kaggle上泰坦尼克的年龄使用tensorflow进行随机森林回归

kaggle上泰坦尼克的年龄转换为int型&＃xff0c;0~80岁一共81个标签。使用随机森林分类准确率为0.47&＃xff0c;然后再用这个模型去获得未知的年龄数据。

import pandas as pd
import numpy as np
import tensorflow as tf from tensorflow.contrib.tensor_forest.python import tensor_forest
from tensorflow.python.ops import resources
path&＃61;r"E:\learn\pc_code\kaggle\Titanic\all"
data_train &＃61; pd.read_csv(path&＃43;&＃39;/&＃39;&＃43;&＃39;train.csv&＃39;)
tf.reset_default_graph()
df&＃61;data_train
age_df&＃61;df[[&＃39;Age&＃39;,&＃39;Fare&＃39;,&＃39;Parch&＃39;,&＃39;SibSp&＃39;,&＃39;Pclass&＃39;]]
known_age &＃61; age_df[age_df.Age.notnull()].values #知道的做训练集
unknown_age &＃61; age_df[age_df.Age.isnull()].values
test_x&＃61;known_age[:,1:]
test_y&＃61;known_age[:,0]#param
num_steps&＃61;1000
batch_size&＃61;100
num_classes&＃61;81
num_features&＃61;4
num_trees&＃61;10
max_nodes&＃61;100000
X&＃61;tf.placeholder(tf.float32,shape&＃61;[None,num_features])
Y&＃61;tf.placeholder(tf.float32,shape&＃61;[None])
hparams&＃61;tensor_forest.ForestHParams(num_classes&＃61;num_classes,num_features&＃61;num_features,num_trees&＃61;num_trees,max_nodes&＃61;max_nodes).fill()forest_graph&＃61;tensor_forest.RandomForestGraphs(params&＃61;hparams)
train_op&＃61;forest_graph.training_graph(X,Y)
loss_op&＃61;forest_graph.training_loss(X,Y)infer_op, _, _&＃61;forest_graph.inference_graph(X)
correct_pre&＃61;tf.equal(tf.argmax(infer_op,1),tf.cast(Y,tf.int64))
accuracy_op &＃61; tf.reduce_mean(tf.cast(correct_pre,tf.float32))
init_vars&＃61;tf.group(tf.global_variables_initializer(),resources.initialize_resources(resources.shared_resources()))
sess&＃61;tf.Session()
sess.run(init_vars)
for i in range(num_steps):permutation &＃61; np.random.permutation(test_y.shape[0])batch_x&＃61;test_x[permutation,:][0:batch_size]batch_y&＃61;test_y[permutation][0:batch_size]_, l&＃61;sess.run([train_op,loss_op],feed_dict&＃61;{X:batch_x,Y:batch_y})if i%50&＃61;&＃61;0 or i&＃61;&＃61;1:acc&＃61;sess.run(accuracy_op,feed_dict&＃61;{X:test_x,Y:test_y})print(&＃39;Step: %i Loss: %f Accuracy: %f&＃39; %(i,l,acc))