Grant: Discovery Grants Program - Individual
University of Waterloo
$14,000.00
Building flexible tree models and ensemble tree models for statistical learning in classification, regression and failure time data analysis
FY 2010-11
Summary
My proposed research is mainly in the direction of building flexible tree models and ensemble tree models for statistical learning in classification, regression and the statistical analysis of failure time data. The proposed research is expected to improve the information learning and prediction in physics, chemistry, drug discovery, biology, engineering and health sciences.
[1] I propose building tree models and ensemble tree models using kernel-induced split rules. This proposal is novel and promising. [2] I propose building survival tree models for failure time data with a genetic algorithm stochastic search. The survival tree models generally use a Cox proportional hazards assumption. [3] I propose building a competing risks survival tree model for failure time data with more than one failure type. This is a novel research idea. We not only estimate the relative risk parameters of the subpopulations in the terminal nodes of the tree but also estimate them according to different failure types. [4] The kernel-type of split rules proposed in [1] can also be used in [2] and [3] to further enhance the performances of the proposed tree models for survival and competing risks analysis. This research proposal is novel. [5] I propose building survival tree models in [2] with the relative risk parameters in the terminal nodes estimated using information from all observations. The specific technique I propose is the soft-node survival tree. [6] For rare target prediction problems which are important in drug discovery and fraud detection, I propose using a ranking-type criterion which is called the average precision to build classification tree models instead of using the traditional Gini index or entropy measure. The proposal is novel and potentially superior to traditional models in this sense.
In summary, I propose flexible tree models and ensemble tree models to solve various practical and theoretical problems in classification, regression and failure time data analysis. Some of the above proposed projects are going to be assigned to graduate students I supervise or co-supervise in the future.
Program
Discovery Grants Program - Individual
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
311980-2008
Name nom
Fan, Guangzhe
Department département
Statistics and Actuarial Science
Organization id
33
Institution établissement
University of Waterloo
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2010
Competition year année de concours
2008
Award amount
14000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
14
Committee name en
Statistical Sciences
Committee name fr
Statistique
Area of application code
1205
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Mathematical sciences
Area of application fr
Sciences mathématiques
Research subject code
3000
Research subject group en
Statistics and probability
Research subject group fr
Statistique et probabilité
Research subject en
Statistics and probability
Research subject fr
Statistique et probabilité
Partie
3
Nb partie
5
Application title
Building flexible tree models and ensemble tree models for statistical learning in classification, regression and failure time data analysis
Keyword
NSERC
Application summary
My proposed research is mainly in the direction of building flexible tree models and ensemble tree models for statistical learning in classification, regression and the statistical analysis of failure time data. The proposed research is expected to improve the information learning and prediction in physics, chemistry, drug discovery, biology, engineering and health sciences.
[1] I propose building tree models and ensemble tree models using kernel-induced split rules. This proposal is novel and promising. [2] I propose building survival tree models for failure time data with a genetic algorithm stochastic search. The survival tree models generally use a Cox proportional hazards assumption. [3] I propose building a competing risks survival tree model for failure time data with more than one failure type. This is a novel research idea. We not only estimate the relative risk parameters of the subpopulations in the terminal nodes of the tree but also estimate them according to different failure types. [4] The kernel-type of split rules proposed in [1] can also be used in [2] and [3] to further enhance the performances of the proposed tree models for survival and competing risks analysis. This research proposal is novel. [5] I propose building survival tree models in [2] with the relative risk parameters in the terminal nodes estimated using information from all observations. The specific technique I propose is the soft-node survival tree. [6] For rare target prediction problems which are important in drug discovery and fraud detection, I propose using a ranking-type criterion which is called the average precision to build classification tree models instead of using the traditional Gini index or entropy measure. The proposal is novel and potentially superior to traditional models in this sense.
In summary, I propose flexible tree models and ensemble tree models to solve various practical and theoretical problems in classification, regression and failure time data analysis. Some of the above proposed projects are going to be assigned to graduate students I supervise or co-supervise in the future.