In the era of "data overload", many areas of society have begun to benefit from exploratory knowledge-mining techniques that are unsupervised, automated and scalable. The extraction processes are, however, often made difficult by a number of complexities: heterogeneous data types, missing information, clutter, prohibitive dimension-count and high bandwidths. In this thesis we focus on exploratory data-mining problems which exhibit these complexities. We present novel, linear-time methods to solve the problems, and empirically evaluate each against the state-of-the-art with respect to synthetically-generated data, real-world data and run-time behavior.
«
In the era of "data overload", many areas of society have begun to benefit from exploratory knowledge-mining techniques that are unsupervised, automated and scalable. The extraction processes are, however, often made difficult by a number of complexities: heterogeneous data types, missing information, clutter, prohibitive dimension-count and high bandwidths. In this thesis we focus on exploratory data-mining problems which exhibit these complexities. We present novel, linear-time methods to solv...
»