The file P17_15.xlsx contains data on 178 wines. They are categorized into three types, labeled A, B, and C. The rest of the variables are numeric properties of the wines. Use Neural-Tools to classify these wines. Use the PNN algorithm, and check the Variable Impact Analysis option in the Train dialog box. This ranks the variables on their impact, which provides some information on which variables might not be needed for the neural net. Then run the algorithm a second time, using only the top five variables in terms of their impact percentages. (In the Data Set Manager, mark the others Unused.) Comment on the results. Is the prediction accuracy much different without the deleted variables?