An Approach for Detecting Noise in Cotton Fiber Properties Data Using Nearest Neighbor Algorithm

The purpose of the current study is to decrease noise of cotton fiber properties data using one of methods for machine learning such as K Nearest Neighbor (KNN). The present investigation was carried out at Egyptian& International Cotton Classification Center (EICCC). The first data was for Giza 87, Giza 88, Giza 86, Giza 90 and Giza 95 each one separately and the second data was for combination of all previous cotton varieties. A wide range of lint cotton grades used in this work. The studied traits were basic fiber properties; length, strength and micronaire value. The highest classification accuracy were 117.65 % for G 87 and 149.25 % for combined data. The integrated statistics among fiber length, strength and micronaire value concluded that spinning consistency value (SCI) which is the most intrinsic technological value were in acceptable range for Giza 87, Giza 88, Giza 86, Giza 90, Giza 95 separately and combined data. For instance, SCI values were 163.74 and 174.85 for G 87 of data treated without KNN and with KNN, respectively. Therefore, any study of cotton fiber properties plays a crucial role in determining spinning performance