Internatonal Journal of Informaton, Security and Systems Management, 2016,Vol.5,No.2, pp. 583-590 IJISSM ABSTRACT In this paper, principles and exising feature selection methods for classifying and clusering data be introduced. To that end, categorizing frameworks for fnding select- ed subsets, namely, search-based and non-search based procedures as well as evaluation criteria and data mining tasks are discussed.In the following, a platform is devel- oped as an intermediate sep toward developing an intelli- gent feature selection sysem, involving crucial, decisive and efective factors in feature selection process. The pro- cedure increases accuracy in classifcation and goodness of clusers. Finally, some of the problems and challenges facing the current and future feature selection processing are also discussed. Keywords Feature selection, classifcation, clusering,categorizing framework, evaluation criteria. 1. Introduction Despite recent advances in data processing technology, the ever-increasing volume of datasets and the number of features, which often increase wase of data, makes it difcult to supply the needed resources, including sorage and processing of data. At the same time, the explosive growth of data has led to increased noise. The misleading and wase data among the mass of useful data, exchanged in social networks, are the manifesations of noise in the data. Curse of dimensionality is the mos important out- come of increasing data dimensions. Additionally, despite the large number of features, learning models are prone to overftting and performance degradation [1]. Several srategies have been proposed in the literature to deal with such consequences. For example, qualitative and targeted data reduction can help to solve the problem in a more limited scale, without removal of useful or meaningful data. Some srategies proposed for data reduction are dimension reduction, data reduction, and data compres- sion[2]. In this article, the authors focus on dimension re- duction techniques, which are among common techniques for reducing the number of features,. 1.1 Feature Extraction In this method, some prominent features are produced through one or more conversions on input features. While mapping points from one space with higher dimensions into another space with lower dimensions, a large number of points may overlap. Feature extraction helps to fnd a new dimension where a minimum number of points may overlap. This approach is associated with the problem area and is commonly used in image processing where specifc features are extracted in accordance with the re- quirements of the problem. 1.2 Feature Selection Proposed in various felds of machine learning and data mining, feature selection is one of the subsidiaries of fea- ture extraction. It is preferable in contexts where read- ability and interpretability are issues of concern, because the discounted values of the main features are preserved in the reduced space [1]. This method of dimension re- duction results in a qualitative database, without remov- al of useful information. It also allows for the features with diferent data models to be combined. The issue is of importance because a large number of features are of- ten used in diferent applications. Therefore, the need to select a limited set from among them becomes apparent. Consraints and considerations such as avoiding the curse of dimensionality, memory limits, reduction of the needed computations, and reduction of the runtime, among others oblige one to select the minimum number of features to be used in prediction of future data. The following general objectives and considerations should be taken into account considering feature selec- tion as well as and the characterisics that the fnal subset should have: 1. Idealization: a minimally sized feature subset that is necessary and sufcient to the target concept should be found [3]. 2. Optimalityof the fnal subset: the amount obtained by Hamid Parvin yasooj branch, islamicazad university Ali AsgharNadri yasooj branch, islamicazad university Farhad Rad yasooj branch, islamicazad university Optimal Feature Selection for Data Classifcation and Clusering: Techniques and Guidelines