CN108520042B - System and method for realizing suspect case-involved role calibration and role evaluation in detection work - Google Patents
System and method for realizing suspect case-involved role calibration and role evaluation in detection work Download PDFInfo
- Publication number
- CN108520042B CN108520042B CN201810291232.5A CN201810291232A CN108520042B CN 108520042 B CN108520042 B CN 108520042B CN 201810291232 A CN201810291232 A CN 201810291232A CN 108520042 B CN108520042 B CN 108520042B
- Authority
- CN
- China
- Prior art keywords
- role
- matrix
- data set
- eci
- record
- 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)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a system and a method for realizing suspect case-involved role calibration and role evaluation in detection work, wherein the method comprises the steps of preprocessing a historical data set and a newly-entered converged data set by using a data normalization processing module, then constructing a digital feature matrix FM and a role re-identification model ECI _ R, then constructing and training a correlation analysis deep neural network CRDN, and constructing a relation network atlas CR by using the role attribute calibration as a centerHISSet; and finally, a weight evaluation model ECI _ W is constructed for weight evaluation, so that a flexible, quick and active response case handling environment can be provided for detection personnel, a case is strived to break the best fighter, the case solving efficiency of the data is improved, the value of various fusion data is fully mined, and the trust of the masses on the case solving of the public security institution is enhanced. Meanwhile, the popularization and application of the system can provide a supporting effect for the safe operation of the national economic system.
Description
Technical Field
The invention relates to the technical field of public security investigation, in particular to the technical field of analysis and study, and specifically relates to a system and a method for realizing role calibration and role evaluation of suspect cases in investigation.
Background
With the rapid development of communication networks such as 4G communication, fiber to the home and the like, internet technology, particularly mobile internet, is rapidly developed, various social networks, payment networks, travel services, shopping websites and game platforms are increasingly closely linked with the lives of people, various data are generated in all links of social life at all times, and related data records generated by the social networks can be found everywhere from the internet, the mobile internet to the internet of things. With the increasing popularization of social networks, people's internet life is increasingly abundant, and the exponential growth of various kinds of APP, in the process, massive data closely related to the monitored work are generated, wherein a lot of valuable information closely related to the national economic activity safety is contained, and the information provides enough support for the acquisition of public security monitored information. Through the information, the roles of production, processing, agency, sale, broker, transportation and the like of the suspected target object in the criminal activity are analyzed and judged, and powerful support can be provided for accurately fighting various economic illegal criminal activities through detection work.
In the traditional method, a surveyor needs to perform manual analysis and judgment based on new converged data and historical case data through a first-line detection to determine the role and weight of a suspected target related to a case. In addition, some invalid redundant information is mixed in valuable data, and a line scout staff wants to manually find various data related to cases from an information system and determine the suspected target roles and weights of the cases, which are very difficult, and all of the cases cause the increase of manpower and the reduction of efficiency. The method for determining roles and weights based on the original information detection system often results in incomplete final analysis and lack of real-time accuracy, so that the attack opportunity is delayed, economic criminals can be seriously caused to continuously act and continuously crime, the situation becomes an inertial criminal, huge economic loss is caused to the country, and harmony and stability of the society are seriously influenced.
Therefore, the role calibration capability and the weight evaluation effect of the investigation information system on the suspected case target are improved, the traditional method of multilayer data filtering and manual analysis and study based on a manual mode is changed, a high-speed and intelligent method is constructed, the role perception and the weight calculation of the suspected case target are automatically carried out in real time, the accuracy and the speed of information processing can be greatly improved, and the investigation information analysis and study platform can be developed towards the direction of intellectualization and high accuracy.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a system and a method for suspect case-related role calibration and role evaluation in detection work, which can intelligently analyze and judge various network data gathered by a platform in real time.
In order to achieve the above object, the system and method for realizing suspect case-involved role calibration and role evaluation in the detection work of the present invention comprises the following components:
the method for realizing suspect involvement role calibration and weight evaluation in the detection work is mainly characterized by comprising the following steps of:
(1) the data normalization processing module is used for preprocessing the historical data set and the newly-entered converged data set and storing the data subjected to tagging processing through the database module after the normalized data set is generated;
(2) constructing a digital feature matrix FM and a role re-identification model ECI _ R;
(3) constructing and training a correlation analysis deep neural network (CRDN) to complete the role attribute calibration of the standard data set;
(4) establishing a relation network atlas CR by taking the role attribute mark as a centerHISSet；
(5) And constructing a weight evaluation model ECI _ W for weight evaluation.
Before the step (1) of the method for realizing the suspect involvement role calibration and weight evaluation in the detection work, the method also comprises the following steps:
(0) the historical data set is aggregated with the new aggregated data set.
In step (1), the normative dataset includes a text normative dataset TNDS and a digital normative dataset DNDS, wherein the text normative dataset is processed as follows:
(1.1) performing labeling processing on the text specification data set TNDS;
(1.2) analyzing the text specification data set TNDS subjected to the labeling processing by a semantic analysis extraction module, and extracting semantic features to generate a semantic analysis matrix SAM;
(1.3) constructing a semantic understanding model SC Mod, and entering the step (2) after finishing calibrating the initialized role attributes of the tagged text specification data set TNDS.
The method for realizing suspect involvement role calibration and weight evaluation in the detection work, which comprises the step (1.2) of generating a semantic analysis matrix SAM, comprises the following steps:
(1.2.1) extracting an original transfer record text RTT and an original shopping record text RST in the text specification data set TNDS after the labeling processing, and completing the extraction and generation of candidate keywords of the original transfer record text RTT and the original shopping record text RST;
(1.2.2) sorting the candidate keywords, and selecting the candidate keywords in a first preset threshold range as document keywords;
(1.2.3) training a translation probability model for the document keywords through alignment;
(1.2.4) training a mapping probability model SUB _ AR between the document keywords and the tagged text specification data set TNDS through the translation probability model;
(1.2.5) generating a semantic analysis matrix SAM through the mapping probability model SUB _ AR.
In step (2), the digital feature matrix FM includes a fund transaction record feature matrix FFM, an online shopping record feature matrix SFM, a logistics record feature matrix LFM, a communication record feature matrix CFM and an instant communication message record feature matrix RTFM; the step of constructing the role re-identification model ECI _ R specifically comprises the following steps:
inputting a fund transaction record characteristic matrix FFM, an online shopping record characteristic matrix SFM, a logistics record characteristic matrix LFM, a communication record characteristic matrix CFM and an instant communication message record characteristic matrix RTFM into a decision tree XG Boost, and sampling and training the decision tree XG Boost to obtain the output of a role re-identification model ECI _ R.
In step (5) of the method for realizing suspect involvement role calibration and weight evaluation in the detection work, the step of constructing the weight evaluation model ECI _ W specifically comprises the following steps:
and (4) calculating a weighted average value of the total times and total indexes of a fund transaction record characteristic matrix FFM, an online shopping record characteristic matrix SFM, a logistics record characteristic matrix LFM, a communication record characteristic matrix CFM and an instant communication message record characteristic matrix RTFM to complete the construction of the weight evaluation model ECI _ W.
The system for realizing suspect case-involved role calibration and role evaluation in the detection work is mainly characterized by comprising the following steps of:
the data normalization processing module is used for carrying out normalization processing on the acquired data;
the semantic analysis and extraction module is used for carrying out standardization processing on relevant data of the suspect and extracting semantic content according to the standardized data;
the database module is connected with the data normalization processing module and the semantic extraction module and is used for storing corrected data and the extracted semantic content;
the role re-identification server ECI _ R is connected with the database module and is used for providing case-related role calibration service;
and the weight evaluation model ECI _ W is connected with the role re-identification server ECI _ R and the database module and is used for providing weight evaluation service.
The system and the method for realizing suspect case-related role calibration and role evaluation in the investigation work can provide powerful support for the investigation work. The method has the advantages of realizing efficient filtering and cleaning of the detected data, providing a flexible, rapid and active response case handling environment for detection personnel, striving to catch the best fighter for case detection, improving the case solving efficiency by using the data, fully mining the value of various fusion data, enhancing the credibility of people to the case solving of public security institutions, improving the confidence of the case handling policemen to detection tasks and having good social benefits. Meanwhile, the popularization and application of the system can provide a supporting effect for the safe operation of the national economic system, and indirectly can bring about more remarkable economic benefits.
Drawings
FIG. 1 is a schematic flow chart of the method for realizing suspect case-involved role calibration and role evaluation in the detection work according to the present invention.
Fig. 2 is a design schematic diagram of a correlation analysis deep neural network CRDN according to the present invention.
Detailed Description
In order to more clearly describe the technical contents of the present invention, the following further description is given in conjunction with specific embodiments.
Fig. 1 is a schematic flow chart showing the method for calibrating and evaluating the role of the suspected person in the detected work according to the present invention.
The method for realizing suspect involvement role calibration and weight evaluation in the detection work is mainly characterized by comprising the following steps of:
(1) the data normalization processing module is used for preprocessing the historical data set and the newly-entered converged data set and storing the data subjected to tagging processing through the database module after the normalized data set is generated;
(2) constructing a digital feature matrix FM and a role re-identification model ECI _ R;
(3) constructing and training a correlation analysis deep neural network (CRDN) to complete the role attribute calibration of the standard data set;
(4) establishing a relation network atlas CR by taking the role attribute mark as a centerHISSet；
(5) And constructing a weight evaluation model ECI _ W for weight evaluation.
Before the step (1) of the method for realizing the suspect involvement role calibration and weight evaluation in the detection work, the method also comprises the following steps:
(0) the historical data set is aggregated with the new aggregated data set.
In step (1), the normative Data set includes a text normative Data set tnds (text Normal Data set) and a digital normative Data set dnds (digital Normal Data set), wherein the text normative Data set is processed as follows:
(1.1) performing labeling processing on the text specification data set TNDS;
(1.2) analyzing the text specification data set TNDS after the labeling processing by a semantic Analysis extraction module, and extracting semantic features to generate a semantic Analysis matrix SAM (semantic Analysis matrix);
(1.3) constructing a Semantic understanding Model SC Mod (Semantic Comprehensive Model), and entering the step (2) after finishing the initialized role attribute calibration of the tagged text specification data set TNDS.
The method for realizing suspect involvement role calibration and weight evaluation in the detection work, which comprises the step (1.2) of generating a semantic analysis matrix SAM, comprises the following steps:
(1.2.1) extracting an original Transfer record text RTT (raw Transfer text) and an original Shopping record text RST (raw Shopping text) in the text specification data set TNDS after the labeling processing, and completing the extraction and generation of candidate keywords of the original Transfer record text RTT and the original Shopping record text RST;
(1.2.2) sorting the candidate keywords, and selecting the candidate keywords in a first preset threshold range as document keywords;
(1.2.3) training a translation probability model for the document keywords through alignment;
(1.2.4) training a mapping probability model SUB _ AR between the document keywords and the tagged text specification data set TNDS through the translation probability model;
(1.2.5) generating a semantic analysis matrix SAM through the mapping probability model SUB _ AR.
In step (2), the digital feature matrix FM includes a fund transaction record feature matrix FFM, an online shopping record feature matrix SFM, a logistics record feature matrix LFM, a communication record feature matrix CFM and an instant communication message record feature matrix RTFM; the step of constructing the role re-identification model ECI _ R specifically comprises the following steps:
inputting a fund transaction record characteristic matrix FFM, an online shopping record characteristic matrix SFM, a logistics record characteristic matrix LFM, a communication record characteristic matrix CFM and an instant communication message record characteristic matrix RTFM into a decision tree XG Boost, and sampling and training the decision tree XG Boost to obtain the output of a role re-identification model ECI _ R.
In step (5) of the method for realizing suspect involvement role calibration and weight evaluation in the detection work, the step of constructing the weight evaluation model ECI _ W specifically comprises the following steps:
the construction of the weight evaluation model ECI _ W is completed by calculating the weighted average of the total times and the total indexes of a capital transaction record characteristic matrix FFM (found Feature matrix), an online shopping record characteristic matrix SFM (shopping Feature matrix), a logistics record characteristic matrix LFM (logistics Feature matrix), a communication record characteristic matrix CFM (call Feature matrix), and an instant communication Message record characteristic matrix RTFM (instant Message Feature matrix).
The system for realizing suspect case-involved role calibration and role evaluation in the detection work is mainly characterized by comprising the following steps of:
the data normalization processing module is used for carrying out normalization processing on the acquired data;
the semantic analysis and extraction module is used for carrying out standardization processing on relevant data of the suspect and extracting semantic content according to the standardized data;
the database module is connected with the data normalization processing module and the semantic extraction module and is used for storing corrected data and the extracted semantic content;
the role re-identification server ECI _ R is connected with the database module and is used for providing case-related role calibration service;
and the weight evaluation model ECI _ W is connected with the role re-identification server ECI _ R and the database module and is used for providing weight evaluation service.
In one embodiment, the method for implementing role calibration and role evaluation of suspected person under investigation in the process of detection work comprises the following steps:
(1) role set together role set { production, processing, brokering, sales, brokering, envelope, terminal }. And defining [1,0,0,0,0,0,0] to represent a production role, [1,0,0,1,0,0,0] to represent a production sales role, [0,0,0,0,0,1,0] to represent a package transportation role, and so on;
(2) preprocessing historical data and a newly-entered convergence data set to generate a text specification data set TNDS, and finishing the operations of extracting, removing duplication, converting and aligning data of different formats and types from various data sources;
(3) extracting { RTT, RST };
(4) completing short text keyword extraction and short text semantic topic calculation from { RTT, RST } to obtain SUB _ AR [0..2 ];
(5) and constructing the SCMOD, and recommending AlternativRole [0..2] through a completion tag based on the SUB _ AR [0..2 ]. Let short text sd be e { RTT, RST }. The purpose of keyword extraction is to rank the candidate keywords according to the maximum likelihood estimate of a given document sd, i.e. to compute Pr (P | sd) for all candidate keywords P ∈ P, where P is the set of candidate keywords. The recommendation calculation method for each label is shown as the following formula:
pr (t | d) ═ Σ w ∈ wd (γ it (w)) + (1- γ) Pr (t | w)) Pr (w | d) (formula one)
In the formula i, it (w) is an indicator function (indicator function), when t is w, that is, when the word in the document is the same as the label, the value of the indicator function is 1, and when t is not equal to w, the value of the indicator function is 0, and γ is a smoothing factor, and the value range is [ 0.0,1.0 ]. The top 3 tags are selected as recommended tags by the system predefined tags.
(6) Constructing a digital feature matrix DNDSHIS(FFM, SFM, LFM, CFM, RTFM), where FFM is a capital transaction record feature matrix, SFM shopping over the internet record feature matrix, LFM is a logistics record feature matrix, CFM is a communications record feature matrix, and RTFM isThe definition of the time-lapse messaging record feature matrix is as follows,
FFM ═ PID _ local, PID _ recip, total number of trades, total revenue trades, total expenditure trades, SUB _ AR [0..2], trade distribution [186] in the last 6 months, type of trade ═ 1: the bank card is matched with the card; 2: the bank card is withdrawn; 3 Payment Pao; 4, the financial payment is to the financial payment; 5, transferring accounts through WeChat; 6 others }; wherein the fund transaction scenario of the last 6 months is defined as every transaction that occurs;
SFM { PID _ local, PID _ recip, total times of sales, total times of shopping, 0, SUB _ AR [0..2], buy and sell distribution [186] in the last 6 months;
LFM { (PID _ local, PID _ recip, total number of posting, total number of receiving, 0,0,0, express distribution of the last 6 months [186] };
CFM { PID _ local, PID _ recip, total number of callers, total number of callees, 0,0,0, call distribution in the last 6 months [186] };
RTFM { PID _ local, PID _ recip, total number of callers, total number of callees, 0, SUB _ AR [0..2], instant messaging profile [186] of nearly 6 months.
Further, the distribution is counted according to 31 days per month, 0 is filled in nonexistent days, 1 is recorded once in each statistical condition, and cumulative addition calculation is carried out for multiple times.
Further, the CRDN is constructed in the step (6), and the CRDN is trained through FFM, SFM, LFM, CFM and RTFM data to obtain a crmod (correlation model) [ FFM.
TABLE 1 parameters for constructing CRDN and details thereof
| Inputting data | Training method | Output model | Model class parameters | Model computation output layer features |
| FFM | CRDN | CRMod[ffm] | ffm | ModFM[ffm] |
| SFM | CRDN | CRMod[sfm] | sfm | ModFM[sfm] |
| LFM | CRDN | CRMod[lfm] | lfm | ModFM[lfm] |
| CFM | CRDN | CRMod[cfm] | cfm | ModFM[cfm] |
| RTFM | CRDN | CRMod[rtfm] | rtfm | ModFM[rtfm] |
In one embodiment, please refer to fig. 2, which is a schematic design diagram of a Deep neural network CRDN (correlation Deep net) for correlation analysis according to the present invention, wherein the Deep neural network CRDN is adopted to train CRDN data to obtain a ModFM model, and the implementation key points of the ModFM model include the following aspects:
(1) and extracting the output characteristics of the basic unit through 3 sequentially cascaded characteristics. The setup method of the convolution layers of 3 sequentially cascaded feature extraction basic units is as follows:
in the first feature extraction basic unit, the size of a convolution kernel is 3, the convolution kernel is used for extracting features with large time span, and the output feature dimension is 32;
in the second feature extraction basic unit, the size of a convolution kernel is 2, the convolution kernel is used for extracting features with medium size in time span, and the output feature dimension is 64;
(2) the first layer performs weighted calculation on FM based on convolution kernels with the size of 3 to obtain a primary feature vector;
(3) the second layer performs pooling operations based on a filter of size 2;
(4) the third layer performs weighted calculation on the eigenvector output by the first layer based on the volume set core with the size of 2 to obtain a second-level eigenvector;
(5) the fourth layer performs pooling operation based on a filter of size 2;
(6) the fifth layer and the sixth layer are full connection layers, and the seventh layer is a SOFTMAX layer;
(7) the integration method cascades a seventh layer, and the output of the seventh layer is Mod FM.
Further, in the step (6), ModFM [ ffm ], ModFM [ sfm ], ModFM [ lfm ], ModFM [ cfm ], and ModFM [ rtfm ] are used as an input data set of XGBoost, boosting sampling is performed on the set, equal weight 1/n is assigned to each training example during initialization, then the training set is trained for t rounds by the mathematical algorithm, after each training, training examples with failed training are assigned with larger weight, that is, the learning algorithm learns training examples with difficult comparison in subsequent learning, so as to obtain a prediction function sequence H _ 1. A decision tree XGboost is respectively established for a plurality of boosting sample sets, and a regular term is added into a cost function of the XGboost and is used for controlling the complexity of a model. The regular term includes the number of leaf nodes of the tree, and score output at each leaf node is the sum of squares of the L2 modules. To limit the growth of the tree, we can add a threshold, which is a coefficient of the number T of leaf nodes in the regularization term, so that xgboost is equivalent to pre-pruning while optimizing the objective function. And finally obtaining the output of the model by a voting method after the training of the single decision tree is finished, and completing the construction of the role re-identification model ECI _ R by the process.
Further, the method for calculating the weight of the role calibration object is defined as follows
The method is characterized in that a role calibration object ROBJ (proportion integration differentiation, PID) is used as an alternative role [0]The central weight matrix of the network link is expanded, wherein, PID (personal Identity number) is the person identification ID, wherein
In (3), w is calculated as follows:
wherein, F (FM) is a weighted average calculation method for FFM, SFM, LFM, CFM and RTFM total times and total amount indexes, and the specific calculation method is as follows:
the calculation process of the W value completes the construction of the weight evaluation model ECI _ W. And completing the calculation of ECI _ R and ECI _ W to obtain a character calibration and weight evaluation color input matrix, ECI _ RW ═ { ECI _ R, ECI _ W }.
The system and the method for realizing suspect case-related role calibration and role evaluation in the investigation work can provide powerful support for the investigation work. The method has the advantages of realizing efficient filtering and cleaning of the detected data, providing a flexible, rapid and active response case handling environment for detection personnel, striving to catch the best fighter for case detection, improving the case solving efficiency by using the data, fully mining the value of various fusion data, enhancing the credibility of people to the case solving of public security institutions, improving the confidence of the case handling policemen to detection tasks and having good social benefits. Meanwhile, the popularization and application of the system can provide a supporting effect for the safe operation of the national economic system, and indirectly can bring about more remarkable economic benefits
In this specification, the invention has been described with reference to specific embodiments thereof. It will, however, be evident that various modifications and changes may be made thereto without departing from the broader spirit and scope of the invention. The specification and drawings are, accordingly, to be regarded in an illustrative rather than a restrictive sense.
Claims (5)
1. A method for realizing suspect case-involved role calibration and weight evaluation in the process of detection work is characterized in that a data standardization processing module preprocesses a historical data set and a newly-entered converged data set, and after a standard data set is generated, the role attribute calibration of personnel is completed through database case data, and the method comprises the following steps:
(1) constructing a digital feature matrix FM and a role re-identification model ECI _ R;
(2) constructing and training a correlation analysis deep neural network (CRDN) to complete the role attribute calibration of the standard data set;
(3) establishing a relation network atlas CR by taking the role attribute mark as a centerHISSet；
(4) Constructing a weight evaluation model ECI _ W for weight evaluation;
in the step (1), the digital feature matrix FM includes a fund transaction record feature matrix FFM, an online shopping record feature matrix SFM, a logistics record feature matrix LFM, a communication record feature matrix CFM, and an instant messaging message record feature matrix RTFM;
the step (1) of constructing the role re-identification model ECI _ R specifically comprises the following steps:
inputting a fund transaction record characteristic matrix FFM, an online shopping record characteristic matrix SFM, a logistics record characteristic matrix LFM, a communication record characteristic matrix CFM and an instant communication message record characteristic matrix RTFM into a decision tree XG Boost, and sampling and training the decision tree XG Boost to obtain the output of a role re-identification model ECI _ R;
the step (2) of constructing and training a correlation analysis deep neural network (CRDN) specifically comprises the following steps:
training data in the input fund transaction record characteristic matrix FFM, the online shopping record characteristic matrix SFM, the logistics record characteristic matrix LFM, the communication record characteristic matrix CFM and the instant communication message record characteristic matrix RTFM to obtain a deep neural network CRDN;
in the step (4), the step of constructing the weight evaluation model ECI _ W specifically includes:
and (4) calculating a weighted average value of the total times and total indexes of a fund transaction record characteristic matrix FFM, an online shopping record characteristic matrix SFM, a logistics record characteristic matrix LFM, a communication record characteristic matrix CFM and an instant communication message record characteristic matrix RTFM to complete the construction of the weight evaluation model ECI _ W.
2. The method for realizing suspect case-involved role calibration and weight evaluation in the spy work according to claim 1, wherein the step (1) is preceded by the following steps:
(0) the historical data set is aggregated with the new aggregated data set.
3. The method of claim 1, wherein the normative dataset comprises a text normative dataset TNDS and a numerical normative dataset DNDS, wherein the text normative dataset is processed by:
(a1) performing labeling processing on the text specification data set TNDS;
(a2) a semantic analysis extraction module analyzes the tagged text specification data set TNDS and extracts semantic features to generate a semantic analysis matrix SAM;
(a3) and (3) constructing a semantic understanding model SCMOD, and entering the step (1) after finishing the initialization role attribute calibration of the tagged text specification data set TNDS.
4. The method for performing suspect case role assignment and weight evaluation through spy work according to claim 3, wherein said step (a2) of generating a semantic analysis matrix SAM comprises the following steps:
(a2.1) extracting an original transfer record text RTT and an original shopping record text RST in the text specification data set TNDS after the labeling processing, and finishing the extraction and generation of candidate keywords of the original transfer record text RTT and the original shopping record text RST;
(a2.2) sorting the candidate keywords, and selecting the candidate keywords in a first preset threshold range as document keywords;
(a2.3) training a translation probability model for the document keywords through alignment;
(a2.4) training a mapping probability model SUB _ AR between the document keywords and the tagged text specification data set TNDS through the translation probability model;
(a2.5) generating a semantic analysis matrix SAM through the mapping probability model SUB _ AR.
5. A system for performing suspect case-involved role calibration and role evaluation in a detected job by using the method of claim 1, wherein the system comprises:
the data normalization processing module is used for carrying out normalization processing on the acquired data;
the semantic analysis and extraction module is used for carrying out standardization processing on relevant data of the suspect and extracting semantic content according to the standardized data;
the database module is connected with the data normalization processing module and the semantic extraction module and is used for storing corrected data and the extracted semantic content;
the role re-identification server ECI _ R is connected with the database module and is used for providing case-related role calibration service;
and the weight evaluation model ECI _ W is connected with the role re-identification server ECI _ R and the database module and is used for providing weight evaluation service.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810291232.5A CN108520042B (en) | 2018-04-03 | 2018-04-03 | System and method for realizing suspect case-involved role calibration and role evaluation in detection work |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810291232.5A CN108520042B (en) | 2018-04-03 | 2018-04-03 | System and method for realizing suspect case-involved role calibration and role evaluation in detection work |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108520042A CN108520042A (en) | 2018-09-11 |
| CN108520042B true CN108520042B (en) | 2022-02-08 |
Family
ID=63431076
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810291232.5A Active CN108520042B (en) | 2018-04-03 | 2018-04-03 | System and method for realizing suspect case-involved role calibration and role evaluation in detection work |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108520042B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570148A (en) * | 2016-10-27 | 2017-04-19 | 浙江大学 | Convolutional neutral network-based attribute extraction method |
| CN107577782A (en) * | 2017-09-14 | 2018-01-12 | 国家计算机网络与信息安全管理中心 | A kind of people-similarity depicting method based on heterogeneous data |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8452725B2 (en) * | 2008-09-03 | 2013-05-28 | Hamid Hatami-Hanza | System and method of ontological subject mapping for knowledge processing applications |
| US8375032B2 (en) * | 2009-06-25 | 2013-02-12 | University Of Tennessee Research Foundation | Method and apparatus for predicting object properties and events using similarity-based information retrieval and modeling |
-
2018
- 2018-04-03 CN CN201810291232.5A patent/CN108520042B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570148A (en) * | 2016-10-27 | 2017-04-19 | 浙江大学 | Convolutional neutral network-based attribute extraction method |
| CN107577782A (en) * | 2017-09-14 | 2018-01-12 | 国家计算机网络与信息安全管理中心 | A kind of people-similarity depicting method based on heterogeneous data |
Non-Patent Citations (1)
| Title |
|---|
| "犯罪学标签理论的应然走向";李明琪等;《中国人民公安大学学报（社会科学版）》;20120615;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108520042A (en) | 2018-09-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110223168B (en) | Label propagation anti-fraud detection method and system based on enterprise relationship map | |
| CN108876600B (en) | Early warning information pushing method, device, computer equipment and medium | |
| WO2020037942A1 (en) | Risk prediction processing method and apparatus, computer device and medium | |
| WO2020119272A1 (en) | Risk identification model training method and apparatus, and server | |
| CN110704572B (en) | Suspected illegal fundraising risk early warning method, device, equipment and storage medium | |
| CN110852856B (en) | Invoice false invoice identification method based on dynamic network representation | |
| CN107657267B (en) | Product potential user mining method and device | |
| CN112053221A (en) | Knowledge graph-based internet financial group fraud detection method | |
| EP3642766A1 (en) | Machine-learning system for servicing queries for digital content | |
| CN112700325A (en) | Method for predicting online credit return customers based on Stacking ensemble learning | |
| CN109102157A (en) | A kind of bank's work order worksheet processing method and system based on deep learning | |
| CN112132233A (en) | Criminal personnel dangerous behavior prediction method and system based on effective influence factors | |
| CN111199474A (en) | Risk prediction method and device based on network diagram data of two parties and electronic equipment | |
| CN112989059A (en) | Method and device for identifying potential customer, equipment and readable computer storage medium | |
| CN112053222A (en) | Knowledge graph-based internet financial group fraud detection method | |
| US20230367821A1 (en) | Machine-learning system for servicing queries for digital content | |
| CN112330328A (en) | Credit card fraud detection method based on feature extraction | |
| CN116402512B (en) | Account security check management method based on artificial intelligence | |
| CN115659966A (en) | Rumor detection method and system based on dynamic heteromorphic graph and multi-level attention | |
| CN115545886A (en) | Overdue risk identification method, overdue risk identification device, overdue risk identification equipment and storage medium | |
| CN116304035A (en) | Multi-notice multi-crime name relation extraction method and device in complex case | |
| CN114387005A (en) | Arbitrage group identification method based on graph classification | |
| CN116821759A (en) | Identification prediction method and device for category labels, processor and electronic equipment | |
| CN107609921A (en) | A kind of data processing method and server | |
| CN117114705A (en) | Continuous learning-based e-commerce fraud identification method and system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |