Article Preview
Top1. Introduction
Object recognition is one of research areas that aims to classify objects. Usually recognition of objects such as building, tree, mountains, roads, and rivers is a very complex task in terms of time and cost. Traditional recognizing methods, such as hand-digitizing, are slow, tedious and the availability of numerous spectral, spatial, and texture features renders the selection of optimal features a time consuming.
The goal of the Feature Selection (FS) is to detect irrelevant and/or redundant features as they harm the learning algorithm performance (Lee & Moore, 2014). A FS algorithm can effectively remove irrelevant and redundant features and take into account feature correlation. This not only leads up to an insight understanding of the data, but also improves the performance of a learner by enhancing the generalization capacity and the interpretability of the learning model (Wang et al., 2014). In other words, no new feature is created, the features that are considered irrelevant or redundant are discarded, and we ideally would end up with the best possible feature subset, that is, the subset with minimum size and which leads to the minimum classification error rate. Feature selection with subset evaluation requires defining how to search the space of feature subsets (search method) and what measure to use when evaluating a feature subset (evaluation criterion) as well as the initial feature set and a termination condition.
Feature Selection methods fall into two broad categories: Wrapper and Filter (Kohavi & John, 1997; Molina, Belanche, & Nebot, 2002). The Wrapper approach uses the accuracy of the classification algorithm as the evaluation function to measure a feature subset as you shown in Figure 1, while the evaluation function of the Filter approach is independent of the classification algorithm. The accuracy of the Wrapper approach is usually high; however, the generality of the result is limited, and the computational complexity is high. In comparison, Filter approach is of generality, and the computational complexity is low. Because the Wrapper approach is computationally expensive (Vafaie & De Jong, 1992), the Filter approach is usually a good choice when the number of features is very large. Thus, we focus on the Wrapper method in our experiment, because we have only 38 features.
Figure 1. Feature selection based on wrapper method (Kohavi & John, 1997)
We use Genetic Algorithm (GA) as an optimization algorithm. Genetic algorithms comprise a search algorithm that guides its search based on a model of evolution. Evolution is the process by which chromosome continually improve over generations, through selection, crossover and mutation. In GA, An initial population is created containing a predefined size (number of chromosomes), each represented by a genetic string. Each chromosome has an associated fitness value, typically representing an accuracy value. The concept that fittest (or best) individuals in a population will produce fitter offspring to be used in the next produced population.
As previously mentioned, we focused on the Wrapper method in our experiment, the wrapper approach was applied using as black box using three classifiers, Artificial Neural Network (ANN), K-Nearest Neighbors (KNN) and J48 Decision tree within optimize search algorithm (Genetic Algorithm).
This paper is organizing as follows. Section 2 present some related works. Section 3 includes the methodology and proposed model. In Section 4, we present and analyze our experimental results. Section 5 will draw the conclusion and summarize the research achievement and future directions.