Experimentation Results

Abstract

In order to meet the requirements discussed in the previous chapters of this book, extensive evaluation of the results of the implementations detailed in chapter 6 were carried out. Initial experiments on the Galaxy Zoo Table 2 data set and on the initial 10 attributes obtained from the work of Baehr et al. [17] were eventually deemed to have not met these requirements. However, our IFS algorithm, developed at the start and deployed to the larger data sets acquired from the SDSS database during the later phases of the project, produced a significant increase on the original accuracies, deeming it a successful implementation and having successfully fulfilled the requirements. The re-labelling and iterative clustering of the galaxies originally labelled as Uncertain was also a remarkable success. These results were reinforced by further experimentation through the use of the Random Forest and Sequential Minimal Optimisation (SMO) algorithms.