CN110428016A - Feature vector generation method and system, user's identification model generation method and system - Google Patents
Feature vector generation method and system, user's identification model generation method and system Download PDFInfo
- Publication number
- CN110428016A CN110428016A CN201910730490.3A CN201910730490A CN110428016A CN 110428016 A CN110428016 A CN 110428016A CN 201910730490 A CN201910730490 A CN 201910730490A CN 110428016 A CN110428016 A CN 110428016A
- Authority
- CN
- China
- Prior art keywords
- user
- vector
- identification model
- data
- feature
- 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
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- 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
Abstract
Present disclose provides a feature vectors generation methods and system, user's identification model generation method and system, comprising: is pre-processed to initial data according to default characteristic format, to obtain the corresponding initial characteristics vector of initial data；It is character string type feature by the numeric type Feature Conversion in initial characteristics vector；The initial characteristics vector for completing data type conversion is converted into the corresponding discrete matrix comprising 1 and 0 two kind of discrete value according to preset algorithm；The corresponding final feature vector of initial data is generated according at least to discrete matrix.The technical solution of the disclosure can realize that by the numeric type Feature Conversion in feature vector be text-type feature, and the feature vector ultimately generated can be applied in corresponding model, simplified model step and calculation amount, lift scheme operational efficiency.
Description
Technical field
This disclosure relates to field of computer technology, in particular to a feature vectors generation method and system, user's identification
Model generating method and system.
Background technique
At this stage based on the identification model of data characteristics, numeric type data in initial data feature is relied heavily on
Characteristic such as compares size of data, and conversion is normalized.Such design relates in one aspect to data type multiplicity, data stabilization
Property cannot measure, be on the other hand related to a variety of logic calculations, to a certain extent influence correlation model operational efficiency.
Summary of the invention
The disclosure aims to solve at least one of the technical problems existing in the prior art, proposes feature vectors generation
Method and system, user's identification model generation method and system.
To achieve the above object, the embodiment of the present disclosure provides a feature vectors generation method, comprising:
Initial data is pre-processed according to default characteristic format, to obtain the corresponding initial characteristics of the initial data
Vector；
It is character string type feature by the numeric type Feature Conversion in the initial characteristics vector；
The initial characteristics vector for completing data type conversion is converted to corresponding according to preset algorithm and includes 1 He
The discrete matrix of 0 two kinds of discrete values；
The corresponding final feature vector of the initial data is generated according at least to the discrete matrix.
It in some embodiments, is that character string type is special in the numeric type Feature Conversion by the initial characteristics vector
Before the step of sign, further includes:
At least one derivative data corresponding with the initial characteristics vector is generated according to the initial characteristics vector；
The step of final feature vector corresponding according at least to the discrete matrix generation initial data, is specific
Include:
The discrete matrix is carried out field at least one described derivative data to merge, to generate the initial data pair
The final feature vector answered.
To achieve the above object, the embodiment of the present disclosure provides a kind of user's identification model generation method, comprising:
Obtain the initial data of multiple broadband users and non-broadband user；
Using described eigenvector generation method any in above-described embodiment, final spy corresponding to each initial data is generated
Levy vector；
, at least partly as training sample, to train corresponding user in all final feature vectors and identify mould
Type, user's identification model can go out this feature vector according to the eigenvector recognition inputted corresponding to user be broadband
User or non-broadband user.
In some embodiments, described, at least partly as training sample, to be instructed in all final feature vectors
The step of practising corresponding user's identification model, specifically includes:
Using the part in all final feature vectors as training sample, corresponding user's identification model is trained；
After described the step of training corresponding user's identification model, further includes:
Using another part in all final feature vectors as test sample, the user trained is identified
Model optimizes processing.
In some embodiments, when using described in previous embodiment, including it is described according to the initial characteristics vector
The feature vector generation method for the step of generating at least one derivative data corresponding with the initial characteristics vector generates each
When final feature vector corresponding to initial data, the initial data includes: user identifier, broadband Contract ID, Yong Huhang
It is characterized data and multiple Dan Yue enters an item of expenditure in the accounts the amount of money；
At least one described derivative data includes: expense summation, expense rate of change, using in saturation degree and stability
At least one.
To achieve the above object, the embodiment of the present disclosure provides a feature vectors and generates system, comprising:
Preprocessing module, for being pre-processed to initial data according to default characteristic format, to obtain the original number
According to corresponding initial characteristics vector；
First conversion module, for being character string type feature by the numeric type Feature Conversion in the initial characteristics vector；
Second conversion module, for converting the initial characteristics vector for completing data type conversion according to preset algorithm
For the corresponding discrete matrix comprising 1 and 0 two kind of discrete value；
Generation module, for generating the corresponding final feature vector of the initial data according at least to the discrete matrix.
In some embodiments, first conversion module includes:
First generation unit, it is corresponding extremely with the initial characteristics vector for being generated according to the initial characteristics vector
A few derivative data；
The generation module includes:
Second generation unit is merged for the discrete matrix to be carried out field at least one described derivative data, with
Generate the corresponding final feature vector of the initial data.
To achieve the above object, the embodiment of the present disclosure provides a kind of user's identification model generation system, comprising:
Module is obtained, for obtaining the initial data of multiple broadband users and non-broadband user；
Any described eigenvector generates system in above-described embodiment, for generating final spy corresponding to each initial data
Levy vector；
Training module, for, at least partly as training sample, to be trained in all final feature vectors pair
The user's identification model answered, it is right that user's identification model can go out this feature vector institute according to the eigenvector recognition inputted
The user answered is broadband user or non-broadband user.
In some embodiments, the training module includes:
Training unit, for training corresponding using the part in all final feature vectors as training sample
User's identification model；
User's identification model generates system further include:
Optimization module, for using another part in all final feature vectors as test sample, to training
User's identification model optimize processing.
In some embodiments, it is described in previous embodiment when described eigenvector generates system, including described the
When the feature vector of one generation unit and second generation unit generates system, the initial data includes: user identifier, width
Band Contract ID, consumption data and user behavior characteristics data；
At least one described derivative data includes: expense summation, expense rate of change, using in saturation degree and stability
At least one.
The disclosure has the advantages that
The embodiment of the present disclosure provides a feature vectors generation method and system, user's identification model generation method and is
System is, it can be achieved that by converting the feature with text property for numeric type feature, when feature vector is applied to corresponding model,
Simplified model step reduces relevant calculation amount, lift scheme efficiency.
Detailed description of the invention
Fig. 1 is the flow chart for the feature vectors generation method that the embodiment of the present disclosure provides；
Fig. 2 is the flow chart for another feature vectors generation method that the embodiment of the present disclosure provides；
Fig. 3 is a kind of flow chart for user's identification model generation method that the embodiment of the present disclosure provides；
Fig. 4 is the flow chart for another user's identification model generation method that the embodiment of the present disclosure provides；
Fig. 5 is the structural block diagram that the feature vectors that the embodiment of the present disclosure provides generate system；
Fig. 6 is the structural block diagram that a kind of user's identification model that the embodiment of the present disclosure provides generates system.
Specific embodiment
To make those skilled in the art more fully understand the technical solution of the disclosure, the disclosure is mentioned with reference to the accompanying drawing
The feature vector generation method and system of confession, user's identification model generation method and system are described in detail.
Feature vector generation method and system provided by the disclosure, user's identification model generation method and system can be used for
Text-type feature is converted by numeric type feature, and feature vector is applied to corresponding model.
Fig. 1 is the flow chart for the feature vectors generation method that the embodiment of the present disclosure provides.As shown in Figure 1, this feature
Vector generation method includes:
Step S1, initial data is pre-processed according to default characteristic format.
In step sl, initial data is pre-processed according to default characteristic format, it is corresponding to obtain initial data
Initial characteristics vector.
In practical applications, default characteristic format includes integer data format and string format, respectively corresponds numeric type feature
With character string type feature.When being pre-processed, integer type is met for data type and numerical value is relatively fixed or lesser feature,
Original numerical value, such as number of paying dues (pay_times can be denoted as) can be directly stored as；Integer type is met for data type,
But numerical value is not fixed or biggish feature, can preset threshold value, according to this feature numerical value whether be greater than preset threshold use compared with
Small natural number is periodically stored, such as set meal expense (can be denoted as survice_fee), the amount of money of entering an item of expenditure in the accounts (can be denoted as
Total_fee) and net duration (online_time can be denoted as).
It is similar, character string type is met for data type and the feature of the relatively fixed or corresponding corresponding encoded of content, it can be straight
It connects and is stored as original value, such as contract type (can be denoted as contract_type) and broadband services (can be denoted as
broadband)；Character string type is met for data type and is directed toward the feature of multiple denumerable types, 1 He of character combination can be used
0 or 1 and 2 are stored, for example, gender (gender can be denoted as), whether be promise to undertake low consumption user (low_ can be denoted as
Consume whether surpass) and continuously set (over_bill can be denoted as).
It step S2, is character string type feature by the numeric type Feature Conversion in initial characteristics vector.
Step S3, the initial characteristics vector for completing data type conversion is converted to corresponding according to preset algorithm includes
1 and 0 two kind of discrete value discrete matrix.
In practical applications, preset algorithm can be will complete data type conversion initial characteristics vector sum preset feature to
Amount carries out same or logical operation algorithm for characteristic value, wherein discrete value 1, two spies are corresponded to if two characteristic values are identical
Value indicative difference then corresponds to discrete value 0.For multiple identification classification types in identification model, which is to meet it
In a type standard feature vector, characteristic value be corresponding various features standard and reference.
Step S4, the corresponding final feature vector of initial data is generated according at least to discrete matrix.
It in practical applications, can be by every a line of discrete matrix when only generating final feature vector according to discrete matrix
As a row vector, the corresponding final feature vector of each row vector, to generate the corresponding final feature of initial data
Vector.
The embodiment of the present disclosure provides a feature vectors generation method, and this method can be used for the numerical value in feature vector
Type Feature Conversion is character string type feature, when being applied in corresponding model, so that spy of the model independent of numeric type feature
Property, simplified model step and calculation amount.
Fig. 2 is the flow chart for another feature vectors generation method that the embodiment of the present disclosure provides.As shown in Fig. 2, the spy
Levying vector generation method not only includes step S1~step S4, before step S2, further includes:
Step S2a, at least one derivative data corresponding with initial characteristics vector is generated according to initial characteristics vector.
In practical applications, which can pass through corresponding arithmetic by the characteristic in initial characteristics vector
Or logical operation obtains.
As shown in Fig. 2, step S4 is specifically included:
Step S401, discrete matrix field is carried out at least one derivative data to merge, it is corresponding to generate initial data
Final feature vector.
In practical applications, the discrete value in discrete matrix can be subjected to word according to the thought that field merges with derivative data
Duan Ronghe forms new discrete matrix, and generates final feature according to discrete matrix in practical applications according to abovementioned steps S4
The method of vector generates final feature vector.
The embodiment of the present disclosure provides a feature vectors generation method, and this method can be used for raw according to initial characteristics vector
Merge at derivative data, and by the two, to form dimension partial data feature abundant.
Fig. 3 is a kind of flow chart for user's identification model generation method that the embodiment of the present disclosure provides.As shown in figure 3, should
User's identification model generation method includes:
Step S5, the initial data of multiple broadband users and non-broadband user are obtained.
In practical applications, broadband user and non-broadband user correspond to multiple identification classification types in identification model, comment
Can it be that identify user as broadband user or non-broadband according to the probability threshold value set that whether the valence model trains successful standard
User.
Step S6, it using any feature vector generation method in above-described embodiment, generates corresponding to each initial data most
Whole feature vector.
Step S7, at least partly as training sample, to train corresponding user in all final feature vectors and know
Other model.
Wherein, user's identification model can go out this feature vector according to the eigenvector recognition inputted corresponding to user
For broadband user or non-broadband user.
In practical applications, by, at least partly as training sample, training process is as follows in all final feature vectors:
Training sample intersect averagely, particularly, is intersected averagely using five foldings, uses LightGBM (Light Gradient
Boosting Machine) algorithm frame carries out model training, prediction classification is compared with actual classification, after obtain model
Classification results, the probability value of output prediction broadband user, the successive value which is 0 to 100%, model is according in advance at this time
The screening threshold value of setting classifies to broadband user and non-broadband user.
The embodiment of the present disclosure provides a kind of user's identification model generation method, and this method can be used for using above-described embodiment
In feature vector generation method generate final feature vector, by final feature vector at least partly as training sample pair
Model is trained, and it is broadband user or non-broadband user that the model after training, which can be used to identify user,.
Fig. 4 is the flow chart for another user's identification model generation method that the embodiment of the present disclosure provides.As shown in figure 4,
Step S7 is specifically included:
Step S701, using the part in all final feature vectors as training sample, corresponding user's identification is trained
Model.
As shown in figure 4, user's identification model generation method not only includes step S5~step S701, step S701 it
It afterwards, further include step S8.
Step S8, using another part in all final feature vectors as test sample, the user trained is identified
Model optimizes processing.
In practical applications, training sample is used as using the part in all final feature vectors, with all final features to
Another part in amount is as test sample, after being trained process according to the method for abovementioned steps S7 in practical applications, In
On the basis of output model classification results, the parameter area of LightGBM algorithm frame major parameter is limited, exports training sample
The F1 score (F1Score) of classification results corresponding with test sample, select the corresponding highest one group of major parameter of F1 score with
Optimal model parameters group is constituted, to reach the purpose of optimization processing.
In some embodiments, when using the feature vector generation method in previous embodiment including step S2a, generation is each
When final feature vector corresponding to initial data, initial data includes: user identifier, broadband Contract ID, user behavior are special
Sign data and multiple Dan Yue enter an item of expenditure in the accounts the amount of money；At least one derivative data includes: expense summation, expense rate of change, using saturation
At least one of degree and stability.
In practical applications, the expense summation amount of money that can be entered an item of expenditure in the accounts by multiple Dan Yue sums up to obtain, and expense rate of change can
By multiple Dan Yue enter an item of expenditure in the accounts the amount of money carry out ring ratio obtain, can be obtained according to user behavior characteristics data using saturation degree and stability.
The embodiment of the present disclosure provides a kind of user's identification model generation method, and this method, which can be used for can recognize in model, to be used
On the basis of broadband user or non-broadband user is in family, processing is optimized to model, to reach better recognition effect.
Fig. 5 is the structural block diagram that the feature vectors that the embodiment of the present disclosure provides generate system.As shown in figure 5, the spy
Sign vector, which generates system, can be used for realizing feature vector generation method provided by the various embodiments described above, and this feature vector generates system
System includes: preprocessing module 1, the first conversion module 2, the second conversion module 3 and generation module 4.
Wherein, preprocessing module 1, it is original to obtain for being pre-processed to initial data according to default characteristic format
The corresponding initial characteristics vector of data.
First conversion module 2, for being character string type feature by the numeric type Feature Conversion in initial characteristics vector.
Second conversion module 3, for being converted to the initial characteristics vector for completing data type conversion according to preset algorithm
The corresponding discrete matrix comprising 1 and 0 two kind of discrete value.
Generation module 4, for generating the corresponding final feature vector of initial data according at least to discrete matrix.
In some embodiments, the first conversion module 2 includes: the first generation unit 2a.
First generation unit 2a, for according to initial characteristics vector generate it is corresponding with initial characteristics vector at least one
Derivative data.
Generation module 4 includes: the second generation unit 401.
Second generation unit 401 is merged for discrete matrix to be carried out field at least one derivative data, to generate original
The corresponding final feature vector of beginning data.
It should be noted that can to the connection between the specific implementation procedure and each module of each module in this present embodiment
Referring to the corresponding contents in preceding feature vector generation method embodiment, details are not described herein again.
The feature vector that the embodiment of the present disclosure provides generates system and can realize the numeric type Feature Conversion in feature vector
For character string type feature, and derivative data is generated according to initial characteristics vector, the two is merged, it is abundant final to form dimension
Feature vector may make model independent of the characteristic of numeric type feature, simplified model step when being applied in corresponding model
And calculation amount, lift scheme efficiency.
Fig. 6 is the structural block diagram that a kind of user's identification model that the embodiment of the present disclosure provides generates system.As shown in fig. 6,
User's identification model, which generates system, can be used for realizing user's identification model generation method provided by the various embodiments described above, the use
It includes: to obtain that module 5, any feature vector generates system and training module 7 in above-described embodiment that family identification model, which generates system,.
Wherein, module 5 is obtained, for obtaining the initial data of multiple broadband users and non-broadband user.
Any feature vector generates system 6 in above-described embodiment, for generating final feature corresponding to each initial data
Vector.
Training module 7, for, at least partly as training sample, to be trained corresponding in all final feature vectors
User's identification model.
Wherein, user's identification model can go out this feature vector according to the eigenvector recognition inputted corresponding to user
For broadband user or non-broadband user.
In some embodiments, training module 7 includes: training unit 701.
Training unit 701, for training corresponding use using the part in all final feature vectors as training sample
Family identification model.
User's identification model generates system further include: optimization module 8.
Optimization module 8, for using another part in all final feature vectors as test sample, to the use trained
Family identification model optimizes processing.
In some embodiments, it is in previous embodiment when included feature vector generates system 6, including first generates
When the feature vector of unit 2a and the second generation unit 401 generates system, initial data includes: user identifier, broadband contract mark
Knowledge, consumption data and user behavior characteristics data；At least one derivative data includes: expense summation, expense rate of change, uses
At least one of saturation degree and stability.
It should be noted that can to the connection between the specific implementation procedure and each module of each module in this present embodiment
Referring to the corresponding contents in aforementioned user's identification model generation method embodiment, details are not described herein again.
The embodiment of the present disclosure provide user's identification model generate system can realize using the feature in above-described embodiment to
It measures generation system and generates final feature vector, model is trained using the part in final feature vector as training sample,
Model is optimized using another part in final feature vector as test sample, the model after training and optimization can be used for
Identify that user is broadband user or non-broadband user.
It is understood that embodiment of above is merely to illustrate that the principle of the disclosure and the exemplary implementation that uses
Mode, however the disclosure is not limited thereto.For those skilled in the art, in the essence for not departing from the disclosure
In the case where mind and essence, various changes and modifications can be made therein, these variations and modifications are also considered as the protection scope of the disclosure.
Claims (10)
1. a feature vectors generation method characterized by comprising
Initial data is pre-processed according to default characteristic format, with obtain the corresponding initial characteristics of the initial data to
Amount；
It is character string type feature by the numeric type Feature Conversion in the initial characteristics vector；
The initial characteristics vector for completing data type conversion is converted to corresponding comprising 1 and 0 liang according to preset algorithm
The discrete matrix of kind discrete value；
The corresponding final feature vector of the initial data is generated according at least to the discrete matrix.
2. feature vector generation method according to claim 1, which is characterized in that described by the initial characteristics vector
In numeric type Feature Conversion be character string type feature the step of before, further includes:
At least one derivative data corresponding with the initial characteristics vector is generated according to the initial characteristics vector；
The step of final feature vector corresponding according at least to the discrete matrix generation initial data, specifically includes:
The discrete matrix is carried out field at least one described derivative data to merge, it is corresponding to generate the initial data
Final feature vector.
3. a kind of user's identification model generation method characterized by comprising
Obtain the initial data of multiple broadband users and non-broadband user；
Using feature vector generation method of any of claims 1 or 2, generate final feature corresponding to each initial data to
Amount；
, at least partly as training sample, to train corresponding user's identification model in all final feature vectors,
User's identification model can go out this feature vector according to the eigenvector recognition inputted corresponding to user be broadband use
Family or non-broadband user.
4. user's identification model generation method according to claim 3, which is characterized in that described with all final spies
Levy in vector at least partly as training sample, the step of training corresponding user's identification model, specifically include:
Using the part in all final feature vectors as training sample, corresponding user's identification model is trained；
After described the step of training corresponding user's identification model, further includes:
Using another part in all final feature vectors as test sample, to the user's identification model trained
Optimize processing.
5. user's identification model generation method according to claim 3 or 4, which is characterized in that when using claim 2 institute
Feature vector generation method is stated, when generating final feature vector corresponding to each initial data, the initial data includes: user
Mark, broadband Contract ID, user behavior characteristics data and multiple Dan Yue enter an item of expenditure in the accounts the amount of money；
At least one described derivative data include: expense summation, expense rate of change, using in saturation degree and stability at least
One.
6. a feature vectors generate system characterized by comprising
Preprocessing module, for being pre-processed to initial data according to default characteristic format, to obtain the initial data pair
The initial characteristics vector answered；
First conversion module, for being character string type feature by the numeric type Feature Conversion in the initial characteristics vector；
Second conversion module is converted to pair for will complete the initial characteristics vector of data type conversion according to preset algorithm
The only discrete matrix comprising 1 and 0 two kind of discrete value answered；
Generation module, for generating the corresponding final feature vector of the initial data according at least to the discrete matrix.
7. feature vector according to claim 6 generates system, which is characterized in that first conversion module includes:
First generation unit, for generating corresponding with the initial characteristics vector at least one according to the initial characteristics vector
A derivative data；
The generation module includes:
Second generation unit is merged for the discrete matrix to be carried out field at least one described derivative data, to generate
The corresponding final feature vector of the initial data.
8. a kind of user's identification model generates system characterized by comprising
Module is obtained, for obtaining the initial data of multiple broadband users and non-broadband user；
Feature vector described in claim 6 or 7 generate system, for generate final feature corresponding to each initial data to
Amount；
Training module, for, at least partly as training sample, to be trained corresponding in all final feature vectors
User's identification model, corresponding to user's identification model can go out this feature vector according to the eigenvector recognition inputted
User is broadband user or non-broadband user.
9. user's identification model according to claim 8 generates system, which is characterized in that the training module includes:
Training unit, for training corresponding user using the part in all final feature vectors as training sample
Identification model；
User's identification model generates system further include:
Optimization module, for using another part in all final feature vectors as test sample, to the institute trained
It states user's identification model and optimizes processing.
10. user's identification model according to claim 8 or claim 9 generates system, which is characterized in that when described eigenvector is raw
When generating system at system for feature vector as claimed in claim 7, the initial data includes: user identifier, broadband contract
Mark, consumption data and user behavior characteristics data；
At least one described derivative data include: expense summation, expense rate of change, using in saturation degree and stability at least
One.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910730490.3A CN110428016A (en) | 2019-08-08 | 2019-08-08 | Feature vector generation method and system, user's identification model generation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910730490.3A CN110428016A (en) | 2019-08-08 | 2019-08-08 | Feature vector generation method and system, user's identification model generation method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110428016A true CN110428016A (en) | 2019-11-08 |
Family
ID=68413318
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910730490.3A Pending CN110428016A (en) | 2019-08-08 | 2019-08-08 | Feature vector generation method and system, user's identification model generation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110428016A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009090584A2 (en) * | 2008-01-18 | 2009-07-23 | Koninklijke Philips Electronics N.V. | Method and system for activity recognition and its application in fall detection |
| CN106845999A (en) * | 2017-02-20 | 2017-06-13 | 百度在线网络技术（北京）有限公司 | Risk subscribers recognition methods, device and server |
| CN107396306A (en) * | 2017-06-30 | 2017-11-24 | 北京奇虎科技有限公司 | User Activity state identification method, device and mobile terminal based on mobile terminal |
| CN109191210A (en) * | 2018-09-13 | 2019-01-11 | 厦门大学嘉庚学院 | A kind of broadband target user's recognition methods based on Adaboost algorithm |
| CN109766932A (en) * | 2018-12-25 | 2019-05-17 | 新华三大数据技术有限公司 | A kind of Feature Selection method and Feature Selection device |
-
2019
- 2019-08-08 CN CN201910730490.3A patent/CN110428016A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009090584A2 (en) * | 2008-01-18 | 2009-07-23 | Koninklijke Philips Electronics N.V. | Method and system for activity recognition and its application in fall detection |
| CN106845999A (en) * | 2017-02-20 | 2017-06-13 | 百度在线网络技术（北京）有限公司 | Risk subscribers recognition methods, device and server |
| CN107396306A (en) * | 2017-06-30 | 2017-11-24 | 北京奇虎科技有限公司 | User Activity state identification method, device and mobile terminal based on mobile terminal |
| CN109191210A (en) * | 2018-09-13 | 2019-01-11 | 厦门大学嘉庚学院 | A kind of broadband target user's recognition methods based on Adaboost algorithm |
| CN109766932A (en) * | 2018-12-25 | 2019-05-17 | 新华三大数据技术有限公司 | A kind of Feature Selection method and Feature Selection device |
Non-Patent Citations (1)
| Title |
|---|
| 吴剑旗: "《先进米波雷达》", 31 August 2015, 国防工业出版社 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Pal et al. | Emotion detection from infant facial expressions and cries | |
| CN107967261A (en) | Interactive question semanteme understanding method in intelligent customer service | |
| CN108717439A (en) | A kind of Chinese Text Categorization merged based on attention mechanism and characteristic strengthening | |
| CN109543502A (en) | A kind of semantic segmentation method based on the multiple dimensioned neural network of depth | |
| CN106919951A (en) | A kind of Weakly supervised bilinearity deep learning method merged with vision based on click | |
| CN112269868B (en) | Use method of machine reading understanding model based on multi-task joint training | |
| CN110232280A (en) | A kind of software security flaw detection method based on tree construction convolutional neural networks | |
| CN110209789A (en) | A kind of multi-modal dialog system and method for user's attention guidance | |
| CN108009481A (en) | A kind of training method and device of CNN models, face identification method and device | |
| Li et al. | Improved sparse least-squares support vector machine classifiers | |
| CN109241518A (en) | A kind of detection network navy method based on sentiment analysis | |
| CN112801297B (en) | Machine learning model adversity sample generation method based on conditional variation self-encoder | |
| CN109657039A (en) | A kind of track record information extraction method based on the double-deck BiLSTM-CRF | |
| CN113157919B (en) | Sentence text aspect-level emotion classification method and sentence text aspect-level emotion classification system | |
| CN110263164A (en) | A kind of Sentiment orientation analysis method based on Model Fusion | |
| CN110415086A (en) | Intelligence financing recommended method based on user's Continuous behavior sequence signature | |
| CN113449303A (en) | Intelligent contract vulnerability detection method and system based on teacher-student network model | |
| CN115422518A (en) | Text verification code identification method based on data-free knowledge distillation | |
| CN110428016A (en) | Feature vector generation method and system, user's identification model generation method and system | |
| CN115334179A (en) | Unknown protocol reverse analysis method based on named entity recognition | |
| CN115482575A (en) | Facial expression recognition method based on label distribution learning | |
| CN113221551B (en) | Fine-grained sentiment analysis method based on sequence generation | |
| CN115291019A (en) | Electric energy quality disturbance classification and identification system based on multi-task learning model | |
| Meng et al. | Human Interaction Understanding with Consistency-Aware Learning | |
| CN108537206A (en) | A kind of face verification method based on convolutional neural networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191108 |