Effect of Normalisation for Gender Identification

This work has considered standard data set for gender identification using machine learning approach. Various approaches are considered to achieve greater accuracy and low error (specificity/MCC). Number of experiments are done to learn optimal weights for different value of K = 5, 10, 15 and 20. From range of experiment, it is clear that higher K fold validation is giving better accuracy all time. When different K folds cross-validation is used, different techniques and different normalisation models are performing in different way. When K = 5 and 15, without normalisation and forward normalisation are performing best, and for K = 10, forward normalisation, and for K = 20 standard normalisation is working best.