Feature selection techniques for disease diagnosis system: A survey

Abstract

Reducing dimensionality as a preprocessing phase towards artificial intelligence effectively eliminates unnecessary and repetitive information, raises the efficiency of learning accuracy, and improving output understandability. Nonetheless, the rapid surge in data dimension presents a real challenge to several current collections of features and extraction methods as regards efficiency and effectiveness. Many researchers build and use plenty of feature selection algorithms. But an emerging field of machine learning is still to be based on data mining and method of analysis for information processing. Due to the recent increase in data variation and speed, many feature selection algorithms face serious efficiency and performance problems. Different kinds of feature selection algorithms are available in research such as algorithms based on filters, algorithms based on wrappers, and algorithms based on hybrids. Additionally, a literature study analyzes some of the existing popular feature selection algorithms and also addresses the strong points and difficulties of those algorithms.