CN106022387A - Method and system for testing damage grade model - Google Patents
Method and system for testing damage grade model Download PDFInfo
- Publication number
- CN106022387A CN106022387A CN201610365761.6A CN201610365761A CN106022387A CN 106022387 A CN106022387 A CN 106022387A CN 201610365761 A CN201610365761 A CN 201610365761A CN 106022387 A CN106022387 A CN 106022387A
- Authority
- CN
- China
- Prior art keywords
- data
- test
- impairment scale
- scale model
- training
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
- G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/285—Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system
Abstract
Provided is a method and system for testing a damage grade model. The method includes selecting characteristic data, dividing the characteristic data into training characteristic data and testing characteristic data based on a certain proportion, and saving the data, selecting a classifier, performing model training on the damage grade data of each component, and testing the damage grade data. The testing result display part is divided into all-component display and single-component display, and the testing result is checked based on the testing demands, and meanwhile, in the single-component display, the performance of the classifier is analyzed based on the evaluate indexes.
Description
Technical field
The invention belongs to data processing field, specifically damage based on long-range setting loss technical testing
The method and system of Grade Model.
Background technology
The recoverable amount of automobile is gradually increasing every year at present, and the constantly planning of road traffic makes car
Travel speed promoted, the incidence rate of vehicle accident also increase, automobile touches again
Claims Resolution flow process main after hitting is: is in danger and--reports a case to the security authorities and--survey--setting loss--verify prices--core damage--
Core is paid for--and paying, wherein setting loss is that the professional sent according to insurance company surveys to scene
After examining, position vestige and degree according to loss carry out on-the-spot preliminary setting loss, or directly arrive
Setting loss is gone at repair shop, 4S shop, setting loss center.This not only consumes substantial amounts of manpower and materials,
And require higher to the specialty of setting loss person during setting loss, so utilize more to become at present
Ripe machine learning method carries out long-range setting loss to car crass, not only can solve manpower and materials
The wasting of resources, and can more rapid more fully to vehicle part damage judge, institute
Have great importance with setting loss long-range to vehicle low speed collision.
During setting loss long-range to slow moving vehicle, mainly vehicle Portable device is adopted
The acceleration of collection, the signal such as angular velocity (hereinafter referred to as vehicle running signal) carry out Treatment Analysis,
Judgment of learning, but the vehicle running signal of actual acquisition is discrete random signal, this signal
It is the signal of a kind of uncertainty, there is feature and the basis of unpredictable future temporal exact value
There is a lot of noises in body, how to be extracted and will to become later stage by the useful information in signal
Practise the important prerequisite judged.
Summary of the invention
The present invention proposes a kind of method and system testing impairment scale model, is for damaging
Hinder the polytropy of factors during grade judgment models determines and design.
On the one hand, the method that the invention provides test impairment scale model, including:
S1: selected characteristic data；
S2: characteristic is divided into according to a certain percentage training characteristics data and test feature number
According to, preserve simultaneously；
S3: choose grader；
S4: the impairment scale data of each part are carried out model training；
S5: impairment scale model is tested.
Concrete, step S1 selected characteristic data, including selecting the characteristic number that handled well
According to call the pretreated characteristic of data processing module.
Concrete, in step S2, characteristic is divided into training data and survey according to the ratio of 7:3
Examination data.
Concrete, when data characteristics dimension is bigger, step S3 also includes entering characteristic
The step of row dimension-reduction treatment.
Concrete, use PCA to carry out dimension-reduction treatment to levying data, its contribution rate has two kinds of acquisitions
Mode, one is acquiescence contribution rate；Another kind is input contribution rate, and input empirical value is 0.8
Between 0.95.
Concrete, grader in step S3, including SVM and RBF.
More specifically, step S4 training process is impairment scale data to each part respectively
Being trained, whether training preserves the training pattern of current part and uses PCA to convert after terminating
Mark.
More specifically, the acquisition mode of impairment scale model measurement data is: to separating, training
During the data transmitted directly carry out testing or the data preserved after separating being read
Take test, after test, test result is shown.
More specifically, test result shows and is divided into single part show and all parts show, singly zero
Part shows and includes the classification situation of all samples, this part under the accuracy rate of this part, this part
The hybrid matrix of all impairment scales and the evaluation index of this part injury grade；All parts show
Show accuracy rate and the Average Accuracy of all parts including each part.
On the other hand, present invention also offers a kind of test impairment scale model system, bag
Include:
Data decimation module for selected characteristic data；
Characteristic is divided into according to a certain percentage training characteristics data and test feature data
Data separating module；
The grader that different impairment scales are correctly classified；
Impairment scale data by each part carry out the training module of impairment scale model training；
Impairment scale model is tested and is shown the test module of test result.
Due to the fact that the above technical method of employing, it is possible to obtain following technique effect:
1. the training data read is the feature extracted, it is to avoid carry out spy for same batch of data
That levies repeats extraction.
2. can verify that training data exists with test ratio data by the difference of segregation ratio
Impact on grader under different situations.
3. show part in test result, be divided into all parts to show and single part shows, permissible
According to testing requirement, test result is checked, simultaneously can be by commenting in single part shows
Valency index carries out the performance evaluation of grader.
Accompanying drawing explanation
For clearer explanation embodiments of the invention or the technical scheme of prior art, below
Introduce the accompanying drawing used required in embodiment or description of the prior art is done one simply, aobvious
And easy insight, the accompanying drawing in describing below is only some embodiments of the present invention, for ability
From the point of view of the those of ordinary skill of territory, on the premise of not paying creative work, it is also possible to according to this
A little accompanying drawings obtain other accompanying drawing.
Fig. 1 is the method flow diagram of test impairment scale model.
Detailed description of the invention
For making the purpose of embodiments of the invention, technical scheme and advantage clearer, knot below
Close the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out the completeest
Whole description:
Embodiment 1
The method that the invention provides test impairment scale model, including:
S1: selected characteristic data；
Characteristic choose two ways: one is the characteristic selecting to have handled well；
Two is to call the pretreated characteristic of data processing module；
S2: data are separated；
Characteristic is divided into training characteristics data and test feature data according to the ratio of 7:3,
Preserve simultaneously；
S3: choose grader；
Different impairment scales are correctly classified；Described grader includes SVM and RBF；
S4: model training and preservation
After above-mentioned parameter is provided with, the impairment scale data of each part are carried out model instruction
Practice；Training process is to be trained the impairment scale data of each part respectively.
S5: impairment scale model is tested；
The acquisition mode of impairment scale model measurement data by: to separate, training during passed
The data passed directly carry out testing or the data preserved after separating being read out test, test
Afterwards test result is shown；Test result shows and is divided into single part show and all parts are aobvious
Show, single part show the classification situation of all samples under the accuracy rate of this part, this part that includes,
The hybrid matrix of all impairment scales of this part and the evaluation index of this part injury grade；All
Part shows accuracy rate and the Average Accuracy of all parts including each part.
On the other hand, present invention also offers a kind of test impairment scale model system, bag
Include:
Data decimation module for selected characteristic data；
Characteristic is divided into according to a certain percentage training characteristics data and test feature data
Data separating module；
The grader that different impairment scales are correctly classified；
Impairment scale data by each part carry out the training module of impairment scale model training；
Impairment scale model is tested and is shown the test module of test result.
Embodiment 2
The method that the invention provides test impairment scale model, including:
S1: selected characteristic data；
Characteristic choose two ways: one is the characteristic selecting to have handled well；
Two is to call the characteristic after data processing module resolves；
S2: data are separated；
Characteristic is divided into training data and test data according to the ratio of 7:3, carries out simultaneously
Preserve；
S3: to characteristic dimension-reduction treatment
When data characteristics dimension is bigger or when intrinsic dimensionality exceedes sample size, to feature
Data carry out dimension-reduction treatment；Using PCA to carry out dimension-reduction treatment to levying data, its contribution rate has two
Planting acquisition mode, one is acquiescence contribution rate, is set as 0.85；Another kind is input contribution
Rate, input value is between 0.8 to 0.95.
S4: choose grader；
Different impairment scales are correctly classified；Described grader includes SVM and RBF；
S5: model training and preservation
After above-mentioned parameter is provided with, the impairment scale data of each part are carried out model instruction
Practice；Training process is to be trained the impairment scale data of each part respectively, and training terminates
The training pattern of the current part of rear preservation and whether use the mark that PCA converts.
S6: impairment scale model is tested；
The acquisition mode of impairment scale model measurement data by: to separate, training during passed
The data passed directly carry out testing or the data preserved after separating being read out test, test
Afterwards test result is shown；Test result shows and is divided into single part show and all parts are aobvious
Show, single part show the classification situation of all samples under the accuracy rate of this part, this part that includes,
The hybrid matrix of all impairment scales of this part and the evaluation index of this part injury grade；All
Part shows accuracy rate and the Average Accuracy of all parts including each part.
On the other hand, present invention also offers a kind of test impairment scale model system, bag
Include:
Data decimation module for selected characteristic data；
Characteristic is divided into according to a certain percentage training characteristics data and test feature data
Data separating module；；
When data characteristics dimension is bigger or when intrinsic dimensionality exceedes sample size, to feature
Data carry out the dimension-reduction treatment module of dimension-reduction treatment；
The grader that different impairment scales are correctly classified；
Impairment scale data by each part carry out the training module of impairment scale model training；
Impairment scale model is tested and is shown the test module of test result.
The above, the only present invention preferably detailed description of the invention, but the protection model of the present invention
Enclosing and be not limited thereto, any those familiar with the art is in the skill of present disclosure
In the range of art, according to technical scheme and inventive concept equivalent in addition thereof or change
Become, all should contain within protection scope of the present invention.
Claims (10)
1. the method testing impairment scale model, it is characterised in that including:
S1: selected characteristic data；
S2: characteristic is divided into according to a certain percentage training characteristics data and test feature number
According to, preserve simultaneously；
S3: choose grader；
S4: the impairment scale data of each part are carried out the training of model；
S5: impairment scale model is tested.
The method of test impairment scale model the most according to claim 1, it is characterised in that
Step S1 selected characteristic data, including selecting the characteristic handled well and calling data
The pretreated characteristic of processing module.
The method of test impairment scale model the most according to claim 1, it is characterised in that
In step S2, characteristic is divided into training characteristics data and test feature number according to the ratio of 7:3
According to.
The method of test impairment scale model the most according to claim 1, it is characterised in that
When data characteristics dimension is bigger, step S3 also includes characteristic is carried out dimension-reduction treatment
Step.
The method of test impairment scale model the most according to claim 4, it is characterised in that
Using PCA to carry out dimension-reduction treatment to levying data, its contribution rate has two kinds of acquisition modes, Yi Zhongshi
Acquiescence contribution rate；Another kind is input contribution rate, and input empirical value is between 0.8 to 0.95.
The method of test impairment scale model the most according to claim 1, it is characterised in that
Grader in step S3, including SVM and RBF.
The method of test impairment scale model the most according to claim 1, it is characterised in that
Step S4 training process is to be trained the impairment scale data of each part respectively, training
The training pattern preserving current part after end and the mark whether using PCA to convert.
The method of test impairment scale model the most according to claim 1, it is characterised in that
The acquisition mode of impairment scale model measurement data by: transmitted to separating, during training
Data directly carry out testing or the data preserved after separating being read out test, right after test
Test result shows.
The method of test impairment scale model the most according to claim 1, it is characterised in that
Test result shows and is divided into single part show and all parts show, list part show include this zero
The classification situation of all samples under the accuracy rate of part, this part, all impairment scales of this part
Hybrid matrix and the evaluation index of this part injury grade；All parts show and include each part
Accuracy rate and the Average Accuracy of all parts.
10. one kind test impairment scale model system, it is characterised in that including:
Data decimation module for selected characteristic data；
Characteristic is divided into according to a certain percentage training characteristics data and test feature data
Data separating module；
The grader that different impairment scales are correctly classified；
Impairment scale data by each part carry out the training module of impairment scale model training；
Impairment scale model is tested and is shown the test module of test result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610365761.6A CN106022387A (en) | 2016-05-27 | 2016-05-27 | Method and system for testing damage grade model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610365761.6A CN106022387A (en) | 2016-05-27 | 2016-05-27 | Method and system for testing damage grade model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106022387A true CN106022387A (en) | 2016-10-12 |
Family
ID=57091379
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610365761.6A Pending CN106022387A (en) | 2016-05-27 | 2016-05-27 | Method and system for testing damage grade model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106022387A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101685006A (en) * | 2009-06-25 | 2010-03-31 | 上海交通大学 | Automatic detection system for vision of contact part feature machine |
| CN102799669A (en) * | 2012-07-17 | 2012-11-28 | 杭州淘淘搜科技有限公司 | Automatic grading method for commodity image vision quality |
| CN104636493A (en) * | 2015-03-04 | 2015-05-20 | 浪潮电子信息产业股份有限公司 | Method for classifying dynamic data on basis of multi-classifier fusion |
| CN105488539A (en) * | 2015-12-16 | 2016-04-13 | 百度在线网络技术（北京）有限公司 | Generation method and device of classification method, and estimation method and device of system capacity |
| CN105488789A (en) * | 2015-11-24 | 2016-04-13 | 大连楼兰科技股份有限公司 | Grading damage assessment method for automobile part |
-
2016
- 2016-05-27 CN CN201610365761.6A patent/CN106022387A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101685006A (en) * | 2009-06-25 | 2010-03-31 | 上海交通大学 | Automatic detection system for vision of contact part feature machine |
| CN102799669A (en) * | 2012-07-17 | 2012-11-28 | 杭州淘淘搜科技有限公司 | Automatic grading method for commodity image vision quality |
| CN104636493A (en) * | 2015-03-04 | 2015-05-20 | 浪潮电子信息产业股份有限公司 | Method for classifying dynamic data on basis of multi-classifier fusion |
| CN105488789A (en) * | 2015-11-24 | 2016-04-13 | 大连楼兰科技股份有限公司 | Grading damage assessment method for automobile part |
| CN105488539A (en) * | 2015-12-16 | 2016-04-13 | 百度在线网络技术（北京）有限公司 | Generation method and device of classification method, and estimation method and device of system capacity |
Non-Patent Citations (1)
| Title |
|---|
| 方匡南: "《随机森林组合预测理论及其在金融中的应用》", 31 May 2012 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104802737B (en) | Mobile phone based vehicle abnormality driving behavior detection method | |
| CN107358250B (en) | Body gait recognition methods and system based on the fusion of two waveband radar micro-doppler | |
| CN105320966A (en) | Vehicle driving state recognition method and apparatus | |
| CN105976449A (en) | Remote automatic damage assessment and collision detection method and system for vehicle | |
| US9128116B2 (en) | Automatic alignment of a vehicle three-axes accelerometer | |
| CN108407816B (en) | Method and system for evaluating driving of automobile driver | |
| CN109916488B (en) | Dynamic vehicle weighing method and device | |
| CN110986938B (en) | Bumpy road identification method and device and electronic equipment | |
| RU2011134065A (en) | IMPROVEMENTS CONCERNING THE NAVIGATION DEVICE USED IN THE VEHICLE | |
| CN106650157B (en) | Method, device and system for estimating fault occurrence probability of vehicle parts | |
| CN106940931B (en) | The tollgate devices quality of data method of inspection based on location data | |
| CN105389985B (en) | A kind of intelligent driving behavior analysis method based on mobile phone sensor | |
| CN106066907B (en) | Loss assessment grading method based on multi-part multi-model judgment | |
| CN109649396B (en) | Safety detection method for commercial vehicle driver | |
| CN105373647B (en) | A method of passing through the pneumatic focus of ground roll-out test identification | |
| Sun et al. | Combining machine learning and dynamic time wrapping for vehicle driving event detection using smartphones | |
| CN107192560A (en) | A kind of vehicle accelerates power method for objectively evaluating | |
| CN106203437B (en) | Individual driving behavior recognition methods and device | |
| CN108401464A (en) | A kind of mobile unit and vehicle collision analysis method and device | |
| CN104700630A (en) | Method and system for monitoring vehicle flow of highway | |
| CN109415057A (en) | Method for preferably identifying object by driver assistance system | |
| US9846174B2 (en) | Computer-implemented methods and computer systems/machines for identifying dependent and vehicle independent states | |
| CN107662613A (en) | A kind of extreme driving behavior recognition methods and system based on mobile intelligent perception | |
| CN106022387A (en) | Method and system for testing damage grade model | |
| CN108981728A (en) | A kind of intelligent vehicle navigation map method for building up |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20161012 |