Grant: Discovery Grants Program - Individual
University of Alberta
$36,000.00
Towards Efficient and Automated Unsupervised and Semi-Supervised Data Analytics
FY 2024-25
Summary
The proposed research program will investigate and advance theory and practice of unsupervised and semi-supervised methods for finding and analyzing "structure" in multi-dimensional data. Many of today's data sets are multidimensional and are typically collected on a large scale, where none or only a small fraction of labelled data is available. These data sets hold the potential for discovering new insights that can lead to new opportunities for improved, data-driven decision making. Exploratory methods for unsupervised and semi-supervised clustering and outlier detection can play an important role in realizing this potential. They are often a first step in a data analytics pipeline that can establish insight into the data distribution, and can be used as a pre-processing step for other types of analyses and tasks. Understanding the (possibly hierarchical) clustering structure and outliers in the data allows for a better data organization (e.g., for more efficient browsing and searching), can lead to the discovery of unknown relationships between groups of data, or can help focus subsequent data analysis on specific subgroups or data points.
Applications of clustering and outlier detection often face significant challenges: (1) the data is often large in size (in number of records and number of attributes/dimensions), while (2) the methods typically do not scale well and (3) require significant involvement of an analyst, who has to choose parameter values for a method and a given data set, and frequently wishes to re-run analyses using different parameter values - which, however, is limited due to (1) and (2).
My proposed research program is trying to advance the theoretical foundations and applicability of clustering and outlier detection in unsupervised and semi-supervised settings - particularly in the density-based paradigm where clusters are considered dense regions in the data space, separated by regions of lower density. We will focus on the aspects of (A) Scalability, in order to deal with large data volumes and to enable fast computation of results for a wide range of parameter values of a method to lay the foundation for drawing more robust conclusions from multiple results and for automating model selection/construction; (B) Automated Model Selection/Construction, based on internal cluster and outlier validation measures, in both unsupervised and semi-supervised scenarios; (C) Modeling the development of cluster structures and outliers over time, in order to adapt results efficiently in dynamic application scenarios, discover and track structural changes in a single data set over time, and enable the discovery of how changes in different data sets can be related to each other.
Based on the gained theoretical insights, we will develop novel and improved algorithms for clustering and outlier detection that overcome some of the limitations of current approaches and extend their applicability to a wider range of useful applications.
Program
Discovery Grants Program - Individual
Recipient
Location
Alberta, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2023-04464
Name nom
Sander, Jörg
Department département
Computing Science
Organization id
9
Institution établissement
University of Alberta
Province en
Alberta
Province fr
Alberta
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2023
Award amount
36000
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
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
1200
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Advancement of knowledge
Area of application fr
Avancement des connaissances
Research subject code
2709
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Other computing methods
Research subject fr
Autres méthodes informatiques
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
Towards Efficient and Automated Unsupervised and Semi-Supervised Data Analytics
Keywords
clustering; outlier detection; density-based clustering; hierarchical clustering; cluster validation; outlier validation; scalable clustering
Application summary
The proposed research program will investigate and advance theory and practice of unsupervised and semi-supervised methods for finding and analyzing "structure" in multi-dimensional data. Many of today's data sets are multidimensional and are typically collected on a large scale, where none or only a small fraction of labelled data is available. These data sets hold the potential for discovering new insights that can lead to new opportunities for improved, data-driven decision making. Exploratory methods for unsupervised and semi-supervised clustering and outlier detection can play an important role in realizing this potential. They are often a first step in a data analytics pipeline that can establish insight into the data distribution, and can be used as a pre-processing step for other types of analyses and tasks. Understanding the (possibly hierarchical) clustering structure and outliers in the data allows for a better data organization (e.g., for more efficient browsing and searching), can lead to the discovery of unknown relationships between groups of data, or can help focus subsequent data analysis on specific subgroups or data points.
Applications of clustering and outlier detection often face significant challenges: (1) the data is often large in size (in number of records and number of attributes/dimensions), while (2) the methods typically do not scale well and (3) require significant involvement of an analyst, who has to choose parameter values for a method and a given data set, and frequently wishes to re-run analyses using different parameter values - which, however, is limited due to (1) and (2).
My proposed research program is trying to advance the theoretical foundations and applicability of clustering and outlier detection in unsupervised and semi-supervised settings - particularly in the density-based paradigm where clusters are considered dense regions in the data space, separated by regions of lower density. We will focus on the aspects of (A) Scalability, in order to deal with large data volumes and to enable fast computation of results for a wide range of parameter values of a method to lay the foundation for drawing more robust conclusions from multiple results and for automating model selection/construction; (B) Automated Model Selection/Construction, based on internal cluster and outlier validation measures, in both unsupervised and semi-supervised scenarios; (C) Modeling the development of cluster structures and outliers over time, in order to adapt results efficiently in dynamic application scenarios, discover and track structural changes in a single data set over time, and enable the discovery of how changes in different data sets can be related to each other.
Based on the gained theoretical insights, we will develop novel and improved algorithms for clustering and outlier detection that overcome some of the limitations of current approaches and extend their applicability to a wider range of useful applications.