Decision tree, data mining

St/pen/7sed Data Mining. Searching large volumes of data for hidden predictive relationships. Supervised analysis requires one or more "dependent" or response variables, to be predicted from a set of "independent" or predictor variables. The techniques used include various classification methods (decision tree, support vector, Bayesian) and various estimation methods (regression, neural nets). 

Data mining is a new methodology for improving the quality and effectiveness of business and scientific decision-making processes.[76] There are currently several data mining techniques available. The decision-tree method is one of the most important techniques.[77] The attribute-splitting criteria of C5.0 decision tree is entropy, which is used to describe uncertainty of a stochastic experiment. X = pi(l), P2(2),..., pn n) is used to describe a stochastic experiment, in which p, denotes the probability of one experiment whose result is i. So the entropy function can be expressed as Equation (7.5) ... 

Decision trees are one of the most versatile tools in data mining. They have been employed for the extraction of patterns in large structure-activity [53-55] or structure-property data sets [56], and for the development of classification models [56,57]. 

A decision tree can be used for the classification of the direct impact of vulnerabibties as Krsul (1998) proved. Such an approach can be used also for data mining in vulnerabbity databases, according to (Schumacher et al. 2000). Software vulnerabbities can be analyzed not only from software engineering point of view, but also from economical perspective as Ozmet (2007) shown. 

The data mining techniques we employed include decision tree, neural networks, and nearest neighbour methods. 

The Decision tree method is widely used for classification and regression. A decision tree is a flow-chart-like tree structure, where each internal node denotes a test on an attribute, each branch represents an outcome of the test, and leaf nodes represent classes or class distributions. In order to classify an unknown sample, the attribute values of the sample are tested according to the decision tree starting from the root until one of the leaves. To build decision trees, a data mining algorithm recursively inspects the available data set to find decisions that optimally split the data into distinguished subsets. An important property of this technique is that its functioning is easily understood. 

For the PCA and PLS-DA, sparse analyses perform a selection from automatic variables. More recently, more complex methods of automatic learning from data mining have been applied to metabolomic data. Decision trees aid the automatic selection of discriminant variables, supply a simple representation of the decision model (the tree) and constitute an exploratory technique to understand complex metabolic profiles. The artificial neuron network was successfully used to classify chemical profiles and is becoming one of the most popular methods for understanding patterns. Data visualization and interactivity are now used to visualize metabolomic data in order to facilitate the interpretation of complex data-sets. XCMS online [GOW 14] offers cloud-plots, PCA and interactive heatmaps (i.e. the heatmaps are graphical representations of correlation matrices). These two types of visualization help the user personalize the display and easily select the most interesting compounds. 

---