Artificial Neural Network (ANN) algorithms have been widely applied to genomic prediction data. Single Nucleotide Polymorphisms(SNPs) represent the most common genetic variations of human genome, it has been shown that they are implicated in many human diseases and they can be used to predict genetic diseases such as diabetes. Developing ANN techniques to handle such kind of data can be considered as a great success in medical field. However, the high dimensionality of genomic data and availability of only a few number of samples can make learning task very complicated. In this work, we propose a New Neural Network classification method based on input perturbation. The idea is first to use SVD to reduce the dimensionality of input data and train a classification network which prediction errors are then reduced by perturbing the SVD projection matrix. The proposed method has been evaluated on individuals genetic ancestry prediction dataset, the experimental results show the efficiency of proposed method. Reaching up to 96.23% of classification accuracy, this approach exceeds the previous deep learning approach evaluated on the same dataset.