#! /usr/bin/env python3
#
def draft_nn_normal ( ):

  from pandas import read_csv
  import matplotlib.pyplot as plt
  import numpy as np
  import pandas as pd

  print ( "draft_nn_normal():" )
  print ( "  Demonstrate nearest neighbor classification with normalized data." )

  filename = "draft.csv"
  df = pd.read_csv ( filename, header = 0 )
#
#  Print column headers.
#
  print ( df.columns )
#
#  Select Draft columns of yes, and of no.
#
  dfyes = df[df['Draft']=='yes']
  dfno  = df[df['Draft']=='no']
#
#  Save only the numeric data.
#
  dfall = df._get_numeric_data ( )
  dfyes = dfyes._get_numeric_data ( )
  dfno  = dfno._get_numeric_data ( )
#
#  Convert the numeric data to numpy arrays.
#
  npall = dfall.to_numpy ( )
  npyes = dfyes.to_numpy ( )
  npno  = dfno.to_numpy ( )
#
#  Get min/max of speed and agility.
#
  xmin = np.min ( npall[:,1] )
  xmax = np.max ( npall[:,1] )
  ymin = np.min ( npall[:,2] )
  ymax = np.max ( npall[:,2] )
  print ( xmin, "<= speed <=", xmax )
  print ( ymin, "<= agility <= ", ymax )
#
#  Introduce a new data item.
#
  npmaybe = np.array ( [ 21, 6.75, 3.00 ] )
#
#  Normalize npall and npmaybe to equalize distances.
#
  npall[:,1] = ( npall[:,1] - xmin ) / ( xmax - xmin )
  npall[:,2] = ( npall[:,2] - ymin ) / ( ymax - ymin )
  npyes[:,1] = ( npyes[:,1] - xmin ) / ( xmax - xmin )
  npyes[:,2] = ( npyes[:,2] - ymin ) / ( ymax - ymin )
  npno[:,1]  = ( npno[:,1] - xmin ) / ( xmax - xmin )
  npno[:,2]  = ( npno[:,2] - ymin ) / ( ymax - ymin )
  npmaybe[1] = ( npmaybe[1] - xmin ) / ( xmax - xmin )
  npmaybe[2] = ( npmaybe[2] - ymin ) / ( ymax - ymin )
#
#  Display the results.
#
  plt.plot ( npyes[:,1], npyes[:,2], 'b^', markersize = 10 )
  plt.plot ( npno[:,1], npno[:,2], 'kv', markersize = 10 )
  plt.grid ( True )
  plt.xlabel ( "Speed" )
  plt.ylabel ( "Agility" )
  plt.title ( "Athletic ratings for drafting" )
  plt.legend ( ( "Drafted", "Rejected" ), loc = "upper left" )
  filename = "draft_results_normal.png"
  plt.savefig ( filename )
  plt.show ( )
  plt.close ( )
#
#  Display data with new item.
#
  plt.plot ( npyes[:,1], npyes[:,2], 'b^', markersize = 10 )
  plt.plot ( npno[:,1], npno[:,2], 'kv', markersize = 10 )
  plt.plot ( npmaybe[1], npmaybe[2], 'r*', markersize = 10 )
  plt.grid ( True )
  plt.xlabel ( "Speed" )
  plt.ylabel ( "Agility" )
  plt.title ( "Athletic ratings for drafting" )
  plt.legend ( ( "Drafted", "Rejected", "New Candidate" ), loc = "upper left" )
  filename = "draft_candidate_normal.png"
  plt.savefig ( filename )
  plt.show ( )
  plt.close ( )
#
#  Compute distances.
#
  dist = np.zeros ( 20 )
  for i in range ( 0, 20 ):
    dist[i] = np.sqrt ( ( npall[i,1]-npmaybe[1] )**2 \
                      + ( npall[i,2]-npmaybe[2] )**2 )

  print ( '' )
  print ( 'index  distance  draft/nodraft:' )
  print ( '' )

  index = np.argsort ( dist )
  distances = ( index, dist[index], df['Draft'][index] )
  print ( np.transpose ( distances ) )

  print ( "" )
  print ( "  Draft/no draft prediction:", df['Draft'][index[0]] )

  print ( "  A 3-nearest neighbor prediction would be YES" )

if ( __name__ == "__main__" ):
  draft_nn_normal ( )