Advantages And Disadvantages Of Feature Selection

1136 Words5 Pages
OPTIMIZED MULTI-LAYER PERCEPTRON BASED STROKE CLASSIFICATION Feature Selection is also termed as feature subset selection, variable selection, or attributes reduction (Roselin 2011) and aims to choose a subset of input variables by eliminating irrelevant features or of no predictive information. Larger number of features is time costly to classify and the efficiency of the classifier also reduces. Feature selection proves theoretically and practically to be effective in improving learning efficiency, reducing complexity of learned results and increasing predictive accuracy (Ramaswami & Bhaskaran 2009). Methodology: The pre-processing steps consist of median filter and histogram equalization. In this work, features are extracted using Watershed…show more content…
GAs work on an individual’s population, and represent candidate solutions to optimization problems. Individuals consist of gene strings (chromosomes). GA applies the principles of survival of the fittest, selection, reproduction, crossover (recombining), and mutation on individuals to ensure better individuals (new solutions). GA’s disadvantage is that it cannot find exact global optimum as there is no guarantee for a best solution. GAs use a number of parameters to control their evolutionary search for the solution to their given problems. Some of these include rate of crossover, rate of mutation, maximum number of generations, number of individuals in the population, and so forth. There are no hard and fast rules for choosing appropriate values for these parameters. An optimal or near-optimal set of control parameters for one GA or GA application does not generalize to all cases. Choosing values for the control parameters is often handled as a problem of trial and error. It is common practice to hand optimize the control parameters by tuning each one at a time. Proposed GA-ICA: ICA is a statistical method and GA is an optimization algorithm. The proposed GA-ICA consists of two stages: • First stage: Samples the large population of individuals uniformly in the solution space. Then ICA is applied on the population to find the independent components. • Second stage: Evolves the population to find the solution. Also it projects the population on the independent components and gets one 1-dimensional sub-population on each independent component; hence it is able to evolve on the independent components

More about Advantages And Disadvantages Of Feature Selection

Open Document