The last two decades there is a huge increase in the amount of data that is stored in digital format. Owing to today's technology data collection ceases to be a problem and in the focus of interest is their analysis and obtaining valuable information from the data (knowledge). Central for this issue is the process of knowledge discovery in data. The process consists of several steps and preparation of the data, which includes data cleaning and feature selection takes away from 60% till 95% total time of the whole process. The purpose of this research is to define new techniques for feature selection in order to reduce the time required to conduct feature selection, and thus removing the bottleneck of the entire process of knowledge discovery in data. For this purpose two new techniques of feature selection are defined, techniques that belong to the group of contrast mining field. The subject of this research is an application and evaluation of contrast mining techniques as a techniques for feature selection.The extensive empirical research is conducted in order to determine data sets characteristics for which contrast mining techniques outperform classical techniques of feature selection, and obtaining general answer can we, and in what kind of data sets, use contrast mining techniques as a superior feature selection techniques, and whether they can eliminate the bottleneck of the entire process of knowledge discovery in data. Results of 1792 analysis showed that in the more than 80% of the analyzed data sets with different characteristics contrast mining techniques resulted with more accurate classification and quickly implemented process of knowledge discovery than previously used feature selection techniques.