Data Mining: Machine Learning and Statistical Techniques

The interdisciplinary field of Data Mining (DM) arises from the confluence of statistics and machine learning (artificial intelligence). It provides a technology that helps to analyze and understand the information contained in a database, and it has been used in a large number of fields or applications. Specifically, the concept DM derives from the similarity between the search for valuable information in databases and mining valuable minerals in a mountain. The idea is that the raw material is the data to analyse, and we use a set of learning algorithms acting as diggers to search for valuable nuggets of information (Bigus, 1996). We offer an applied vision of DM techniques, in order to provide a didactic perspective of the data analysis process of these techniques. We analyze and compare the results from applying machine learning algorithms and statistical techniques, under DM methodology, in searching for knowledge models that show the structures and regularities underlying the data analysed. In this sense, some authors have pointed out that DM consists of “the analysis of (often large) observational datasets to find unsuspected relationships and to summarize the data in novel ways that are both understandable and useful to the data owner” (Hand, Mannila & Smyth, 2001), or, more simply, “the search for valuable information in large volumes of data” (Weiss & Indurkhya, 1998), or “the discovery of interesting, unexpected or valuable structures in large databases” (Hand, 2007). Other authors define DM as “the exploration and analysis of large quantities of data in order to discover meaningful patterns and rules” (Berry & Linoff, 2004). These definitions make it clear that DM is an appropriate process for detecting relationships and patterns in large databases (although we point out that it can also be applied in relatively small databases). In this sense, the concept of Knowledge Discovery in Databases (KDD) has been frequently used in the literature to define this process (Han & Kamber, 2000, 2006; Hand et al., 2001), specifying that DM is a stage of the process, and highlighting the need for a previous stage of integration and collection of data (if we start with large raw databases), and also the stage of cleaning and preparing data (data pre-processing) before building descriptive/predictive models in the DM stage (applying suitable techniques to the analysis requirements). On the other hand, several authors have used the concept of DM (instead of KDD) to refer to the complete process (Bigus, 1996; Two Crows, 1999; Paul, Guatam & Balint, 2002; Kantardzic, 2003; Ye, 2003; Larose, 2005).