Machine data that are generated will be in the form of heterogeneous. It can be in the combination of both human-readable and unreadable format where it contains the huge number of features or variable values along with that analyzing all these features which increases the execution time, as well as the accuracy level of the discovering hidden patterns, will also be in very low level (Ibrahim, 2010). To overcome the above-said issue the several types of research are been on-going on identifying or extracting the relevant features from the given set of features which increases the accuracy level that can be increased. This section is focusing on to have state-of-art about various feature selection methods which are been currently used along with the mining algorithms. Table 1 will provide the details about the list of feature selection methods along with the supervised and unsupervised algorithms which are proposed and applied to the network dataset. Most of the above-said methods were tried to apply their classification process in a multiple-level decision fusion approach. Also, the proposed system has been attempted to focus on designing the solution for a multiclass problem using a feature weighting method (Jyoothsna et al., 2011).