Features Selection Algorithms for Classification of Voice Signals

In data mining problems, the high dimensionality of input features can affect performance process. this way, selection methods appear as a solution to problems encountered when analyzing databases with large dimensions. This article presents implementation Pearson’s linear correlation, ReliefF, Welch’s t-test and multilinear regression based algorithms forwards backward elimination direction for acoustic task voice pathologies identification. The best set selected improved accuracy F1-score from 83% 92% (9 points percentage), using ReliefF algorithm.