CN104899430A - Multisource example transfer learning-based terror act prediction method - Google Patents
Multisource example transfer learning-based terror act prediction method Download PDFInfo
- Publication number
- CN104899430A CN104899430A CN201510246797.8A CN201510246797A CN104899430A CN 104899430 A CN104899430 A CN 104899430A CN 201510246797 A CN201510246797 A CN 201510246797A CN 104899430 A CN104899430 A CN 104899430A
- Authority
- CN
- China
- Prior art keywords
- tissue
- candidate prediction
- source
- sample
- behavior
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a multisource example transfer learning-based terror act prediction method, comprising the steps of: preprocessing background data, wherein the background data consists of background knowledge and behavior knowledge and marking as a vector pair. In order to realize transfer among organizations, the background data are preprocessed, and multiple behavior attributes are combined into one behavior attribute; candidate prediction models are trained, union sets are taken from multiple source organization sample sets and a target organization sample set, iteration training is performed on each union set by adopting an SVM, and wrong samples are subject to weight correction in each time of iteration training to obtain a candidate prediction model set; a final prediction model is integrated, a part exceeding a threshold value relative to the error of the target data set in the candidate prediction model sets is filtered, and surplus models are integrated in a weighing voting manner. The prediction method aims at the behavior prediction of newly generated terror organizations, and solves the problem that the newly generated organizations are hard to predict due to lack of samples, and the prediction precision is improved.
Description
Technical field
The present invention relates to computer data to excavate and application, in particular to one based on the terroristic prediction algorithm of multi-source instance migration study prediction.
Background technology
On September 11 calendar year 2001, USA New York, Washington suffer terrorist attacks, cause people more than 3100 dead.This time event is considered to the most serious attack of terrorism that the U.S. after Pearl Harbor Incident is subjected to, and indicates that terrorist's organization has become and causes significant impact non-countries ' power to world security.The act of terrorism how utilizing existing information prediction to occur, becomes an important research direction.
Act of terrorism prediction is the typical apply that prediction knowledge excavates, and it utilizes the correlation technique of data mining and machine learning, carries out the statistical study of science, then predict its development trend to the terroristic situation of past, the enforcement of terroristic organization's planning now.Terrified prediction is not confirm in the past, reality to be described, but to start with from the data of dominant terroristic organization and the attack of terrorism, find out the useful information of some recessiveness.The feature may hidden from attack of terrorism data, behavior or other because usually extracting relevant pattern, give a clue with this, prediction terrorist's organization development trend.Its object is to as taking effective preventive measure to provide decision support.
In early days mainly to the act of terrorism, long-term forecasting is carried out for the statistics of authorities to the research of terror prediction, but because data message at that time only considers the factor such as time, place, target that terrorist incident occurs, and the factor not considering to cause the sociology of terrorist incident etc. profound, can not effectively to predict it from data-driven therefore merely.And traditional analytical approach too relies on the analysis of sociology expert, does not possess operability for big data quantity.
At present, start with interdisciplinary (computer data digging technology and sociology to the research of terror prediction, criminology etc.) based on, the information of statistics specifically, not only comprise the information that traditional terrorist activity occurs, but also gather information and data mining from aspects such as economic conflict, political contradiction (error as national conflict, religious values difference and ethnic policy), cultural contradictions so that by the analysis of these data for authorities provide more effective forecast analysis.Therefore, the impact of contextual factor on its behavior by analyzing terroristic organization becomes the focus of research.
In current research, the research object chosen often life period for a long time and sample size compared with the terroristic organization of horn of plenty, but along with terroristic globalization, constantly have new terroristic organization to produce in recent years.The data that this class loading is collected because generation time short-range missile causes are very rare, make to lack foundation for this histioid prediction, and precision of prediction is low.But, be correlated with often between terroristic organization, in the process of their Existence and development, have general character more or less.Therefore how effectively can utilize the general character between tissue, help new the carrying out producing tissue and predict, be the instant problem that the prediction of the current act of terrorism needs to solve.
Summary of the invention
For problems of the prior art, the present invention is intended to propose a kind of act of terrorism Forecasting Methodology based on the study of multi-source instance migration, by carrying out instance migration study between tissue, utilize the useful knowledge in source tissue to help destination organization and carry out behavior prediction, solve the new tissue that produces to lack according to the low problem of the precision of prediction caused because sample rareness makes to predict, effectively raise the degree of accuracy of prediction algorithm.
For reaching above-mentioned purpose, technical solution of the present invention is:
Based on an act of terrorism Forecasting Methodology for multi-source instance migration study, comprise the following steps:
Step 1, the pre-service of background data: background data is made up of background knowledge and behavior knowledge, is labeled as vector to (CS, AS), wherein CS=(C
1, C
2..., C
m) represent background attribute in background data, AS=(A
1, A
2..., A
n) represent the behavior property related in background data, in order to realize migration between tissue, pre-service being carried out to background data, multiple behavior property is merged into a behavior property, form the data set of (CS, A), wherein A=A
1|| A
2|| ... || A
nfor the behavior property merged;
Step 2, training candidate prediction model: obtain shape as (CS by step 1, A) source tissue and destination organization sample set, with destination organization sample set, union is got respectively to multiple source tissues sample set, each union adopt SVM carry out the training of iteration, all divide sample to carry out weight correction to mistake in each repetitive exercise, and then obtain candidate prediction Models Sets;
Step 3, integrated final forecast model: the part in the candidate prediction Models Sets that filtration step 2 obtains, the error of target data set being exceeded to threshold value, remaining model realizes behavior prediction in the mode of Nearest Neighbor with Weighted Voting.
Further, in step 2, candidate prediction model is trained specifically to comprise the following steps:
Step 2.1, the weight vectors of initialization source tissue and destination organization
wherein
for kGe source tissue sample weights vector,
for the sample weights vector of destination organization, in order to avoid weight mismatch problem, give higher weight to destination organization sample;
Step 2.2, for first source tissue's sample set
with destination organization sample set D
tget union
?
the training of enterprising row iteration obtains the candidate prediction model of first source tissue;
Step 2.3, training institute's active tissue being completed to iteration obtains candidate prediction Models Sets G.
Further, in step 2.2, the training of iteration comprises the following steps:
S1 is right
sample weights vector
utilize formula
carry out weight normalized, make the weight sum of all samples be 1;
S2, adopts svm classifier algorithm to exist according to the weight distribution after normalization
upper training classification forecast model h:C → A;
S3, utilizes formula
calculate h at destination organization sample set D
ton error ε;
S4, if
wherein n
sfor source tissue's sample size, utilize formula
revise source tissue's sample weights
S5, if α
t=ε/(1-ε), in order to ensure α
t> 1, so ε must be less than 0.5, utilizes formula
revise goal tissue samples weights W
t;
S6, error F ← (h, the ε) that record forecast model is corresponding completes first round iteration, and carries out second according to revised sample weights and take turns iteration, until cause maximum iteration time M;
S7, according to formula
obtain the candidate prediction model of this source tissue, according to formula
obtain candidate prediction model at D
ton average error, record candidate prediction model and average error G ← (h
sum, ε
avg).
Further, in described step S2, svm classifier algorithm is standard svm classifier algorithm, and the kernel function of the SVM of use is gaussian kernel function.
Further, in described step 3, integrated final forecast model specifically comprises the following steps:
Step 3.1, carries out zero setting process h according to threshold value thr to the model that its average error in the forecast model in candidate prediction Models Sets G exceedes threshold value
sum=0;
Step 3.2, according to source tissue's weight factor
formula is utilized to be weighted integrated
wherein
for the weight factor of kGe source tissue, if current time is t, then
result (i) in formula, for kGe source tissue is in the result of moment i to destination organization migration prediction, is successfully 1, otherwise is 0; F (x) is time attenuation function, and x is larger, and then f (x) is less; And give tacit consent to result (0)=1.
From the above technical solution of the present invention shows that, beneficial effect of the present invention is: the act of terrorism Forecasting Methodology based on the study of multi-source instance migration that the present invention proposes, and is assisted the behavior prediction of the new tissue produced by the sample moving relevant tissue.In the process of training, revise the weight of wrong point sample iteratively, the sample weights that obstruction goal task is predicted reduces, and helps large sample to have higher weight to goal task.And by arranging strobe utility and being weighted the integrated impact reducing mismatching model according to source tissue's weight factor, make the useful information in linked groups can participate in the behavior prediction of new generation tissue as much as possible, improve the precision of the behavior prediction for the new tissue produced with this, enhance the performance of prediction algorithm.
Accompanying drawing explanation
Fig. 1 is the prediction schematic flow sheet of the embodiment of the present invention.
Fig. 2 is the overall framework schematic diagram of prediction.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely described.
Be described below in detail embodiments of the invention, the example of described embodiment is shown in the drawings, and wherein same or similar label represents same or similar element or has element that is identical or similar functions from start to finish.Being exemplary below by the embodiment be described with reference to the drawings, only for explaining the present invention, and can not limitation of the present invention being interpreted as.
As shown in Figure 1, according to embodiments of the invention, the act of terrorism Forecasting Methodology based on terrorist organization background knowledge subspace comprises three basic steps: the pre-service of background data; Training candidate prediction model; Integrated final forecast model.
One, the pre-service of background data
Background data is made up of background knowledge and behavior knowledge, is labeled as vector to (CS, AS), wherein CS=(C
1, C
2..., C
m) represent background attribute in background data, AS=(A
1, A
2..., A
n) represent the behavior property related in background data.In order to realize migration between tissue, pre-service being carried out to background data, multiple behavior property is merged into a behavior property and A=A
1|| A
2|| ... || A
n, form the data set of (CS, A);
Background data subset as shown in table 1 below, has 11 fields, is labeled as ID, C respectively
1, C
2, C
3, C
4, C
5, C
6and A
1, A
2, A
3, Action.ID is labeled as the numbering be recorded in table.(C
1, C
2, C
3, C
4, C
5, C
6)=CS (g) represents background attribute, (A
1, A
2, A
3)=AS (g) represents behavior property, and Action is A
1, A
2, A
3the behavior property merged.Attribute in this example for predicting is C
1, C
2, C
3, C
4, C
5, C
6and Action, wherein C
1, C
2, C
3, C
4, C
5, C
6for background characteristics, Action is class label.
Two, candidate prediction model is trained
Training candidate prediction model, shape is obtained as (CS by pre-service, A) source tissue and destination organization sample set, to the multiple source tissue tissue of sample-rich (existing) sample set respectively same destination organization (the new tissue produced) sample set get union, each union adopt SVM to train iteratively, all divide sample to carry out weight correction to mistake in each repetitive exercise, and then obtain candidate prediction Models Sets, specifically comprise the following steps:
Step 1: the weight vectors of initialization source tissue and destination organization
wherein
for kGe source tissue sample weights vector,
for the sample weights vector of destination organization.In order to avoid weight mismatch problem, give higher weight to destination organization sample;
Step 2: for first source tissue's sample set
with destination organization sample set D
tget union
?
the training of enterprising row iteration obtains the candidate prediction model of first source tissue;
In the present embodiment, the training of data iteration is comprised the following steps:
1.1 right
sample weights vector
utilize formula
carry out weight normalized, make the weight sum of all samples be 1;
1.2 adopt support vector machine (SVM) sorting algorithm to exist according to the weight distribution after normalization
upper training classification forecast model h:C → A; The SVM that wherein the present embodiment adopts is standard SVM, and the kernel function of use is gaussian kernel function.
1.3 utilize formula
calculate h at destination organization sample set D
ton error ε;
1.4 establish
wherein n
sfor source tissue's sample size.Utilize formula
revise source tissue's mistake point sample weights
1.5 establish α
t=ε/(1-ε), in order to ensure α
t> 1, so ε must be less than 0.5.Utilize formula
revise goal tissue mistake point sample weights W
t;
Error F ← (h, the ε) that 1.6 record forecast models are corresponding complete first round iteration, and carry out second according to revised sample weights and take turns iteration, until reach maximum iteration time M; Maximum iteration time wherein in the present embodiment is set as 20.
1.7 according to formula
obtain the candidate prediction model of this source tissue, according to formula
obtain candidate prediction model at D
ton average error.Record candidate prediction model and average error G ← (h
sum, ε
avg).
Step 3: training institute's active tissue being completed to iteration obtains candidate prediction Models Sets G.
Three, integrated final forecast model
In the present embodiment, integrated final forecast model comprises the following steps:
Step 1: zero setting process hsum=0 is carried out to the model that its average error in the forecast model in candidate prediction Models Sets G exceedes threshold value according to threshold value thr; Thr wherein in the present embodiment is set to 0.3.
Step 2: as shown in Figure 2, according to source tissue's weight factor
formula is utilized to be weighted integrated
wherein
for the weight factor of kGe source tissue, if current time is t, then
result (i) in formula, for kGe source tissue is in the result of moment i to destination organization migration prediction, is successfully 1, otherwise is 0; F (x) is time attenuation function, and x is larger, and then f (x) is less; And give tacit consent to result (0)=1.The time attenuation function used in this example is f (x)=0.9
x.
The statistics predicted and as shown in table 2 with the contrast predicted the outcome of other traditional algorithm is carried out according to the present embodiment.The main evaluation criteria that algorithm evaluation adopts mainly comprises recall ratio (recall), precision ratio (precision) and F value (F-measure).What recall ratio was assessed is that in sample to be tested, minority class predicts successful ratio; Precision ratio assessment be the ratio predicted in the sample being predicted as minority class shared by correct sample; F value is the comprehensive assessment based on recall ratio and precision ratio.Above-mentioned several standard is also widely applied in the field such as text classification, information retrieval.
As can be seen from the experimental result shown in table 2, have employed the prediction effect of the behavior prediction method based on the study of multi-source instance migration in this paper significantly better than traditional prediction algorithm, because CONVEX and SVM does not move to any knowledge from source tissue, a small amount of sample of destination organization is only relied on to cause prediction effect low.And the SVM prediction effect of structure based least risk to compare CONVEX algorithm slightly better.
In sum, the present invention proposes a kind of act of terrorism Forecasting Methodology based on the study of multi-source instance migration, comprises the following steps: step 1: the pre-service of background data, background data is made up of background knowledge and behavior knowledge, be labeled as vector to (CS, AS), wherein CS=(C
1, C
2..., C
m) represent background attribute in background data, AS=(A
1, A
2..., A
n) represent the behavior property related in background data.In order to realize migration between tissue, pre-service being carried out to background data, multiple behavior property is merged into a behavior property, form the data set of (CS, A); Step 2: training candidate prediction model, with destination organization sample set, union is got respectively to multiple source tissues sample set, each union adopts SVM to train iteratively, in each repetitive exercise, all divides sample to carry out weight correction to mistake, and then obtain candidate prediction Models Sets; Step 3: integrated final forecast model, filter candidate forecast model concentrates the error for target data set to exceed the part of threshold value, and remaining model carries out integrated in the mode of Nearest Neighbor with Weighted Voting, finally realizes behavior prediction.Forecasting Methodology of the present invention is the behavior prediction for the new terroristic organization produced, and solves the problem that the new tissue produced causes owing to lacking sample being difficult to predict, improves the precision of prediction.
Act of terrorism Forecasting Methodology based on the study of multi-source instance migration of the present invention, utilizes the knowledge of linked groups to help the new tissue produced and carries out behavior prediction.And by the wrong point sample weights of the correction of iteration, strobe utility is set and adds the information that source tissue's weight factor makes source tissue useful and participate in as far as possible among goal task prediction, effectively solve the new tissue produced because generation time is short, lack the problem low to the behavior prediction precision organized that usable samples causes.
In the description of this instructions, specific features, structure, material or feature that the description of reference term " embodiment ", " some embodiments ", " illustrative examples ", " example ", " concrete example " or " some examples " etc. means to describe in conjunction with this embodiment or example are contained at least one embodiment of the present invention or example.In this manual, identical embodiment or example are not necessarily referred to the schematic representation of above-mentioned term.And the specific features of description, structure, material or feature can combine in an appropriate manner in any one or more embodiment or example.
Although illustrate and describe embodiments of the invention, those having ordinary skill in the art will appreciate that: can carry out multiple change, amendment, replacement and modification to these embodiments when not departing from principle of the present invention and aim, scope of the present invention is by claim and equivalents thereof.
Claims (5)
1., based on an act of terrorism Forecasting Methodology for multi-source instance migration study, it is characterized in that, comprise the following steps:
Step 1, the pre-service of background data: background data is made up of background knowledge and behavior knowledge, is labeled as vector to (CS, AS), wherein CS=(C
1, C
2..., C
m) represent background attribute in background data, AS=(A
1, A
2..., A
n) represent the behavior property related in background data, in order to realize migration between tissue, pre-service being carried out to background data, multiple behavior property is merged into a behavior property, form the data set of (CS, A), wherein A=A
1|| A
2|| ... || A
nfor the behavior property merged;
Step 2, training candidate prediction model: obtain shape as (CS by step 1, A) source tissue and destination organization sample set, with destination organization sample set, union is got respectively to multiple source tissues sample set, each union adopt SVM carry out the training of iteration, all divide sample to carry out weight correction to mistake in each repetitive exercise, and then obtain candidate prediction Models Sets;
Step 3, integrated final forecast model: the part in the candidate prediction Models Sets that filtration step 2 obtains, the error of target data set being exceeded to threshold value, remaining model realizes behavior prediction in the mode of Nearest Neighbor with Weighted Voting.
2. the act of terrorism Forecasting Methodology based on the study of multi-source instance migration according to claim 1, is characterized in that, in step 2, training candidate prediction model specifically comprises the following steps:
Step 2.1, the weight vectors of initialization source tissue and destination organization
wherein
for kGe source tissue sample weights vector,
for the sample weights vector of destination organization, in order to avoid weight mismatch problem, give higher weight to destination organization sample;
Step 2.2, for first source tissue's sample set
with destination organization sample set D
tget union
?
the training of enterprising row iteration obtains the candidate prediction model of first source tissue;
Step 2.3, training institute's active tissue being completed to iteration obtains candidate prediction Models Sets G.
3. the act of terrorism Forecasting Methodology based on the study of multi-source instance migration according to claim 2, it is characterized in that, in step 2.2, the training of iteration comprises the following steps:
S1 is right
sample weights vector
Utilize formula
Carry out weight normalized, make the weight sum of all samples be 1;
S2, adopts svm classifier algorithm to exist according to the weight distribution after normalization
upper training classification forecast model h:C → A;
S3, utilizes formula
calculate h at destination organization sample set D
ton error ε;
S4, if
Wherein n
sfor source tissue's sample size, utilize formula
Revise source tissue's sample weights
S5, if α
t=ε/(1-ε), in order to ensure α
t> 1, so ε must be less than 0.5, utilizes formula
revise goal tissue samples weights W
t;
S6, error F ← (h, the ε) that record forecast model is corresponding completes first round iteration, and carries out second according to revised sample weights and take turns iteration, until cause maximum iteration time M;
S7, according to formula
obtain the candidate prediction model of this source tissue, according to formula
obtain candidate prediction model at D
ton average error, record candidate prediction model and average error G ← (h
sum, ε
avg).
4. the act of terrorism Forecasting Methodology based on the study of multi-source instance migration according to claim 3, it is characterized in that, in described step S2, svm classifier algorithm is standard svm classifier algorithm, and the kernel function of the SVM of use is gaussian kernel function.
5. the act of terrorism Forecasting Methodology based on the study of multi-source instance migration according to claim 1, it is characterized in that, in described step 3, integrated final forecast model specifically comprises the following steps:
Step 3.1, carries out zero setting process h according to threshold value thr to the model that its average error in the forecast model in candidate prediction Models Sets G exceedes threshold value
sum=0;
Step 3.2, according to source tissue's weight factor
formula is utilized to be weighted integrated
wherein
for the weight factor of kGe source tissue, if current time is t, then
result (i) in formula, for kGe source tissue is in the result of moment i to destination organization migration prediction, is successfully 1, otherwise is 0; F (x) is time attenuation function, and x is larger, and then f (x) is less; And give tacit consent to result (0)=1.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510246797.8A CN104899430A (en) | 2015-05-14 | 2015-05-14 | Multisource example transfer learning-based terror act prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510246797.8A CN104899430A (en) | 2015-05-14 | 2015-05-14 | Multisource example transfer learning-based terror act prediction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104899430A true CN104899430A (en) | 2015-09-09 |
Family
ID=54032092
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510246797.8A Pending CN104899430A (en) | 2015-05-14 | 2015-05-14 | Multisource example transfer learning-based terror act prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104899430A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106776884A (en) * | 2016-11-30 | 2017-05-31 | 江苏大学 | A kind of act of terrorism Forecasting Methodology that multi-categorizer is combined based on multi-tag |
CN108109096A (en) * | 2017-12-29 | 2018-06-01 | 航天科工智慧产业发展有限公司 | A kind of method of wisdom inspection business prevention occupational crime |
CN108256759A (en) * | 2018-01-11 | 2018-07-06 | 中国科学院地理科学与资源研究所 | A kind of attack of terrorism risk Temporal Simulation and intelligent analysis method |
CN108764537A (en) * | 2018-05-14 | 2018-11-06 | 浙江工业大学 | A kind of multi-source community label prediction of the development trend method based on A-TrAdaboost algorithms |
CN109214436A (en) * | 2018-08-22 | 2019-01-15 | 阿里巴巴集团控股有限公司 | A kind of prediction model training method and device for target scene |
CN110969197A (en) * | 2019-11-22 | 2020-04-07 | 上海交通大学 | Quantile prediction method for wind power generation based on instance migration |
CN114999182A (en) * | 2022-05-25 | 2022-09-02 | 中国人民解放军国防科技大学 | Vehicle flow prediction method, device and equipment based on LSTM feedback mechanism |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102955894A (en) * | 2012-05-24 | 2013-03-06 | 华东师范大学 | Customer segmentation-based method for controlling churn rate prediction |
CN103065047A (en) * | 2012-12-27 | 2013-04-24 | 江苏大学 | Terrorism behavior prediction method based on terrorist organization background knowledge subspace |
US20140324752A1 (en) * | 2013-03-15 | 2014-10-30 | Alexander Statnikov | Data Analysis Computer System and Method For Fast Discovery Of Multiple Markov Boundaries |
-
2015
- 2015-05-14 CN CN201510246797.8A patent/CN104899430A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102955894A (en) * | 2012-05-24 | 2013-03-06 | 华东师范大学 | Customer segmentation-based method for controlling churn rate prediction |
CN103065047A (en) * | 2012-12-27 | 2013-04-24 | 江苏大学 | Terrorism behavior prediction method based on terrorist organization background knowledge subspace |
US20140324752A1 (en) * | 2013-03-15 | 2014-10-30 | Alexander Statnikov | Data Analysis Computer System and Method For Fast Discovery Of Multiple Markov Boundaries |
Non-Patent Citations (1)
Title |
---|
张倩: "基于知识表达的迁移学习研究", 《中国博士学位论文全文数据库信息科技辑》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106776884A (en) * | 2016-11-30 | 2017-05-31 | 江苏大学 | A kind of act of terrorism Forecasting Methodology that multi-categorizer is combined based on multi-tag |
CN106776884B (en) * | 2016-11-30 | 2021-04-20 | 江苏大学 | Terrorism prediction method based on multi-label combination and multi-classifier |
CN108109096A (en) * | 2017-12-29 | 2018-06-01 | 航天科工智慧产业发展有限公司 | A kind of method of wisdom inspection business prevention occupational crime |
CN108256759A (en) * | 2018-01-11 | 2018-07-06 | 中国科学院地理科学与资源研究所 | A kind of attack of terrorism risk Temporal Simulation and intelligent analysis method |
CN108764537A (en) * | 2018-05-14 | 2018-11-06 | 浙江工业大学 | A kind of multi-source community label prediction of the development trend method based on A-TrAdaboost algorithms |
CN108764537B (en) * | 2018-05-14 | 2021-11-23 | 浙江工业大学 | A-TrAdaboost algorithm-based multi-source community label development trend prediction method |
CN109214436A (en) * | 2018-08-22 | 2019-01-15 | 阿里巴巴集团控股有限公司 | A kind of prediction model training method and device for target scene |
CN110969197A (en) * | 2019-11-22 | 2020-04-07 | 上海交通大学 | Quantile prediction method for wind power generation based on instance migration |
CN110969197B (en) * | 2019-11-22 | 2022-01-04 | 上海交通大学 | Quantile prediction method for wind power generation based on instance migration |
CN114999182A (en) * | 2022-05-25 | 2022-09-02 | 中国人民解放军国防科技大学 | Vehicle flow prediction method, device and equipment based on LSTM feedback mechanism |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104899430A (en) | Multisource example transfer learning-based terror act prediction method | |
Pollock et al. | Large conservation gains possible for global biodiversity facets | |
Jetz et al. | The interplay of past diversification and evolutionary isolation with present imperilment across the amphibian tree of life | |
Banker et al. | Some models for estimating technical and scale inefficiencies in data envelopment analysis | |
Zia et al. | Cross-scale value trade-offs in managing social-ecological systems: the politics of scale in Ruaha National Park, Tanzania | |
Maypa et al. | Marine protected area management effectiveness: progress and lessons in the Philippines | |
CN106776884A (en) | A kind of act of terrorism Forecasting Methodology that multi-categorizer is combined based on multi-tag | |
Fuller et al. | Spatial assessment of threats to biodiversity within East Kalimantan, Indonesia | |
CN107180283A (en) | A kind of behavior prediction system and method for being in hospital again combined based on optimal characteristics | |
CN110347836A (en) | Incorporate the more bilingual the emotion of news classification method of the Chinese of viewpoint sentence feature | |
Sun et al. | Financial distress prediction based on similarity weighted voting CBR | |
Pelletier et al. | REDD+ emissions estimation and reporting: dealing with uncertainty | |
CN105760633A (en) | Green architectural design method applicable to information assistance | |
Zhang et al. | Scientific relatedness and intellectual base: a citation analysis of un-cited and highly-cited papers in the solar energy field | |
Donovan et al. | Safety in numbers: cost-effective endangered species management for viable populations | |
Narawish et al. | Importance of cost efficiency in critical aspect of influences the decision-making process in banks | |
Nost | Performing nature's value: software and the making of Oregon's ecosystem services markets | |
Shan et al. | Incorporating user behavior flow for user risk assessment | |
Davidson | Labour on the leading edge: A critical review of labour rights and standards in renewable energy | |
Arunachalam | Optimization using differential evolution | |
CN105956982A (en) | Method of predicting act of terror based on background change | |
CN106202299A (en) | A kind of people with disability authority user based on people with disability's feature recommends method | |
CN107644268B (en) | Open source software project incubation state prediction method based on multiple features | |
Amici et al. | Selecting focal species in ecological network planning following an expert-based approach: a case study and a conceptual framework | |
CN103065047A (en) | Terrorism behavior prediction method based on terrorist organization background knowledge subspace |
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: 20150909 |