Grant: Discovery Grants Program - Individual
Université Laval
$38,000.00
Taking advantage of the longitudinal or spatial correlation structure with complex or unstructured data
FY 2024-25
Summary
The objective of this research program is to develop methods for the analysis of data with a complex structure that can also handle certain types of unstructured data such as texts or simple images (polygons). The idea is to allow analysts to take full advantage of the connections among units or individuals in a database and of the information that could be available in the form of polygons or texts. Two important applications of those methods will drive the research. First, because the daily perimeters of a wildfire are often recorded as polygons on a map, we need to develop methods that can use these types of records to calibrate more realistic and efficient wildfire long-range simulation models. Second, email exchanges between a client and a company, or medical records of patients at a clinic, contain important information on the satisfaction of the clients with the services that they receive or on how the patients cope with their treatments, and this information should be included in models that analysts use to determine the best services or the best treatments for their clients or patients. It is also hoped that methods that extract information from text can, when properly reweighted, lead to predictions or decisions that will be as efficient for minority groups as for majority groups.
While there has been much research on how to include textual or image data into statistical learning algorithms and just as much work on how to handle correlated data, not much has been done to try to take advantage of both of these sources of information simultaneously. The first application is important, as wildfire spread models for short-term simulations are computationally heavy and are not best suited to assess long-term risk, such as the probability that a town or infrastructure will be reached by a wildfire at least once over the next 10 years. Moreover, methods to calibrate the spread models do not generally make full use of newly available data obtained by satellite photos or remote-sensing. The second application also has value, because information on how clients are satisfied with a product or how patients react to a treatment can sometimes only be found in textual data, and methods to extract information from text and use it in models that can handle correlated individuals do not exist.
The potential benefits to Canada are important. In the case of the first application, more than 200,000 Canadians have been evacuated from their homes this summer due to the threat of wildfires. The availability of more accurate long-range risk assessment models should enable governments, aboriginal communities, insurance companies and other entities to provide Canadians with improved protection, infrastructure and guidelines to adapt to the rapidly evolving risk that wildfires pose to many of them. For the second application, the tools developed should help Canadians receive offers, products or treatments more beneficial for their financial and physical health.
Program
Discovery Grants Program - Individual
Recipient
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2024-04204
Name nom
Duchesne, Thierry
Department département
Mathématiques et statistique
Organization id
51
Institution établissement
Université Laval
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
38000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
Area of application code
1002
Area of application group en
Northern development
Area of application group fr
Développement du nord
Area of application en
Environment
Area of application fr
Environnement
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
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Taking advantage of the longitudinal or spatial correlation structure with complex or unstructured data
Keywords
longitudinal data; time-to-event data; correlation structure; joint modeling; unstructured data; spatial data; natural language processing; wildfires; insurance; networks
Application summary
The objective of this research program is to develop methods for the analysis of data with a complex structure that can also handle certain types of unstructured data such as texts or simple images (polygons). The idea is to allow analysts to take full advantage of the connections among units or individuals in a database and of the information that could be available in the form of polygons or texts. Two important applications of those methods will drive the research. First, because the daily perimeters of a wildfire are often recorded as polygons on a map, we need to develop methods that can use these types of records to calibrate more realistic and efficient wildfire long-range simulation models. Second, email exchanges between a client and a company, or medical records of patients at a clinic, contain important information on the satisfaction of the clients with the services that they receive or on how the patients cope with their treatments, and this information should be included in models that analysts use to determine the best services or the best treatments for their clients or patients. It is also hoped that methods that extract information from text can, when properly reweighted, lead to predictions or decisions that will be as efficient for minority groups as for majority groups.
While there has been much research on how to include textual or image data into statistical learning algorithms and just as much work on how to handle correlated data, not much has been done to try to take advantage of both of these sources of information simultaneously. The first application is important, as wildfire spread models for short-term simulations are computationally heavy and are not best suited to assess long-term risk, such as the probability that a town or infrastructure will be reached by a wildfire at least once over the next 10 years. Moreover, methods to calibrate the spread models do not generally make full use of newly available data obtained by satellite photos or remote-sensing. The second application also has value, because information on how clients are satisfied with a product or how patients react to a treatment can sometimes only be found in textual data, and methods to extract information from text and use it in models that can handle correlated individuals do not exist.
The potential benefits to Canada are important. In the case of the first application, more than 200,000 Canadians have been evacuated from their homes this summer due to the threat of wildfires. The availability of more accurate long-range risk assessment models should enable governments, aboriginal communities, insurance companies and other entities to provide Canadians with improved protection, infrastructure and guidelines to adapt to the rapidly evolving risk that wildfires pose to many of them. For the second application, the tools developed should help Canadians receive offers, products or treatments more beneficial for their financial and physical health.