Chess

 # coding=utf-8
 import pandas as pd
 import numpy as np
 from sklearn import cross_validation
 import tensorflow as tf

 global flag
 flag=0

 def DataPreprocessing():
     abalone = pd.read_csv("train_data.csv", sep=',', header=0, keep_default_na=True)
     X_train=np.array(abalone.iloc[:,:6])
     Y_train_=np.array(abalone.iloc[:,6:])
     print(X_train)
     Y_train=[]
     for i in range(len(X_train)):

         X_train[i][0] = ord(X_train[i][0])-97
         X_train[i][2] = ord(X_train[i][2])-97
         X_train[i][4] = ord(X_train[i][4])-97

     # for i in range (len(X_train)):
     #     for j in range(6):
     #         X_train[i][j]=X_train[i][j]-0.0
     #
     #X_train.astype(np.float64)
   #  print(X_train,type(X_train),X_train[0][0],type(X_train[0][0]))

     #binary classifier
     for i in range(len(Y_train_)):

         if Y_train_[i][0]=="draw":
             Y_train.append(0)
         else:
             Y_train.append(1)

     # multiple classifer

     return cross_validation.train_test_split(X_train,Y_train,test_size=0.25,random_state=0,stratify=Y_train)

 def GetInputs():
     global flag
     X_train, X_test, Y_train, Y_test = DataPreprocessing()

     #print(type(X_train),type(X_train[0][0]))
     #print(X_train)
     # print(len(X_test))
     # print(len(Y_train))
     # print(len(Y_test))

     #X_train[X_train.isnull().any(axis=1)]
     #X_train.fillna('',inplace=True)

     # print(X_train)
     # print(Y_test)

     x_train=tf.constant(X_train)
     y_train=tf.constant(Y_train)
     x_test=tf.constant(X_test)
     y_test=tf.constant(Y_test)
     #
     # print(x_train)
     # print(y_train)
     # print(x_test)
     # print(y_test)

     if flag==0:
         return x_train,y_train
     else:
         return x_test,y_test

 def Main():

     global flag

     feature_columns=[tf.contrib.layers.real_valued_column("",dimension=6)]

     clf=tf.contrib.learn.DNNClassifier(feature_columns=feature_columns,hidden_units=[20,40,20],n_classes=2,model_dir="/home/jiangjing/TensorflowModel/chess")

     clf.fit(input_fn=GetInputs,steps=2000)

     flag=1
     accuracy_score=clf.evaluate(input_fn=GetInputs,steps=1)["accuracy"]

     print("nTest Accuracy:{0:f}".format(accuracy_score))

 if __name__ =="__main__":
     #DataPreprocessing()

     Main()

 exit(0)