No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

A Quantitative Structure-Activity Relationship (QSAR) study is an attempt to model some biological activity over a collection of chemical compounds in terms of their structural properties A QSAR model may be constructed through (typically linear) multivariate regression analysis of the biological activity data against a number of features or 'descriptors' of chemical structure. As with any regression model, there are a number of issues emerging in real applications, including (a) domain of applicability of the model, (b) validation of the model within its domain of applicability, and (c) possible non-linearity of the QSAR Unfortunately the existing methods commonly used in QSAR for overcoming these issues all suffer from problems such as computational inefficiency and poor treatment of non- linearity. In practice this often results in the omission of proper analysis of them altogether. In this thesis we develop methods for tackling the issues listed above using K-means clustering. Specifically, we model the shape of a dataset in terms of intelligent K-means clustering results and use this to develop a non- parametric estimate for the domain of applicability of a QSAR model. Next we propose a 'hybrid' variant of K-means, incorporating a regression-wise element, which engenders a technique for non-linear QSAR modelling. Finally we demonstrate how to partition a dataset into training and testing subsets, using the K-means clustering to ensure that the partitioning respects the overall distribution Our experiments involving real QSAR data confirm the effectiveness of the methods developed in the project.

The problem of combining multiple classifiers, referred to as a classifier ensemble, is one sub-domain of machine learning which has received significant attention in recent years. A classifier ensemble is an integration of classification models, referred to as base classifiers, whose individual decisions are combined in order to obtain a final prediction. The aim of using a classifier ensemble is to provide an overall level of performance which is superior to the performance of any of the single base classifiers. This Thesis studies the problem of constructing a classifier ensemble from different perspectives with the aim of improving the overall level of performance. Two novel ensemble techniques were introduced and evaluated within this study. The first approach, referred to as Classification by Cluster Analysis, was proposed as an alternative solution to the Stacking technique. The new method applies a clustering technique for the purpose of combining base classifier outputs. This approach offers benefits with reduced classification time compared with existing ensemble methods. In addition, it outperformed other ensemble methods in terms of classification accuracy. As an extension to the concept the method was adapted to incorporate semisupervised learning which is subsequently considered as a new research direclion within the domain of ensemble learning. The second method, referred to as Cluster-Based Classifier Ensemble, was proposed as an alternative to the Nearest Neighbour classifier ensemble. It applies a clustering technique for the purpose of generating base classifiers. A new combining function was proposed to be applied with the method as an alternative to the conventional majority voting technique. The new approach outperforms existing ensemble methods in terms of accuracy and efficiency. Both methods were evaluated in an activity recognition problem considered within the work as a case study. The effectiveness of the two methods was further supported by the findings from an experimental evaluation with a real world data set.