Survey on Feature Selection for Data Reduction

Abstract

The storage capabilities and advanced in data collection has led to an information load and the size of databases increases in dimensions, not only in rows but also in columns. Data reduction (DR) plays a vital role as a data prepossessing techniques in the area of knowledge discovery from the huge collection of data. Feature selection (FS) is one of the well known data reduction techniques, which deals with the reduction of attributes from the original data without affecting the main information content. Based on the training data used for different applications of knowledge discovery, FS technique falls into supervised, unsupervised. In this paper an extensive survey on supervised FS technique describing the different searching approach, methods and application areas with an outline of a comparative study is covered.

A comparative study on dimensionality reduction between principal component analysis and k-means clustering

“…757 may act as a way to reduce the dimensionality and ease the computation of KNN model but it does not influence the performance of the model. This result is supported by previous studies where they claimed there is a high tendency that the complexity of the computation is being reduced without affecting the performance of a classification model [23][24]. Hence, this study found that the accuracy of KNN model remains the same with the application of feature selection towards normally distributed data set.…”

Time-Series Identification on Fish Feeding Behaviour

“…Nevertheless, it is clear from this work that the use of PCA with varimax rotation factor loads contributed to a higher identification performance than the traditional PCA methodology. Even though the classification reliability of PCA varimax and the choice of all features are shown to be poor, it is important to note that around half of the information is decreased and that the processing time used for non-trivial real-time classification decreases dramatically [17][18][19]. The k-NN method gives, ultimately, 3.5% error of the system tested by choosing features using PCA with a varimax rotation, the highest classification level.…”

Variable selection for prediction of time series from smart city

“…Despite the fact that the feature subset selection methods are well known for more than four decades, a huge number of new feature selection surveys was published during last few years ( [5], [6], [7], [8], [9], [10], [11], [12]). This supports the well known and still increasing importance of a good feature selection for the design of prediction and classification systems.…”

“…This supports the well known and still increasing importance of a good feature selection for the design of prediction and classification systems. For wrapper approaches, the reviews mostly adopt the following claims: (1) wrappers often reach better results than filters, because their feature selection criteria are more natural representatives of the prediction quality and are tailored to the 978-1-4673-6727-1/15$31.00 c 2015 IEEE particular predictor [5], [11] 1 , (2) wrappers are often much slower than filters, because of their time consuming feature selection criteria [6], [5], (3) wrappers exhibit higher risk of over-fitting [11]. This paper proves significant benefits of two particular wrapper approaches in terms of high dimensionality reduction and for most datasets also negligible performance degradation.…”