CN109543986A - The pre- methods of risk assessment of prison convict three and system based on user's portrait - Google Patents
The pre- methods of risk assessment of prison convict three and system based on user's portrait Download PDFInfo
- Publication number
- CN109543986A CN109543986A CN201811363792.3A CN201811363792A CN109543986A CN 109543986 A CN109543986 A CN 109543986A CN 201811363792 A CN201811363792 A CN 201811363792A CN 109543986 A CN109543986 A CN 109543986A
- Authority
- CN
- China
- Prior art keywords
- criminal
- risk assessment
- lightgbm
- disaggregated model
- xgboost
- 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
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- 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
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
Abstract
The invention belongs to machine learning techniques fields, and in particular to a kind of pre- methods of risk assessment of prison convict three and system based on user's portrait.This method comprises: step 1, obtaining from Prison Management System Developed the initial data of criminal, and data cleansing is carried out, obtain sample data；Step 2, analysis sample data establish criminal user's portrait label；Step 3, foundation prison officer form the pre- risk assessment label of criminal three to the artificial risk assessment situation of criminal；Step 4, drawn a portrait according to the criminal user label and the pre- risk assessment label of criminal three, training XGBoost, LightGBM disaggregated model；Step 5 is obtained to risk assessment data, carries out three pre- risk assessment using XGBoost, LightGBM disaggregated model after training.The present invention also provides a kind of pre- risk evaluating systems of prison convict three based on user's portrait.Assessment result of the present invention is more accurate, can be improved the accuracy of criminal's risk assessment, and can timely early warning, effectively prevent accident.
Description
Technical field
The invention belongs to machine learning techniques fields, and in particular to a kind of pre- risk of prison convict three based on user's portrait
Appraisal procedure and system.
Background technique
Prison is the place put in prison, reformed the wrong-doers, and is important one of state apparatus.How prison has been managed by superintendent
Member, prevention occur that person under surveillance three is pre-, that is, run away, violence, all kinds of influence supervision places safety and stability such as suicide accident, structure
The safe precaution system for building up supervision is extremely important.It, can be to being supervised by carrying out three pre- risk assessment to person under surveillance
The running away of personnel, violence and suicide risk are assessed.
At present when carrying out three pre- risk assessment to prison convict, subjectivity often is carried out according to the experience of basic-level policemen and is commented
Sentence, this traditional passive type manually monitors, regulatory format has been unable to satisfy supervision demand under the new situation, especially to prominent
Hair malignant event can not accomplish the prevention and processing of first time.
Prison system is mounted with a large amount of digitizers, produces in daily life and management from personal information, daily
Behavior, the data system for the multidimensional all standing such as checking and supervising.Multidimensional data has recorded long-term a large amount of behaviors of criminal, the criminal of foring
The various dimensions tag set of people, i.e. user's portrait of convict.The risk assessment of convict is substantially a classification identification process, therefore
User's Portrait brand technology applied to the credit scoring of enterprise, fraud identification, lean operation is equally applicable to convict's risk assessment.
Summary of the invention
In order to solve the above technical problems, the present invention provides a kind of criminal's Risk Forecast Method based on user's portrait, lead to
User's Portrait brand technology is crossed, person under surveillance's label system is established；It is predicted on this basis by machine learning method by superintendent
Member commit suiside, run away, the risk degree of violence, for the later period implementation Supervision Measures decision support is provided.To achieve the goals above,
The technical solution adopted by the present invention is as follows.
A kind of pre- methods of risk assessment of prison convict three based on user's portrait, comprising the following steps:
Step 1 obtains sample data: the initial data of criminal obtained from Prison Management System Developed, and carries out data cleansing,
Obtain sample data；
Step 2, analysis sample data establish criminal user's portrait label, and criminal user's portrait label includes nature
ATTRIBUTE INDEX, is supervising situation index and psychologic status index at criminal information index；
Step 3, foundation prison officer form the pre- risk assessment mark of criminal three to the artificial risk assessment situation of criminal
Label, the pre- risk assessment label of the criminal three includes normal tag, label of committing suiside, label of running away, violence label；
Step 4, drawn a portrait according to the criminal user label and the pre- risk assessment label of criminal three, training XGBoost,
LightGBM disaggregated model；
Step 5, the initial data that criminal is obtained from Prison Management System Developed, by data cleansing, as to risk assessment
Data carry out three pre- risk assessment using XGBoost, LightGBM disaggregated model after training.
Preferably, in step 4, training pattern XGBoost, LightGBM disaggregated model specifically comprises the following steps:
Step 41, sequentially in time using before sample data 80% data as training set, will be every in training set
Input of a criminal user's portrait label as XGBoost, LightGBM disaggregated model, by the described three pre- risks of each criminal
Assessment tag is as model label, training XGBoost, LightGBM disaggregated model parameter；
Step 42 selects after sample data 20% data as test set, after training sequentially in time
XGBoost, LightGBM disaggregated model carry out three pre- risk assessment to the sample data in test set, obtain three pre- risk assessment
As a result；
Step 43, the three pre- wind exported according to the described three pre- risk assessment labels and step 42 of criminal each in test set
Dangerous assessment result calculates the classification accuracy and recall rate of XGBoost, LightGBM disaggregated model；Judge classification accuracy and
The size relation of recall rate and setting value, if classification accuracy and recall rate are to terminate to train more than or equal to setting value
Process；If any one of classification accuracy and recall rate value are less than setting value, 44 are thened follow the steps；
Step 44 carries out upper and lower sampling by the training data to XGBoost, LightGBM disaggregated model with right
XGBoost, LightGBM disaggregated model parameter are adjusted, and use three pre- risks of XGBoost, LightGBM disaggregated model
The result of assessment is merged, and step 41 to 43 is repeated, until point for XGBoost, LightGBM disaggregated model that training obtains
Class accuracy rate and recall rate are all larger than equal to setting value.
Preferably, the calculation formula of the classification accuracy and recall rate is as follows:
P=TP/ (TP+FP)
R=TP/ (TP+FN)
Wherein, P presentation class accuracy rate, R indicate recall rate, and TP is that positive class determines to be positive the quantity of class in test set, FP
The quantity of class of being positive is determined for class negative in test set, and FN is that positive class determines to be negative the quantity of class in test set.
The present invention also provides a kind of pre- risk evaluating system of prison convict three based on user's portrait, which includes:
Sample data obtains module, for counting to the various initial data for obtaining criminal from the management system in prison
According to cleaning, sample data is obtained；
User's portrait label establishes module, for establishing criminal user's portrait label according to the sample data；
Three pre- risk assessment labels form module, are drawn for prison officer according to the sample data and the user
As label filtration is used to carry out the data of criminal's risk assessment, the three pre- risk assessment labels of criminal are formed；
Disaggregated model training module, for according to described three pre- risk assessment labels and user portrait label training
XGBoost, LightGBM disaggregated model；
Criminal's risk evaluation module carries out three pre- risk assessment to data to be assessed, uses for obtaining data to be assessed
XGBoost, LightGBM disaggregated model after training carries out three pre- risk assessment.
Preferably, the disaggregated model training module includes:
Disaggregated model parameter training module, for selecting before sample data sequentially in time 80% as training set,
By input of each criminal user portrait as XGBoost, LightGBM disaggregated model in training set, by the institute of each criminal
Three pre- risk assessment labels are stated as model label, training XGBoost, LightGBM disaggregated model parameter；
Computing module, for selecting rear the 20% of sample data sequentially in time as test set, after training
XGBoost, LightGBM disaggregated model carry out three pre- risk assessment to the sample data in test set, obtain three pre- risk assessment
As a result；
Determining module, for being exported according to the described three pre- risk assessment labels and computing module of criminal each in test set
Three pre- risk evaluation results determine whether the classification accuracy of XGBoost, LightGBM disaggregated model and recall rate reach and set
Definite value；
Module is adjusted, for being adopted above and below by the training data to XGBoost, LightGBM disaggregated model
Sample uses the three of XGBoost, LightGBM disaggregated model to be adjusted to XGBoost, LightGBM disaggregated model parameter
The result of pre- risk assessment is merged, the classification accuracy of XGBoost, LightGBM disaggregated model that obtains until training and
Recall rate reaches setting value.
Using the present invention obtain the utility model has the advantages that using the method for the present invention carry out criminal's risk assessment, pass through this method will
The much information of criminal is integrated, and the multidimensional data generated in present prison administration is made full use of, and forms the personalization of convict
Description, using big data and artificial intelligence technology, beneficial exploration and practice is implemented in the pre- risk assessment of convict three at the prison.Assessment
As a result relatively more accurate, compared with traditional artificial regulatory format, it can be improved the accuracy of criminal's risk assessment, and can be timely
Early warning effectively prevent accident.The present invention establishes user's portrait label of criminal by user's Portrait brand technology, on this basis
The behavior of criminal is predicted by machine learning method, can implement supervision for the later period and decision support is provided.
Detailed description of the invention
Fig. 1 is the flow chart of the method for the present invention；
Fig. 2 is the structural framing figure of the prison convict three pre- risk evaluating system provided by the invention based on user's portrait；
Fig. 3 is the structural framing figure of disaggregated model training module provided by the invention.
Shown by reference numeral in above-mentioned attached drawing are as follows:
100, the pre- risk evaluating system of prison convict three based on user's portrait, 1, sample data acquisition module, 2, user
Portrait label establishes module, and 3, three pre- risk assessment labels form module, 4, disaggregated model training module, 5, criminal's risk assessment
Module, 41, disaggregated model parameter training module, 42, computing module, 43, determining module, 44, adjustment module.
Specific embodiment
Present invention will be further explained below with reference to the attached drawings and examples.
It is as shown in Figure 1 the flow chart of the method for the present invention, a kind of pre- risk assessment of prison convict three based on user's portrait
Method.
Step 1, it obtains sample data: extracting criminal's basic information, shopping information, mail note from Prison Management System Developed
Record, education memcon, emotional affection telephony recording, emotional affection meet with record, the convict that is put in prison assessment, rewards and punishments score, people's police's report information,
Criminal's report information carries out missing values cleaning, format content cleaning, logic error cleaning etc. as initial data, to initial data
Data cleansing, obtains sample data, and the sample data includes convict's basic information, shopping information, mail record, education talk
Record, emotional affection telephony recording, emotional affection meet with record, the convict that is put in prison assessment, rewards and punishments score, people's police's report information, criminal's report letter
Breath；
Step 2, criminal user's portrait label is established；Sample data is analyzed, forms criminal's label system, including certainly
Right ATTRIBUTE INDEX, is supervising situation index, psychologic status index at criminal information index.Natural quality index mainly includes age, property
Not, before native place, residence, marital status, children's situation, primary family status, schooling, paving occupation, whether have suicide from
Hurt autotomy history；Criminal information index include case by, prison term, cause of criminality, number of serving a sentence, whether clique, whether used recidivist, whether
Concurrence of offences, the age of crime for the first time enter to supervise the age, aberrant biological factor, past medical history, enter to supervise the time etc.；In prison situation index
It is met with including emotional affection mail, emotional affection phone, emotional affection, education talk, rewards and punishments situation, people's police's report, criminal's report, purchase and consumption；
Psychologic status index is subject to criminal and is put in prison and assesses information table.In embodiment, particular content is as shown in table 1.
1 criminal user of table portrait label
Step 3, form the pre- risk assessment label of criminal three: at the prison in management, each convict's every month needs prison to manage
Reason personnel (people's police) carry out multiple risk assessment: i.e. assessment convict is normally, has suicide risk, violence risk, risk of running away.It is real
It is fewer, i.e. normal sample and wind that convict, which is assessed as having suicide, running away with the data of violence risk, in the historical data of border
Dangerous sample proportion is very unbalanced, therefore is assessed as having the data of three pre- risks that can all use in all historical samples.Just
Regular data selection principle are as follows: preferential to select the criminal for being once assessed as having three pre- risks other all are assessed as normal sample
This and other criminals nearlyr time are assessed as normal sample, to form criminal's assessment tag, including normally mark
It signs, label of committing suiside, label of running away, violence label；
Step 4, drawn a portrait according to the criminal user label and the pre- risk assessment label of criminal three, training XGBoost,
LightGBM disaggregated model；XGBoost, LightGBM disaggregated model are the common models in existing machine learning techniques, described
Training XGBoost, LightGBM disaggregated model includes the following steps:
Step 41, it because criminal's risk assessment can only use past data to predict future, selects sequentially in time
Preceding 80% data in sample data are selected as training set, using each criminal user in training set draw a portrait label as
The input of XGBoost, LightGBM model, using each criminal's risk assessment label as model label, training XGBoost,
LightGBM model parameter, using installing xgboost0.71, lightgbm2.1.1 work in python environment in the present embodiment
Have the training of packet implementation model and assessment；
Step 42, using 20% data behind in sample data as test set, each criminal calculated in test set is
Normally, it commits suiside, still violence of running away；Three pre- risk evaluation results are obtained as by disaggregated model after training；
Step 43, according to the pre- risk evaluation result of output three of the label of each criminal in test set and step 42, really
Determine whether the classification accuracy of disaggregated model and recall rate in step 41 reach setting value；If classification accuracy and recall rate are equal
More than or equal to setting value, then to train process to terminate；If any one of classification accuracy and recall rate value are less than setting value,
Execute step 44；
Step 44 carries out upper and lower sampling by the training data to XGBoost, LightGBM disaggregated model with right
XGBoost, LightGBM disaggregated model parameter are adjusted, and use three pre- risks of XGBoost, LightGBM disaggregated model
The result of assessment is merged, and is repeated step 41 and is arrived step 43, until XGBoost, LightGBM disaggregated model that training obtains
Classification accuracy and recall rate be all larger than equal to setting value.
Preferably, the calculation formula of criminal's risk assessment accuracy rate and recall rate is as follows:
P=TP/ (TP+FP)
R=TP/ (TP+FN)
Wherein, TP is that positive class determines to be positive the quantity of class in test set, and FP is that negative class determines to be positive the number of class in test set
Amount, FN are that positive class determines to be negative the quantity of class in test set.It is with the accuracy rate and recall rate for risk of committing suiside in specific embodiment
Example, TP are that all suicide risks that are judged as are really to have the quantity of suicide risk in test set, and FP is practical in test set
Risk of not committing suiside is judged as the quantity of suicide risk, and FN is actually to have suicide risk to be judged as not certainly in test set
Kill the quantity of risk.
As shown in Fig. 2, the present invention also provides a kind of pre- risk evaluating systems of prison convict three based on user's portrait
100, which includes that sequentially connected sample data obtains module 1, user's portrait label establishes module 2, three pre- risk assessment
Label forms module 3, disaggregated model training module 4, criminal's risk evaluation module 5.It is provided by the invention described based on user's picture
The pre- methods of risk assessment of prison convict three of picture is applicable to the pre- risk assessment system of prison convict three based on user's portrait
System 100.
The sample data obtain module 1 be used for obtained from the management system in prison the various initial data of criminal into
Row data cleansing, obtains sample data.Wherein, various initial data include criminal's basic information, shopping information, mail record,
Educate memcon, emotional affection telephony recording, emotional affection meeting record, the convict that is put in prison assessment, rewards and punishments score, talk information, people's police's remittance
It notifies breath and criminal's report information.
User's portrait label establishes module 2 for establishing criminal user's portrait label according to the sample data.Its
In, user label of drawing a portrait includes natural quality index, criminal information index, refers in prison situation index and psychologic status
Mark.
The three pre- risk assessment label forms module 3 and is used for label sieve of drawing a portrait according to the sample data and the user
It is selected to carry out the data of criminal's risk assessment, forms the three pre- risk assessment labels of criminal.Wherein, described three pre- risk assessment
Label includes normal tag, label of committing suiside, run away label and violence label.
The disaggregated model training module 4, for according to described three pre- risk assessment labels and user portrait label
Training XGBoost, LightGBM disaggregated model；
Criminal's risk evaluation module 5 for obtaining data to be assessed, using after training XGBoost,
LightGBM disaggregated model carries out three pre- risk assessment.
As shown in figure 3, the disaggregated model training module 4 include disaggregated model parameter training module 41, computing module 42,
Determining module 43 and adjustment module 44.
Before in user's portrait of the disaggregated model parameter training module 41 for selecting criminal sequentially in time
80% is used as training set, and the user of each criminal in training set is drawn a portrait as the defeated of XGBoost, LightGBM disaggregated model
Enter, using described the three of each criminal pre- risk assessment labels as model label, training XGBoost, LightGBM disaggregated model
Parameter.
The computing module 42 is used to select sequentially in time rear 20% in user's portrait of criminal as test set,
Three pre- risk assessment are carried out to the sample data in test set using XGBoost, LightGBM disaggregated model after training, are obtained
Three pre- risk evaluation results；
The determining module 43 is used for the described three pre- risk assessment labels according to criminal each in test set and calculates mould
The calculated result of block output determines whether the classification accuracy of XGBoost, LightGBM disaggregated model and recall rate reach setting
Value.
The adjustment module 44 for carried out by the training data to XGBoost, LightGBM disaggregated model,
Down-sampling uses XGBoost, LightGBM disaggregated model to be adjusted to XGBoost, LightGBM disaggregated model parameter
The results of three pre- risk assessment merged, until the classification for XGBoost, LightGBM disaggregated model that training obtains is accurate
Rate and recall rate reach setting value.The present invention obtains sample number after carrying out data cleansing by the various initial data to criminal
According to, based on sample data training XGBoost, LightGBM disaggregated model, and criminal's risk assessment is carried out by Model Fusion,
So that assessment result is more acurrate, according to three pre- risk assessment as a result, can give warning in advance, it effectively prevent accident.
Contain the explanation of the preferred embodiment of the present invention above, this be for the technical characteristic that the present invention will be described in detail, and
Be not intended to for summary of the invention being limited in concrete form described in embodiment, according to the present invention content purport carry out other
Modifications and variations are also protected by this patent.The purport of the content of present invention is to be defined by the claims, rather than by embodiment
Specific descriptions are defined.
Claims (7)
1. a kind of pre- methods of risk assessment of prison convict three based on user's portrait, it is characterised in that: the following steps are included:
Step 1 obtains sample data: obtaining the initial data of criminal from Prison Management System Developed, and carries out data cleansing, obtains
Sample data；
Step 2, analysis sample data establish criminal user's portrait label, and criminal user's portrait label includes natural quality
Index, is supervising situation index and psychologic status index at criminal information index；
Step 3, foundation prison officer form the pre- risk assessment label of criminal three to the artificial risk assessment situation of criminal,
The pre- risk assessment label of the criminal three includes normal tag, label of committing suiside, label of running away, violence label；
Step 4, drawn a portrait according to the criminal user label and the pre- risk assessment label of criminal three, training XGBoost, LightGBM
Disaggregated model；
Step 5, the initial data that criminal is obtained from Prison Management System Developed, by data cleansing, as to risk assessment data,
Three pre- risk assessment are carried out using XGBoost, LightGBM disaggregated model after training.
2. a kind of pre- methods of risk assessment of prison convict three based on user's portrait as described in claim 1, it is characterised in that institute
It states in step 4, training pattern XGBoost, LightGBM disaggregated model specifically comprises the following steps:
Step 41, sequentially in time using before sample data 80% data as training set, by each crime in training set
Violate input of user's portrait label as XGBoost, LightGBM disaggregated model, by the described three pre- risk assessment of each criminal
Label is as model label, training XGBoost, LightGBM disaggregated model parameter；
Step 42 selects after sample data 20% data as test set, after training sequentially in time
XGBoost, LightGBM disaggregated model carry out three pre- risk assessment to the sample data in test set, obtain three pre- risk assessment
As a result；
Step 43 is commented according to the three pre- risks that the described three pre- risk assessment labels and step 42 of criminal each in test set export
Estimate as a result, calculating the classification accuracy and recall rate of XGBoost, LightGBM disaggregated model；Judge classification accuracy and recalls
The size relation of rate and setting value, if classification accuracy and recall rate are to terminate process more than or equal to setting value；If
Any one of classification accuracy and recall rate value are less than setting value, then follow the steps 44；
Step 44 carries out upper and lower sampling by the training data to XGBoost, LightGBM disaggregated model with right
XGBoost, LightGBM disaggregated model parameter are adjusted, and use three pre- risks of XGBoost, LightGBM disaggregated model
The result of assessment is merged, and step 41 to 43 is repeated, until point for XGBoost, LightGBM disaggregated model that training obtains
Class accuracy rate and recall rate are all larger than equal to setting value.
3. a kind of pre- methods of risk assessment of prison convict three based on user's portrait as described in claim 1, it is characterised in that: institute
The calculation formula for stating classification accuracy and recall rate is as follows:
P=TP/ (TP+FP)
R=TP/ (TP+FN)
Wherein, P presentation class accuracy rate, R indicate recall rate, and TP is that positive class determines to be positive the quantity of class in test set, and FP is to survey
Examination concentrates negative class to determine to be positive the quantity of class, and FN is that positive class determines to be negative the quantity of class in test set.
4. a kind of pre- methods of risk assessment of prison convict three based on user's portrait as described in claim 1, it is characterised in that: institute
Stating the data cleansing in step 1 includes missing values cleaning, format content cleaning, logic error cleaning.
5. a kind of pre- methods of risk assessment of prison convict three based on user's portrait as described in claim 1, it is characterised in that: institute
Stating sample data in step 1 includes sample data, including convict's basic information, shopping information, mail record, education are talked and remembered
Record, emotional affection telephony recording, emotional affection meet with record, the convict that is put in prison assessment, rewards and punishments score, people's police's report information and criminal and report letter
Breath.
6. a kind of pre- risk evaluating system of prison convict three based on user's portrait, which is characterized in that the system comprises:
Sample data obtains module, clear for carrying out data to the various initial data for obtaining criminal from the management system in prison
It washes, obtains sample data；
User's portrait label establishes module, for establishing criminal user's portrait label according to the sample data；
Three pre- risk assessment labels form module, are drawn a portrait and are marked according to the sample data and the user for prison officer
Label screen the data for carrying out criminal's risk assessment, form the three pre- risk assessment labels of criminal；
Disaggregated model training module, for according to described three pre- risk assessment labels and user portrait label training
XGBoost, LightGBM disaggregated model；
Criminal's risk evaluation module, for data to be assessed carry out three pre- risk assessment, using after training XGBoost,
LightGBM disaggregated model carries out three pre- risk assessment.
7. a kind of pre- risk evaluating system of prison convict three based on user's portrait as claimed in claim 6, which is characterized in that
The disaggregated model training module includes:
Disaggregated model parameter training module will be instructed for selecting before sample data 80% sequentially in time as training set
Practice the input that each criminal user concentrated draws a portrait as XGBoost, LightGBM disaggregated model, by described the three of each criminal
Pre- risk assessment label is as model label, training XGBoost, LightGBM disaggregated model parameter；
Computing module, for selecting rear the 20% of sample data sequentially in time as test set, after training
XGBoost, LightGBM disaggregated model carry out three pre- risk assessment to the sample data in test set, obtain three pre- risk assessment
As a result；
Determining module, three for being exported according to the described three pre- risk assessment labels and computing module of criminal each in test set
Pre- risk evaluation result determines whether the classification accuracy of XGBoost, LightGBM disaggregated model and recall rate reach setting value；
Adjust module, for by training data to XGBoost, LightGBM disaggregated model carry out upper and lower sampling with
XGBoost, LightGBM disaggregated model parameter are adjusted, and use three pre- wind of XGBoost, LightGBM disaggregated model
The result nearly assessed is merged, until training the classification accuracy of obtained XGBoost, LightGBM disaggregated model and recalling
Rate reaches setting value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811363792.3A CN109543986A (en) | 2018-11-16 | 2018-11-16 | The pre- methods of risk assessment of prison convict three and system based on user's portrait |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811363792.3A CN109543986A (en) | 2018-11-16 | 2018-11-16 | The pre- methods of risk assessment of prison convict three and system based on user's portrait |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109543986A true CN109543986A (en) | 2019-03-29 |
Family
ID=65847858
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811363792.3A Pending CN109543986A (en) | 2018-11-16 | 2018-11-16 | The pre- methods of risk assessment of prison convict three and system based on user's portrait |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109543986A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110134722A (en) * | 2019-05-22 | 2019-08-16 | 北京小度信息科技有限公司 | Target user determines method, apparatus, equipment and storage medium |
| CN110826852A (en) * | 2019-09-27 | 2020-02-21 | 安徽四创电子股份有限公司 | Risk assessment method and system for forced isolation drug rehabilitation personnel |
| CN111950585A (en) * | 2020-06-29 | 2020-11-17 | 广东技术师范大学 | XGboost-based underground comprehensive pipe gallery safety condition assessment method |
| CN112037010A (en) * | 2020-08-12 | 2020-12-04 | 无锡锡商银行股份有限公司 | Application method and device of multi-scene risk rating model based on SSR-Net in personal loan and storage medium |
| CN112036483A (en) * | 2020-08-31 | 2020-12-04 | 中国平安人寿保险股份有限公司 | Object prediction classification method and device based on AutoML, computer equipment and storage medium |
| CN112508311A (en) * | 2021-02-05 | 2021-03-16 | 浙江连信科技有限公司 | Data processing method and device for risk prediction |
| CN113077197A (en) * | 2021-06-08 | 2021-07-06 | 泰豪信息技术有限公司 | Escort personnel consumption supervision method, system, storage medium and equipment |
| CN115329909A (en) * | 2022-10-17 | 2022-11-11 | 上海冰鉴信息科技有限公司 | User portrait generation method and device and computer equipment |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103761449A (en) * | 2014-02-20 | 2014-04-30 | 杭州华亭科技有限公司 | Criminal propensity and risk degree quantifying method and system based on AHP |
| CN104182805A (en) * | 2014-08-22 | 2014-12-03 | 杭州华亭科技有限公司 | Dangerous tendency prediction method based on prisoner behavior characteristic ensemble learning model |
| CN105139029A (en) * | 2015-08-14 | 2015-12-09 | 哈尔滨华夏矿安科技有限公司 | Activity recognition method and activity recognition device for persons serving sentences |
| CN106096623A (en) * | 2016-05-25 | 2016-11-09 | 中山大学 | A kind of crime identifies and Forecasting Methodology |
| CN106304015A (en) * | 2015-05-28 | 2017-01-04 | 中兴通讯股份有限公司 | The determination method and device of subscriber equipment |
| CN106447194A (en) * | 2016-09-27 | 2017-02-22 | 上海中信信息发展股份有限公司 | Assessment method for criminal risk factor by combining static and dynamic information |
| CN107895245A (en) * | 2017-12-26 | 2018-04-10 | 国网宁夏电力有限公司银川供电公司 | A kind of tariff recovery methods of risk assessment based on user's portrait |
| CN108363714A (en) * | 2017-12-21 | 2018-08-03 | 北京至信普林科技有限公司 | A kind of method and system for the ensemble machine learning for facilitating data analyst to use |
| CN108399190A (en) * | 2018-01-24 | 2018-08-14 | 山东中磁视讯股份有限公司 | A kind of panorama image space method applied to prison convict |
| CN108492173A (en) * | 2018-03-23 | 2018-09-04 | 上海氪信信息技术有限公司 | A kind of anti-Fraud Prediction method of credit card based on dual-mode network figure mining algorithm |
-
2018
- 2018-11-16 CN CN201811363792.3A patent/CN109543986A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103761449A (en) * | 2014-02-20 | 2014-04-30 | 杭州华亭科技有限公司 | Criminal propensity and risk degree quantifying method and system based on AHP |
| CN104182805A (en) * | 2014-08-22 | 2014-12-03 | 杭州华亭科技有限公司 | Dangerous tendency prediction method based on prisoner behavior characteristic ensemble learning model |
| CN106304015A (en) * | 2015-05-28 | 2017-01-04 | 中兴通讯股份有限公司 | The determination method and device of subscriber equipment |
| CN105139029A (en) * | 2015-08-14 | 2015-12-09 | 哈尔滨华夏矿安科技有限公司 | Activity recognition method and activity recognition device for persons serving sentences |
| CN106096623A (en) * | 2016-05-25 | 2016-11-09 | 中山大学 | A kind of crime identifies and Forecasting Methodology |
| CN106447194A (en) * | 2016-09-27 | 2017-02-22 | 上海中信信息发展股份有限公司 | Assessment method for criminal risk factor by combining static and dynamic information |
| CN108363714A (en) * | 2017-12-21 | 2018-08-03 | 北京至信普林科技有限公司 | A kind of method and system for the ensemble machine learning for facilitating data analyst to use |
| CN107895245A (en) * | 2017-12-26 | 2018-04-10 | 国网宁夏电力有限公司银川供电公司 | A kind of tariff recovery methods of risk assessment based on user's portrait |
| CN108399190A (en) * | 2018-01-24 | 2018-08-14 | 山东中磁视讯股份有限公司 | A kind of panorama image space method applied to prison convict |
| CN108492173A (en) * | 2018-03-23 | 2018-09-04 | 上海氪信信息技术有限公司 | A kind of anti-Fraud Prediction method of credit card based on dual-mode network figure mining algorithm |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110134722A (en) * | 2019-05-22 | 2019-08-16 | 北京小度信息科技有限公司 | Target user determines method, apparatus, equipment and storage medium |
| CN110826852A (en) * | 2019-09-27 | 2020-02-21 | 安徽四创电子股份有限公司 | Risk assessment method and system for forced isolation drug rehabilitation personnel |
| CN111950585A (en) * | 2020-06-29 | 2020-11-17 | 广东技术师范大学 | XGboost-based underground comprehensive pipe gallery safety condition assessment method |
| CN112037010A (en) * | 2020-08-12 | 2020-12-04 | 无锡锡商银行股份有限公司 | Application method and device of multi-scene risk rating model based on SSR-Net in personal loan and storage medium |
| CN112036483A (en) * | 2020-08-31 | 2020-12-04 | 中国平安人寿保险股份有限公司 | Object prediction classification method and device based on AutoML, computer equipment and storage medium |
| CN112036483B (en) * | 2020-08-31 | 2024-03-15 | 中国平安人寿保险股份有限公司 | AutoML-based object prediction classification method, device, computer equipment and storage medium |
| CN112508311A (en) * | 2021-02-05 | 2021-03-16 | 浙江连信科技有限公司 | Data processing method and device for risk prediction |
| CN113077197A (en) * | 2021-06-08 | 2021-07-06 | 泰豪信息技术有限公司 | Escort personnel consumption supervision method, system, storage medium and equipment |
| CN115329909A (en) * | 2022-10-17 | 2022-11-11 | 上海冰鉴信息科技有限公司 | User portrait generation method and device and computer equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109543986A (en) | The pre- methods of risk assessment of prison convict three and system based on user's portrait | |
| CN105184315B (en) | A kind of quality inspection processing method and system | |
| CN106611169B (en) | A kind of dangerous driving behavior real-time detection method based on deep learning | |
| CN110619568A (en) | Risk assessment report generation method, device, equipment and storage medium | |
| US20170004128A1 (en) | Device and method for analyzing reputation for objects by data mining | |
| CN109492026A (en) | A kind of Telecoms Fraud classification and Detection method based on improved active learning techniques | |
| CN105225420B (en) | Air traffic controller's fatigue detection method based on principal component analysis, device and system | |
| CN106295565A (en) | Monitor event identifications based on big data and in real time method of crime prediction | |
| CN107766797A (en) | A kind of stranger's recognition methods and system | |
| CN109543542A (en) | A kind of determination method whether particular place personnel dressing standardizes | |
| CN110197332A (en) | A kind of overall control of social public security evaluation method | |
| CN209543514U (en) | Monitoring and alarm system based on recognition of face | |
| CN110009224A (en) | Suspect's violation probability prediction technique, device, computer equipment and storage medium | |
| CN111783086A (en) | Internal threat detection method and system based on anti-production behavior characteristics | |
| CN110245603A (en) | A kind of group abnormality behavior real-time detection method | |
| CN106952190A (en) | False source of houses typing Activity recognition and early warning system | |
| CN113411542A (en) | Intelligent working condition monitoring equipment | |
| CN113408464A (en) | Behavior detection method and device, electronic equipment and storage medium | |
| CN106845916A (en) | A kind of intelligent APP attendance management and method for monitoring state based on water purifier | |
| CN114692593A (en) | Network information safety monitoring and early warning method | |
| CN103970727B (en) | Anti- cheat method, device and server based on topic | |
| Adderley | The use of data mining techniques in operational crime fighting | |
| CN112528952B (en) | Working state intelligent recognition system for electric power business hall personnel | |
| CN104301330B (en) | The Honey net detection method measured with member's cohesion is monitored based on abnormal behaviour | |
| CN114186118A (en) | Network public opinion topic information processing system, method, storage medium and terminal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190329 |
|
| RJ01 | Rejection of invention patent application after publication |