turicreate学习笔记2-clustering

1.生成sframe文件

import turicreate as tc

sf_functional = tc.SFrame.read_csv('train/train_FNC.csv')
sf_morphometry = tc.SFrame.read_csv('train/train_SBM.csv')

sf = sf_functional.join(sf_morphometry, on='Id')
sf = sf.remove_column('Id')

sf.save('schizophrenia_clean.sframe')

train_FNC.csv和train_SBM.csv在kaggle上可以下载

sframe file

2.create kmeans model

# create kmeans model
from math import sqrt

K = int(sqrt(sf.num_rows() / 2.0))

kmeans_model = tc.kmeans.create(sf, num_clusters=K)
print kmeans_model.summary()

kmeans model

3.使用

kmeans_model['cluster_info'].print_rows(
  num_columns=5,
  max_row_width=80,
  max_column_width=10
)

new_clusters = kmeans_model.predict(sf[:5])
print new_clusters