CN108093405A - A kind of fraudulent call number analysis method and apparatus - Google Patents
A kind of fraudulent call number analysis method and apparatus Download PDFInfo
- Publication number
- CN108093405A CN108093405A CN201711079207.2A CN201711079207A CN108093405A CN 108093405 A CN108093405 A CN 108093405A CN 201711079207 A CN201711079207 A CN 201711079207A CN 108093405 A CN108093405 A CN 108093405A
- Authority
- CN
- China
- Prior art keywords
- fraudulent call
- call number
- fraudulent
- telephone number
- characteristic
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W12/00—Security arrangements; Authentication; Protecting privacy or anonymity
- H04W12/12—Detection or prevention of fraud
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/22—Traffic simulation tools or models
Abstract
The present invention provides a kind of fraudulent call number analysis method and apparatus, and analysis method includes：Obtain the abnormal behaviour data and characteristic of telephone number in Original CDR；Abnormal behaviour data include one or more in the anomalous sign code number of calls, dead number call number and the strange number number of calls；Characteristic includes active degree and communicating data, wherein, active degree includes calling the frequency and/or calling intensive period day, and communicating data includes percent of call completed, the latest earliest air time, air time, average call duration, major call area and one or more in average ring time；Two kinds of data are inputted into trained fraudulent call number analysis model, by Weighted Naive Bayes Classification Algorithm, obtain fraudulent call number analysis result.Whether the present invention can be that fraudulent call number carries out Accurate Analysis to telephone number；Various dimensions fraudulent call number analysis model is obtained, so as to whether be comprehensively that fraudulent call number is analyzed to telephone number.
Description
Technical field
The present invention relates to telecommunication fraud monitoring protection field, more particularly, to a kind of fraudulent call number analysis method
And equipment.
Background technology
The event of the swindle carried out in recent years using phone is deceived wide in outburst trend, and victim is deceived great amount,
Telecommunication fraud becomes the pain spot of telecommunication user.Current fraudulent call means are also varied, and swindler is often carried out by software
The number of changing, and counterfeit bank, operator, acquaintance, social security etc. carry out fraud.2013, whole nation communication fraud case had more than 30 ten thousand
It rises, victim is deceived nearly 10,000,000,000 yuan；2014, whole nation communication fraud case had more than 40 ten thousand, and victim is deceived nearly 10,700,000,000 yuan；
Telecommunication fraud case 590,000 is found altogether by national public security organ within 2015, rises 32.5% on year-on-year basis, causes economic loss nearly 22,200,000,000
Member.
In the past 10 years, telecommunication fraud case in China's is every year with the speed rapid growth of 20%-30%.It is national that quilt occurs altogether
It deceives ten million yuan or more of telecommunication fraud case 104 to rise, million yuan or more of case 2392 rises." the endowment money " of many masses " is rescued
Order money " it is cheated, it goes bankrupt or broke, is with one's family broken up；Some business capitals are cheated, break, trigger many Mass disturbances.In mesh
Among preceding telecommunication fraud, swindler possesses various fraud tactics, makes people hard to guard against, has seriously affected telecommunication security and has led to
Order is talked about, compromises the interests of telecommunication user and the prestige of telecommunications network.
It takes place frequently in view of above-mentioned telecommunication fraud case, under the severe background of social influence, occurs some for swindle electricity
The method that words number carries out analysis comparison, still, in existing method, mostly simply to the telephone number of the sampling in ticket
It is repeatedly screened and is compared, and carry out the computing of a series of complex, these methods are more complicated, not simple comprehensive enough.
The content of the invention
The present invention provides one kind and overcomes in above-mentioned existing method, more complicated, a kind of not simple enough comprehensive swindle electricity
Talk about number analysis method and apparatus.
According to an aspect of the present invention, a kind of fraudulent call number analysis method is provided, including：It obtains in Original CDR
Telephone number abnormal behaviour data and characteristic；The abnormal behaviour data include the anomalous sign code number of calls, spacing
One or more in the number of calls and the strange number number of calls；The characteristic includes active degree and communicating data,
Wherein, the active degree includes the day calling frequency and/or calling intensive period, and the communicating data includes percent of call completed, earliest
Air time, the latest air time, average call duration, major call area and the one or more in average ring time；
The abnormal behaviour data and the characteristic are inputted into trained fraudulent call number analysis model, it is simple by weighting
Bayesian Classification Arithmetic obtains fraudulent call number analysis result.
Preferably, the trained fraudulent call number analysis model is obtained by following steps：According to known swindleness
Deceive telephone number storehouse and non-fraudulent call number storehouse structure training set；Abnormal behaviour data are extracted to the data in the training set
And characteristic, and input fraudulent call number analysis model and successively trained：Determine abnormal behaviour data and described
The abnormal value range of characteristic, and determine the weight of each characteristic；Telephone number is obtained by weight coefficient algorithm
Belong to the conditional probability of fraudulent call number, and pass through Weighted Naive Bayes Classification Algorithm and obtain the trained swindle electricity
Talk about number analysis model.
Preferably, it is described that the abnormal behaviour data and the characteristic are inputted into trained fraudulent call number point
Model is analysed, by Weighted Naive Bayes Classification Algorithm, fraudulent call number analysis result is obtained and further comprises：In the swindleness
It deceives whether the telephone number judged in the first layer of telephone number resolutions model in the ticket has abnormal behaviour, will have different
Fraudulent call number analysis described in the telephone number typing of Chang Hangwei is as a result, in addition to the telephone number of typing, remaining electricity
Talk about the second layer that number enters the fraudulent call number analysis model；In the second layer of the fraudulent call number analysis model
It is middle to obtain the fraudulent call number analysis result.
Preferably, it is described judge the ticket in the first layer of the fraudulent call number analysis model in phone number
Whether code has abnormal behaviour, and fraudulent call number analysis result described in the telephone number typing with abnormal behaviour is further
Including：In the first layer of the fraudulent call number analysis model, if abnormal described in any telephone number whithin a period of time
The value of behavioral data is in the abnormal value range, it is determined that any telephone number has abnormal behaviour, described will appoint
One telephone number is identified as fraudulent call number, and by fraudulent call number analysis knot described in the fraudulent call number typing
Fruit.
Preferably, it is described that the fraudulent call number point is obtained in the second layer of the fraudulent call number analysis model
Analysis result further comprises：In the second layer of the fraudulent call number analysis model, according to the characteristic and described different
Normal value range, by weight coefficient algorithm, it is fraudulent call number to obtain the corresponding telephone number of any characteristic
Conditional probability；According to the conditional probability and the weight, by Weighted Naive Bayes Classification Algorithm, by any spy
The corresponding telephone number of sign data is categorized into any in fraudulent call number, doubtful fraudulent call number and normal telephone number
Kind, and using the type of the corresponding telephone number of any characteristic as the fraudulent call number analysis result.
Preferably, it is described that the abnormal behaviour data and the characteristic are inputted into trained fraudulent call number point
Model is analysed, by Weighted Naive Bayes Classification Algorithm, fraudulent call number analysis result is obtained and further includes afterwards：Described in analysis
The location of fraudulent call number and counterfeit type, and fraudulent call number analysis result described in typing.
Preferably, the fraudulent call number analysis result includes：It is recorded in the first layer of fraudulent call number analysis model
The fraudulent call number entered；In the second layer of fraudulent call number analysis model, by the corresponding phone of any characteristic
It is number sorted into any one of fraudulent call number, doubtful fraudulent call number and normal telephone number；Each swindle
The location of telephone number and counterfeit type.
Preferably, the abnormal behaviour data for obtaining the telephone number in Original CDR and characteristic are further wrapped
It includes：The telephone number in the Original CDR is obtained from big data, and the telephone number in the Original CDR is carried out pre-
Processing, obtains the abnormal behaviour data and the characteristic.
Preferably, the location of the analysis fraudulent call number and counterfeit type, and fraudulent call described in typing
It is further included after number analysis result：The fraudulent call number analysis result is imported into the training set.
According to another aspect of the present invention, a kind of fraudulent call number analysis equipment is provided, including：At least one processing
Device；And at least one processor being connected with the processor communication, wherein：The memory storage has can be by the processing
The program instruction that device performs, the processor call described program instruction to be able to carry out such as any of the above-described analysis method.
A kind of fraudulent call number analysis method and apparatus provided by the invention, by being arranged on trained fraudulent call
By Weighted Naive Bayes Classification Algorithm in number analysis model, fraudulent call number analysis is obtained as a result, it is possible to simply
Whether it is that fraudulent call number carries out Accurate Analysis to telephone number, safeguards the safety of telecommunications network；By to abnormal behaviour data
Classify with characteristic, the fraudulent call number analysis model for various dimensions can be obtained, so as to comprehensively right
Whether telephone number is that fraudulent call number is analyzed.
Description of the drawings
Fig. 1 is a kind of flow chart of fraudulent call number analysis method in the embodiment of the present invention；
Fig. 2 is a kind of structure diagram of fraudulent call number analysis equipment in the embodiment of the present invention.
Specific embodiment
With reference to the accompanying drawings and examples, the specific embodiment of the present invention is described in further detail.Implement below
Example is not limited to the scope of the present invention for illustrating the present invention.
Fig. 1 is a kind of flow chart of fraudulent call number analysis method in the embodiment of the present invention, as shown in Figure 1, described
Analysis method includes：Obtain the abnormal behaviour data and characteristic of the telephone number in Original CDR；The abnormal behaviour number
According to including the one or more in the abnormal number number of calls, dead number call number and the strange number number of calls；The feature
Data include active degree and communicating data, wherein, the active degree includes the day calling frequency and/or calling intensive period,
The communicating data include percent of call completed, the earliest air time, the latest the air time, average call duration, major call area and
One or more in average ring time；By the abnormal behaviour data and the trained swindle electricity of characteristic input
Number analysis model is talked about, by Weighted Naive Bayes Classification Algorithm, obtains fraudulent call number analysis result.
Specifically, the abnormal behaviour data are exhaled including the anomalous sign code number of calls, dead number call number and strange number
It is number.Wherein, the abnormal number call refers to the telephone number of mobile phone None- identified or special number section telephone number
It is called.The dead number call refers to that swindler calls spacing using software, spacing be in telecommunications network due to
Family cancellation, the number of changing are reported the loss and telephone number section that situations such as defaulting subscriber occurs is empty situation.The strange number calls
Refer to the strange number calling that swindler carries out certain number section number subsequenct call generation.
Further, the characteristic includes active degree and communicating data.Wherein, the active degree is exhaled including day
It is the frequency and calls the intensive period, but not limited to this.The communicating data includes percent of call completed, and the earliest air time converses the latest
Time, average call duration, major call area and average ring time, but not limited to this.
Further, the Weighted Naive Bayes Classification Algorithm refers to divide sample by calculating posterior probability
Class.Such as：It is assumed that X represents that caller was averaged ring duration less than 0.5 second and called average call duration is less than 1 second, H represents false
It is harassing call to determine X, then P (H | X) is represented be averaged duration less than 0.5 second when the caller of telephone number X and is called average talk
When duration is less than 1 second, telephone number X is that fraudulent call number firmly believes degree.
In Weighted Naive Bayes Classification Algorithm, weights are assigned for each characteristic according to importance, characteristic
Weights are bigger, and it is bigger to represent influence of this feature data to classification.Preferably, the active degree in characteristic and call feelings
Condition is to judge the important evidence of fraudulent call number, can be increased its weight setting proportion, to improve recognition efficiency.
A kind of fraudulent call number analysis method provided by the invention, by being arranged on trained fraudulent call number point
It analyses in model through Weighted Naive Bayes Classification Algorithm, obtains fraudulent call number analysis as a result, it is possible to simply to phone
Whether number is that fraudulent call number carries out Accurate Analysis, safeguards the safety of telecommunications network；By to abnormal behaviour data and feature
Data are classified, and the fraudulent call number analysis model for various dimensions can be obtained, so as to comprehensively to phone number
Whether code is that fraudulent call number is analyzed.A kind of fraudulent call number analysis method provided by the invention passes through setting pair
Telephone number in Original CDR carries out format conversion so that the telephone number in the ticket after format conversion of acquisition is convenient for
Subsequent fraudulent call number analysis.
Based on above-described embodiment, the trained fraudulent call number analysis model is obtained by following steps：According to
Known fraudulent call number storehouse and non-fraudulent call number storehouse structure training set；Data in the training set are extracted abnormal
Behavioral data and characteristic, and input fraudulent call number analysis model and successively trained：Determine the abnormal behaviour number
According to the abnormal value range with the characteristic, and determine the weight of each characteristic；It is obtained by weight coefficient algorithm
Telephone number belongs to the conditional probability of fraudulent call number, and by being trained described in Weighted Naive Bayes Classification Algorithm acquisition
Fraudulent call number analysis model.
Specifically, the abnormal value range of abnormal behaviour data and characteristic is determined by statistics.
Specifically, defining characteristic attribute value with the abnormal value range has the fraudulent call of intersection and normal master
Phone is, and in the case of characteristic attribute does not have intersection, it directly can be into according to the abnormal value range of specific properties
Row judges.The weight is the weight of the active degree and the weight of the communicating data.
It should be noted that there is higher weight without the characteristic overlapped for abnormal value range.
Further, successively training has periodically for the progress.
A kind of fraudulent call number analysis method provided by the invention, by setting periodically to fraudulent call number analysis
Model is trained, its efficiency and accuracy can be continuously improved.
It is described by the abnormal behaviour data and the trained swindle electricity of characteristic input based on above-described embodiment
Number analysis model is talked about, by Weighted Naive Bayes Classification Algorithm, fraudulent call number analysis result is obtained and further comprises：
Judge whether the telephone number in the ticket has abnormal behaviour in the first layer of the fraudulent call number analysis model,
By fraudulent call number analysis described in the telephone number typing with abnormal behaviour as a result, in addition to the telephone number of typing,
Remaining telephone number enters the second layer of the fraudulent call number analysis model；In the fraudulent call number analysis model
The second layer in obtain the fraudulent call number analysis result.
It is described to judge in the first layer of the fraudulent call number analysis model in the ticket based on above-described embodiment
Telephone number whether have abnormal behaviour, by fraudulent call number analysis knot described in the telephone number typing with abnormal behaviour
Fruit further comprises：In the first layer of the fraudulent call number analysis model, if whithin a period of time in any telephone number
The value of the abnormal behaviour data is in the abnormal value range, it is determined that and any telephone number has abnormal behaviour,
Any telephone number is identified as fraudulent call number, and by fraudulent call number described in the fraudulent call number typing
Analysis result.
Specifically, the abnormal behaviour includes abnormal number call, dead number call and strange number calling.The anomalous sign
Code calling, the dead number call and strange number calling are the event of small probability.If whithin a period of time, any phone
Abnormal behaviour continuous trigger described in number, i.e., whithin a period of time described in any telephone number at the value of abnormal behaviour data
In the abnormal value range, it is determined that any telephone number has abnormal behaviour.It should be noted that the exception row
Value for data be the abnormal number number of calls, dead number call number and the strange number number of calls and value.Exception takes
It is worth scope by being obtained to fraudulent call number analysis model training.
Further, the high frequency strange number number of calls in telephone number history call can be as judgement swindle number
Important evidence.If any telephone number is identified as fraudulent call number in the first layer of fraudulent call number analysis model,
Then any telephone number without the second layer of follow-up fraudulent call number analysis model analysis.
It is described that the swindle electricity is obtained in the second layer of the fraudulent call number analysis model based on above-described embodiment
Words number analysis result further comprises：In the second layer of the fraudulent call number analysis model, according to the characteristic
With the abnormal value range, by weight coefficient algorithm, it is swindle to obtain the corresponding telephone number of any characteristic
The conditional probability of telephone number；According to the conditional probability and the weight, by Weighted Naive Bayes Classification Algorithm, will appoint
The corresponding telephone number of one characteristic is categorized into fraudulent call number, doubtful fraudulent call number and normal telephone number
Any one of, and using the type of the corresponding telephone number of any characteristic as the fraudulent call number analysis knot
Fruit.
Specifically, the present embodiment further comprises design conditions probability P (A | B), wherein, A represents that the telephone number is swindleness
Number is deceived, B represents that the corresponding characteristic of the telephone number falls into the value range of the corresponding characteristic of the telephone number.
Further, by weight coefficient algorithm, the corresponding telephone number of any characteristic is obtained as swindle electricity
The conditional probability of words number specifically includes：According to weight coefficient algorithm, by the posterior probability and any feature of any feature data
The multiplied by weight of data obtains the conditional probability that the corresponding telephone number of any characteristic is fraudulent call number.
Further, according to the conditional probability and the weight, by Weighted Naive Bayes Classification Algorithm, calculate and appoint
The prior probability of one characteristic, and the weights of any characteristic are obtained, if any characteristic
With high weight, then fraudulent call number is categorized into；If any characteristic has middle weights, it is categorized into doubtful
Fraudulent call number；Any characteristic has low weights, then is categorized into normal telephone number.
Further, the fraudulent call number analysis result refers to：The type and words of any telephone number in ticket
Any telephone number in list is fraudulent call number either doubtful fraudulent call number or normal telephone number.
A kind of fraudulent call number analysis method provided by the invention, by setting to fraudulent call number analysis model point
Whether layer analysis can be exactly that fraudulent call number is analyzed to telephone number.
It is described by the abnormal behaviour data and the trained swindle electricity of characteristic input based on above-described embodiment
Number analysis model is talked about, by Weighted Naive Bayes Classification Algorithm, fraudulent call number analysis result is obtained and further includes afterwards：
Analyze the location of the fraudulent call number and counterfeit type, and fraudulent call number analysis result described in typing.
Specifically, the location refers to：The place region situation of fraudulent call number, such as home or overseas.The institute
It can be as the foundation of fraudulent call number retrospect, for judging the swindle wildness degree of each department on ground.
Further, the counterfeit type refers to：The possible swindle type of fraudulent call number, such as counterfeit bank, society
It protects, acquaintance etc..
Further, location is as after telephone number is identified as fraudulent call, and further analysis swindle is normal
Send out the foundation in area；Counterfeit type judges according to the counterfeit telephone number-type of fraudulent call number to be identified.
Based on above-described embodiment, the fraudulent call number analysis result includes：In fraudulent call number analysis model
The fraudulent call number of first layer typing；In the second layer of fraudulent call number analysis model, by any characteristic pair
The telephone number answered is categorized into any one of fraudulent call number, doubtful fraudulent call number and normal telephone number；It is each
The location of the fraudulent call number and counterfeit type.
Further, the fraudulent call number analysis result includes：The class of each telephone number in the ticket
The location and counterfeit type of type and each fraudulent call number.
Based on above-described embodiment, it is described obtain Original CDR in telephone number abnormal behaviour data and characteristic into
One step includes：The telephone number in the Original CDR is obtained from big data, and to the telephone number in the Original CDR
It carries out pretreatment and obtains the abnormal behaviour data and the characteristic.
Specifically, the big data (big data), refer to can not in the range of certain time with conventional software instrument into
The data acquisition system that row is captured, manages and handled.
Further, big data brings three subversiveness Concept Changes：It is total data rather than stochastical sampling；It is
General direction rather than precise guidance；It is correlativity rather than causality.
Described is that total data rather than stochastical sampling refer to：In the big data epoch, more data can be analyzed, are had
When can even handle with some special relevant all data of phenomenon, and be no longer dependent on stochastical sampling.
Described is that general direction rather than accurate system refer to：Data is so more, so that high precision need not be pursued
Degree；Need the data analyzed seldom before, so must quantify to record as accurately as possible, with the expansion of scale, to accurate
The requirement of degree reduces；Have big data, then no longer need to get to the bottom to a phenomenon, the development side of to master substantially
To suitably ignoring the accuracy in microcosmic point, macroscopic aspect can be caused to possess better insight.
Described is that correlativity rather than causality refer to：It is the custom of the mankind for a long time to find causality,
The big data epoch without the causality dig-inned between things, and look for the correlativity between things；Correlativity is perhaps
It cannot inform why something can occur exactly, but this can be reminded to occur.
A kind of fraudulent call number analysis method provided by the invention obtains the original words by setting from big data
Telephone number in list can comprehensively analyze ticket information from multiple dimensions, and one is trained among the data of magnanimity
Efficient, the high fraudulent call number analysis model of discrimination is realized effective identification to fraudulent call, has been evaded to a other
Telephone number carries out analysis is difficult that the problem of whether telephone number is fraudulent call number distinguished.
Based on above-described embodiment, the location of the analysis fraudulent call number and counterfeit type, and described in typing
It is further included after fraudulent call number analysis result：The fraudulent call number analysis result is imported into the training set.
Specifically, the fraudulent call number analysis result is also preserved before export.
A kind of fraudulent call number analysis method provided by the invention, by setting the fraudulent call number analysis knot
Tab phenolphthaleinum enters the training set, can increase the accuracy and recognition efficiency of fraudulent call number analysis model.
As a preferred embodiment, below to a kind of specific step of fraudulent call number analysis method provided by the invention
Suddenly explain：
First, Original CDR is pre-processed, format conversion is carried out to the telephone number in Original CDR, described in acquisition
The abnormal behaviour data and characteristic of telephone number in ticket and corresponding relation, the abnormal behaviour data include abnormal
Number call number, dead number call number and the strange number number of calls, the characteristic include active degree and call number
According to；Wherein, the active degree includes the day calling frequency and calls intensive period, and the communicating data includes percent of call completed, earliest
Air time, the latest air time, average call duration, major call area and average ring time.
Secondly, training set is built according to known fraudulent call number storehouse；Data in the training set are inputted into swindle
Telephone number resolutions model is successively trained, and obtains the trained fraudulent call number analysis model.
Again, fraudulent call number analysis is carried out to real-time ticket, and obtains fraudulent call number analysis as a result, will swindleness
Telephone number resolutions result is deceived to preserve and export.
Last, the fraudulent call number analysis result is imported into the training set, and carries out subsequent processing, will such as be divided
The fraudulent call number identified is analysed to add in blacklist or report relevant department.
Based on above-described embodiment, Fig. 2 is that a kind of structure of fraudulent call number analysis equipment in the embodiment of the present invention is shown
It is intended to, as shown in Fig. 2, the equipment includes：At least one processor 301；And with the processor 301 communication connection at least
One memory 302, wherein：The memory 302 is stored with the program instruction that can be performed by the processor 301, the place
Reason device 301 calls described program instruction to be able to carry out the fraudulent call number analysis method that the various embodiments described above are provided, such as
Including：The abnormal behaviour data and characteristic of the telephone number in ticket are obtained, the abnormal behaviour data include anomalous sign
The code number of calls, dead number call number and the strange number number of calls, the characteristic include active degree and communicating data；
The abnormal behaviour data and the characteristic are inputted into trained fraudulent call number analysis model, it is simple by weighting
Bayesian Classification Arithmetic obtains fraudulent call number analysis result.
A kind of fraudulent call number analysis method and apparatus provided by the invention, by being arranged on trained fraudulent call
By Weighted Naive Bayes Classification Algorithm in number analysis model, fraudulent call number analysis is obtained as a result, it is possible to simply
Whether it is that fraudulent call number carries out Accurate Analysis to telephone number, safeguards the safety of telecommunications network；By to abnormal behaviour data
Classify with characteristic, the fraudulent call number analysis model for various dimensions can be obtained, so as to comprehensively right
Whether telephone number is that fraudulent call number is analyzed.
Finally, method of the invention is only preferable embodiment, is not intended to limit the scope of the present invention.It is all
Within the spirit and principles in the present invention, any modifications, equivalent replacements and improvements are made should be included in the protection of the present invention
Within the scope of.
Claims (10)
- A kind of 1. fraudulent call number analysis method, which is characterized in that including：Obtain the abnormal behaviour data and characteristic of the telephone number in Original CDR；The abnormal behaviour data include abnormal One or more in number call number, dead number call number and the strange number number of calls；The characteristic includes living Jump degree and communicating data, wherein, the active degree includes the day calling frequency and/or calling intensive period, the call number During according to including percent of call completed, earliest air time, air time, average call duration, major call area and average ring the latest Between in one or more；The abnormal behaviour data and the characteristic are inputted into trained fraudulent call number analysis model, pass through weighting Naive Bayes Classification Algorithm obtains fraudulent call number analysis result.
- 2. analysis method according to claim 1, which is characterized in that the trained fraudulent call number analysis model It is obtained by following steps：According to known fraudulent call number storehouse and non-fraudulent call number storehouse structure training set；To the data extraction abnormal behaviour data and characteristic in the training set, and input fraudulent call number analysis model Successively trained：It determines the abnormal value range of the abnormal behaviour data and the characteristic, and determines the power of each characteristic Weight；Telephone number is obtained by weight coefficient algorithm and belongs to the conditional probability of fraudulent call number, and passes through the simple pattra leaves of weighting This sorting algorithm obtains the trained fraudulent call number analysis model.
- 3. analysis method according to claim 2, which is characterized in that described by the abnormal behaviour data and the feature Data input trained fraudulent call number analysis model, by Weighted Naive Bayes Classification Algorithm, obtain fraudulent call Number analysis result further comprises：It is abnormal to judge whether the telephone number in the ticket has in the first layer of the fraudulent call number analysis model Behavior, by fraudulent call number analysis described in the telephone number typing with abnormal behaviour as a result, the phone number except typing Code is outer, and remaining telephone number enters the second layer of the fraudulent call number analysis model；In the fraudulent call number point It analyses in the second layer of model and obtains the fraudulent call number analysis result.
- 4. analysis method according to claim 3, which is characterized in that described in the fraudulent call number analysis model Judge whether the telephone number in the ticket has abnormal behaviour in first layer, by the telephone number typing with abnormal behaviour The fraudulent call number analysis result further comprises：In the first layer of the fraudulent call number analysis model, if abnormal row described in any telephone number whithin a period of time The abnormal value range is in for the value of data, it is determined that any telephone number has abnormal behaviour, will be described any Telephone number is identified as fraudulent call number, and by fraudulent call number analysis result described in the fraudulent call number typing.
- 5. analysis method according to claim 3, which is characterized in that described in the fraudulent call number analysis model The fraudulent call number analysis result is obtained in the second layer to further comprise：In the second layer of the fraudulent call number analysis model, according to the characteristic and the abnormal value range, lead to Weight coefficient algorithm is crossed, obtains the conditional probability that the corresponding telephone number of any characteristic is fraudulent call number；According to the conditional probability and the weight, by Weighted Naive Bayes Classification Algorithm, by any characteristic Corresponding telephone number is categorized into any one of fraudulent call number, doubtful fraudulent call number and normal telephone number, and Using the type of the corresponding telephone number of any characteristic as the fraudulent call number analysis result.
- 6. analysis method according to claim 5, which is characterized in that described by the abnormal behaviour data and the feature Data input trained fraudulent call number analysis model, by Weighted Naive Bayes Classification Algorithm, obtain fraudulent call It is further included after number analysis result：Analyze the location of the fraudulent call number and counterfeit type, and fraudulent call number analysis result described in typing.
- 7. analysis method according to claim 1, which is characterized in that the fraudulent call number analysis result includes：In the fraudulent call number of the first layer typing of fraudulent call number analysis model；In the second layer of fraudulent call number analysis model, the corresponding telephone number of any characteristic is categorized into swindle Any one of telephone number, doubtful fraudulent call number and normal telephone number；The location of each fraudulent call number and counterfeit type.
- 8. analysis method according to claim 1, which is characterized in that described to obtain the different of the telephone number in Original CDR Normal behavioral data and characteristic further comprise：The telephone number in the Original CDR is obtained from big data, and the telephone number in the Original CDR is carried out pre- Processing, obtains the abnormal behaviour data and the characteristic.
- 9. analysis method according to claim 6, which is characterized in that the location of the analysis fraudulent call number With counterfeit type, and fraudulent call number analysis result described in typing after further include：The fraudulent call number analysis result is imported into the training set.
- 10. a kind of fraudulent call number analysis equipment, which is characterized in that including：At least one processor；And at least one processor being connected with the processor communication, wherein：The memory is deposited The program instruction that can be performed by the processor is contained, the processor calls described program instruction to be able to carry out such as claim 1 to 9 any analysis method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711079207.2A CN108093405A (en) | 2017-11-06 | 2017-11-06 | A kind of fraudulent call number analysis method and apparatus |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711079207.2A CN108093405A (en) | 2017-11-06 | 2017-11-06 | A kind of fraudulent call number analysis method and apparatus |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108093405A true CN108093405A (en) | 2018-05-29 |
Family
ID=62171941
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711079207.2A Pending CN108093405A (en) | 2017-11-06 | 2017-11-06 | A kind of fraudulent call number analysis method and apparatus |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108093405A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108924333A (en) * | 2018-06-12 | 2018-11-30 | 阿里巴巴集团控股有限公司 | Fraudulent call recognition methods, device and system |
| CN109274834A (en) * | 2018-09-27 | 2019-01-25 | 杭州东信北邮信息技术有限公司 | A kind of express delivery number identification method based on call behavior |
| CN109327627A (en) * | 2018-11-27 | 2019-02-12 | 深圳声笑科技有限公司 | Telephone number recognition methods, device and storage medium based on block chain |
| CN109711984A (en) * | 2019-01-23 | 2019-05-03 | 北京市天元网络技术股份有限公司 | Risk monitoring and control method and device before a kind of loan based on collection |
| CN110087230A (en) * | 2019-04-26 | 2019-08-02 | 同盾控股有限公司 | Data processing method, device, storage medium and electronic equipment |
| CN110213449A (en) * | 2019-05-17 | 2019-09-06 | 国家计算机网络与信息安全管理中心 | A kind of recognition methods of roaming swindle number |
| CN111031546A (en) * | 2019-11-29 | 2020-04-17 | 武汉烽火众智数字技术有限责任公司 | LR model training method applied to telephone number analysis and using method |
| CN111385420A (en) * | 2018-12-29 | 2020-07-07 | 中兴通讯股份有限公司 | User identification method and device |
| CN111465021A (en) * | 2020-04-01 | 2020-07-28 | 北京中亦安图科技股份有限公司 | Graph-based crank call identification model construction method |
| CN113163057A (en) * | 2021-01-20 | 2021-07-23 | 北京工业大学 | Method for constructing dynamic identification interval of fraud telephone |
| CN113727351A (en) * | 2020-05-12 | 2021-11-30 | 中国移动通信集团广东有限公司 | Communication fraud identification method and device and electronic equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150269941A1 (en) * | 2014-03-21 | 2015-09-24 | Wells Fargo Bank, N.A. | Fraud detection |
| CN106686261A (en) * | 2017-01-19 | 2017-05-17 | 腾讯科技（深圳）有限公司 | Information processing method and system |
| CN106791220A (en) * | 2016-11-04 | 2017-05-31 | 国家计算机网络与信息安全管理中心 | Prevent the method and system of telephone fraud |
-
2017
- 2017-11-06 CN CN201711079207.2A patent/CN108093405A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150269941A1 (en) * | 2014-03-21 | 2015-09-24 | Wells Fargo Bank, N.A. | Fraud detection |
| CN106791220A (en) * | 2016-11-04 | 2017-05-31 | 国家计算机网络与信息安全管理中心 | Prevent the method and system of telephone fraud |
| CN106686261A (en) * | 2017-01-19 | 2017-05-17 | 腾讯科技（深圳）有限公司 | Information processing method and system |
Non-Patent Citations (1)
| Title |
|---|
| 刘剑: "基于数据挖掘技术实现骚扰电话识别", 《CNKI中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108924333A (en) * | 2018-06-12 | 2018-11-30 | 阿里巴巴集团控股有限公司 | Fraudulent call recognition methods, device and system |
| CN109274834A (en) * | 2018-09-27 | 2019-01-25 | 杭州东信北邮信息技术有限公司 | A kind of express delivery number identification method based on call behavior |
| CN109274834B (en) * | 2018-09-27 | 2021-04-20 | 杭州东信北邮信息技术有限公司 | Express number identification method based on call behavior |
| CN109327627A (en) * | 2018-11-27 | 2019-02-12 | 深圳声笑科技有限公司 | Telephone number recognition methods, device and storage medium based on block chain |
| CN111385420B (en) * | 2018-12-29 | 2022-04-29 | 中兴通讯股份有限公司 | User identification method and device, storage medium and electronic device |
| CN111385420A (en) * | 2018-12-29 | 2020-07-07 | 中兴通讯股份有限公司 | User identification method and device |
| CN109711984A (en) * | 2019-01-23 | 2019-05-03 | 北京市天元网络技术股份有限公司 | Risk monitoring and control method and device before a kind of loan based on collection |
| CN110087230A (en) * | 2019-04-26 | 2019-08-02 | 同盾控股有限公司 | Data processing method, device, storage medium and electronic equipment |
| CN110087230B (en) * | 2019-04-26 | 2020-09-15 | 同盾控股有限公司 | Data processing method, data processing device, storage medium and electronic equipment |
| CN110213449A (en) * | 2019-05-17 | 2019-09-06 | 国家计算机网络与信息安全管理中心 | A kind of recognition methods of roaming swindle number |
| CN111031546A (en) * | 2019-11-29 | 2020-04-17 | 武汉烽火众智数字技术有限责任公司 | LR model training method applied to telephone number analysis and using method |
| CN111031546B (en) * | 2019-11-29 | 2023-09-19 | 武汉烽火众智数字技术有限责任公司 | LR model training method applied to telephone number analysis and application method |
| CN111465021A (en) * | 2020-04-01 | 2020-07-28 | 北京中亦安图科技股份有限公司 | Graph-based crank call identification model construction method |
| CN111465021B (en) * | 2020-04-01 | 2023-06-09 | 北京中亦安图科技股份有限公司 | Graph-based crank call identification model construction method |
| CN113727351A (en) * | 2020-05-12 | 2021-11-30 | 中国移动通信集团广东有限公司 | Communication fraud identification method and device and electronic equipment |
| CN113727351B (en) * | 2020-05-12 | 2024-03-19 | 中国移动通信集团广东有限公司 | Communication fraud identification method and device and electronic equipment |
| CN113163057B (en) * | 2021-01-20 | 2022-09-30 | 北京工业大学 | Method for constructing dynamic identification interval of fraud telephone |
| CN113163057A (en) * | 2021-01-20 | 2021-07-23 | 北京工业大学 | Method for constructing dynamic identification interval of fraud telephone |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108093405A (en) | A kind of fraudulent call number analysis method and apparatus | |
| CN108924333B (en) | Fraud telephone identification method, device and system | |
| CN106791220B (en) | Method and system for preventing telephone fraud | |
| Becker et al. | Fraud detection in telecommunications: History and lessons learned | |
| Xing et al. | Employing latent dirichlet allocation for fraud detection in telecommunications | |
| CN109451182B (en) | Detection method and device for fraud telephone | |
| CN107506776A (en) | A kind of analysis method of fraudulent call number | |
| CN110401779A (en) | A kind of method, apparatus and computer readable storage medium identifying telephone number | |
| Lu et al. | Telecom fraud identification based on ADASYN and random forest | |
| CN108243049A (en) | Telecoms Fraud recognition methods and device | |
| CN109819126A (en) | A kind of exception number identification method and device | |
| CN110493476B (en) | Detection method, device, server and storage medium | |
| CN101686444A (en) | System and method for detecting spam SMS sender number in real time | |
| Arafat et al. | Detection of wangiri telecommunication fraud using ensemble learning | |
| CN110213448A (en) | Malice number identification method, device, storage medium and computer equipment | |
| CN111626754B (en) | Card-keeping user identification method and device | |
| CN108810290A (en) | A kind of method and system of the identification of fraudulent call | |
| JP2000507765A (en) | Fraud monitoring in telecommunications networks | |
| Qayyum et al. | Fraudulent call detection for mobile networks | |
| CN114513791A (en) | Telecom anti-fraud method based on machine learning | |
| CN111131627B (en) | Method, device and readable medium for detecting personal harmful call based on streaming data atlas | |
| CN111105064A (en) | Method and device for determining suspected information of fraud event | |
| KR20130083286A (en) | System for processing spam for mobile phone | |
| CN109525739B (en) | Telephone number identification method and device and server | |
| CN107222319A (en) | A kind of traffic operation analysis method and device |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180529 |