Clustering algorithms for rock porosity categorization

In this thesis we will discuss about about the result of a research project conducted for Politecnico di Torino about the classification of geological pores in rock samples. We analyzed the data collected by the experts about geological pores by using the most well-known clustering techniques of data mining. The aim is to understand which clustering algorithms give the best result in pores data categorization. In general, clustering studies sets of objects properties by finding relationships between these objects based on the similarities of their properties. In our case pores, have a great variety of attributes mainly based on their shape and size. An important aspect is that some of their attributes have values ​​distributed over very big range and this means that it is possible to create groups where locate pores characterized by similar attributes values. For example, in nature exist pores of very different sizes, so measuring their diameter, which can assume values in a big range of values, it possible to create distinct groups by using this feature. The thesis will be organized in parts where we will present the problem in more detail, discuss the used techniques from a mathematical and implementation point of view, then the most significant results we have obtained. This study can be thought as a starting point for classifying the pores through the clustering methods, moreover can be used as an approach to solve, in an alternative way, some of the several known problems that concern the analysis of geological pores.