal transaction identification method of claim 1, wherein: in step S1, an abnormal transaction detection model is constructed based on the isolated forest algorithm.
3. The AI artificial intelligence technology-based abnormal transaction identification method of claim 2, wherein: in step S2, data preprocessing and data cleaning are performed on the transaction order data, so as to establish a corresponding transaction sample set.
4. The AI artificial intelligence technology-based abnormal transaction identification method of claim 3, wherein: the data preprocessing and data cleaning comprise data deduplication, data completion, outlier screening and data filtering.
5. The AI artificial intelligence technology-based abnormal transaction identification method of claim 3, wherein: in step S3, the abnormality degree of each transaction sample in the transaction sample set is first determined; then, feature construction and feature selection are carried out based on the abnormal degree so as to obtain the category of each data in the transaction sample and abnormal labels for indicating whether the data are normal or not; and finally, taking the data with the abnormal labels as training data to construct a corresponding training data set.
6. The AI artificial intelligence technology-based abnormal transaction identification method of claim 5, wherein: and when determining the abnormality degree of the transaction sample, acquiring the corresponding attribute of each transaction flow in the transaction sample, and then determining the abnormality degree of the corresponding transaction sample according to the attribute of the transaction flow.
7. The AI artificial intelligence technology-based abnormal transaction recognition method of claim 5, wherein in step S4, the abnormal transaction detection model is trained by:
s401: sorting the training data in the training data set in an abnormal descending manner;
s402: in the model tuning stage, model parameters of the abnormal transaction detection model are adjusted based on the training data set;
s403: in the model fusion stage, model fusion is carried out in a simple fusion, weighted fusion and/or model fusion mode;
s404: and constructing a corresponding loss function according to the classification result of the abnormal transaction detection model and the class marking data error, and optimizing the model parameters of the abnormal transaction detection model based on the loss function until the model converges.
8. The AI artificial intelligence technology-based abnormal transaction identification method of claim 7, wherein: and after the abnormal transaction detection model is trained, performing model verification through a cross verification method.
9. The AI artificial intelligence technology-based abnormal transaction identification method of claim 1, wherein: and packaging the abnormal transaction detection model into a corresponding identification module by adopting a micro-service framework, and realizing the cloud deployment of the identification module by adopting an algorithm scheduling service, an FTP service, a monitoring process and a heartbeat report mode so as to be capable of opening on each transaction service line.
10. The AI artificial intelligence technology-based abnormal transaction identification method of claim 9, wherein: and after the identification module identifies the abnormal transaction, intercepting the abnormal transaction by adopting a corresponding intercepting mode based on the intercepting gateway according to the transaction service type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111523767.9A CN114140248A (en) | 2021-12-14 | 2021-12-14 | AI artificial intelligence technology-based abnormal transaction identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111523767.9A CN114140248A (en) | 2021-12-14 | 2021-12-14 | AI artificial intelligence technology-based abnormal transaction identification method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114140248A true CN114140248A (en) | 2022-03-04 |
Family
ID=80382560
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111523767.9A Pending CN114140248A (en) | 2021-12-14 | 2021-12-14 | AI artificial intelligence technology-based abnormal transaction identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114140248A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116185315A (en) * | 2023-04-27 | 2023-05-30 | 美恒通智能电子(广州)股份有限公司 | Hand-held printer data monitoring and early warning system and method based on artificial intelligence |
| CN117131445A (en) * | 2023-07-28 | 2023-11-28 | 深圳市财富趋势科技股份有限公司 | Abnormal transaction detection method and system |
| CN117372076A (en) * | 2023-08-23 | 2024-01-09 | 广东烟草广州市有限公司 | Abnormal transaction data monitoring method, device, equipment and storage medium |
-
2021
- 2021-12-14 CN CN202111523767.9A patent/CN114140248A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116185315A (en) * | 2023-04-27 | 2023-05-30 | 美恒通智能电子(广州)股份有限公司 | Hand-held printer data monitoring and early warning system and method based on artificial intelligence |
| CN117131445A (en) * | 2023-07-28 | 2023-11-28 | 深圳市财富趋势科技股份有限公司 | Abnormal transaction detection method and system |
| CN117372076A (en) * | 2023-08-23 | 2024-01-09 | 广东烟草广州市有限公司 | Abnormal transaction data monitoring method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10735285B2 (en) | Systems and methods for identifying and mitigating outlier network activity | |
| CN114140248A (en) | AI artificial intelligence technology-based abnormal transaction identification method | |
| US11093519B2 (en) | Artificial intelligence (AI) based automatic data remediation | |
| CN110956273A (en) | Credit scoring method and system integrating multiple machine learning models | |
| CN110008343A (en) | File classification method, device, equipment and computer readable storage medium | |
| CN110287316A (en) | A kind of Alarm Classification method, apparatus, electronic equipment and storage medium | |
| CN109034194A (en) | Transaction swindling behavior depth detection method based on feature differentiation | |
| CN106408325A (en) | User consumption behavior prediction analysis method based on user payment information and system | |
| CN110147389A (en) | Account number treating method and apparatus, storage medium and electronic device | |
| CN111986027A (en) | Abnormal transaction processing method and device based on artificial intelligence | |
| CN116485020B (en) | Supply chain risk identification early warning method, system and medium based on big data | |
| Dabab et al. | A decision model for data mining techniques | |
| CN117196630A (en) | Transaction risk prediction method, device, terminal equipment and storage medium | |
| CN116611911A (en) | Credit risk prediction method and device based on support vector machine | |
| CN113837481B (en) | Financial big data management system based on block chain | |
| CN116151857A (en) | Marketing model construction method and device | |
| CN114529255A (en) | Loan automatic approval method and system based on wind control scoring card | |
| CN108009224A (en) | The sorting technique and device of power customer | |
| CN106682985A (en) | Financial fraud identification method and system thereof | |
| CN112907254A (en) | Fraud transaction identification and model training method, device, equipment and storage medium | |
| Vijayarani et al. | Predicting Stock Market Using Machine Learning Algorithms | |
| CN115082071A (en) | Abnormal transaction account identification method and device and storage medium | |
| CN111612302A (en) | Group-level data management method and equipment | |
| Zhang et al. | Reliability-oriented product manufacturing quality improvement approach based on R-KQC and DMAIC | |
| Wijesinghe et al. | Machine learning for pre-auction sample selection |
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 |