Pattern visualization in heatmaps

With the need of more and more data analysis in today’s world, the patterns visualisation algorithms take a more important place in our society. This approach of data analyzis focuses on reordering rows and columns of datasets in double entry tables or matrices in order to highlight patterns that may expose informations on the subject the data refer. The informations exposed could come from either the ordering of the rows and columns that brings the patterns together or from the patterns themselves. This thesis aims at improving one of those pattern visualisation algorithm known as Iterative Matrix Ordering. This algorithm is a local search algorithm that was made with purpose of ordering greater and noisier datasets that what was possible at its time. The goal of this thesis will then to go even farther and allow this marvelous tool to work on even greater, noisier or more sparse datasets. Indeed, the original algorithm would encounter difficulties when trying to order the huge dataset we can find nowadays. This thesis tries to allow datasets too big for the algorithm memory to be ordered by the algorithm by preprocessing the dataset or other tricks. It also tries to bring as many improvements possible to the functioning of the algorithm and analyze some of its results on dataset with specific layouts.