这个图看起来一团糟，但是每一部分都代表不同的含义：
1，这个图形结合了本项目的股票数据，GraphLassoCV图结构模型，近邻传播算法的分类结果，故而可以说是整个项目的结晶。
2，图中每一个节点代表一只股票，旁边有股票名称，节点的颜色表示该股票所属类别的种类，用节点颜色来区分股票所属簇群。
3，GraphLassoCV图结构模型中的稀疏逆协方差信息用节点之间的线条来表示，线条越粗，表示股票之间的关联性越强。
4，股票在图形中的位置是由2D嵌套算法来决定的，距离越远，表示其相关性越弱，簇间距离越远。
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline
np.random.seed(37) # 使得每次运行得到的随机数都一样
# 准备数据集，使用tushare来获取股票数据
# 准备几个函数，用来获取数据
import tushare as ts
def get_K_dataframe(code,start,end):
'''get day-K data of code, from start date to end date
params:
code: stock code eg: 600123, 002743
start: start date, eg: 2016-10-01
end: end date, eg: 2016-10-31
return:
dataframe with columns [date, open, close, high, low]
'''
df=ts.get_k_data(code,start=start,end=end)
df.drop(['volume'],axis=1, inplace=True)
return df
def get_batch_K_df(codes_list,start,end):
'''get batch stock K data'''
df=pd.DataFrame()
print('fetching data. pls wait...')
for code in codes_list:
# print('fetching K data of {}...'.format(code))
df=df.append(get_K_dataframe(code,start,end))
return df
# 数据规整函数，用于对获取的df进行数据处理
def preprocess_data(stock_df,min_K_num=1000):
'''preprocess the stock data.
Notice: min_K_num: the minimum stock K number.
because some stocks was halt trading in this time period,
the some K data was missing.
if the K data number is less than min_K_num, the stock is discarded.'''
df=stock_df.copy()
df['diff']=df.close-df.open # 此处用收盘价与开盘价的差值做分析
df.drop(['open','close','high','low'],axis=1,inplace=True)
result_df=None
#下面一部分是将不同的股票diff数据整合为不同的列，列名为股票代码
for name, group in df[['date','diff']].groupby(df.code):
if len(group.index)<min_K_num: continue
if result_df is None:
result_df=group.rename(columns={'diff':name})
else:
result_df=pd.merge(result_df,
group.rename(columns={'diff':name}),
on='date',how='inner') # 一定要inner，要不然会有很多日期由于股票停牌没数据
result_df.drop(['date'],axis=1,inplace=True)
# 然后将股票数据DataFrame转变为np.ndarray
stock_dataset=np.array(result_df).astype(np.float64)
# 数据归一化，此处使用相关性而不是协方差的原因是在结构恢复时更高效
stock_dataset/=np.std(stock_dataset,axis=0)
return stock_dataset,result_df.columns.tolist()
# 上面准备了各种函数，下面开始准备数据集
# 我们此处分析上证50指数的成分股，看看这些股票有哪些特性
sz50_df=ts.get_sz50s()
stock_list=sz50_df.code.tolist()
# print(stock_list) # 没有问题
batch_K_data=get_batch_K_df(stock_list,start='2013-09-01',end='2018-09-01') # 查看最近五年的数据
print(batch_K_data.info())
stock_dataset,selected_stocks=preprocess_data(batch_K_data,min_K_num=1100)
print(stock_dataset.shape) # (603, 41) 由此可以看出得到了603个交易日的数据，其中有41只股票被选出。
# 其他的9只股票因为不满足最小交易日的要求而被删除。这603个交易日是所有41只股票都在交易，都没有停牌的数据。
print(selected_stocks) # 这是实际使用的股票列表
# 从相关性中学习其图形结构
from sklearn.covariance import GraphLassoCV
edge_model=GraphLassoCV()
edge_model.fit(stock_dataset)
# 使用近邻传播算法构建模型，并训练LassoCV graph
from sklearn.cluster import affinity_propagation
_,labels=affinity_propagation(edge_model.covariance_)
n_labels=max(labels)
# 对这41只股票进行了聚类，labels里面是每只股票对应的类别标号
print('Stock Clusters: {}'.format(n_labels+1)) # 10，即得到10个类别
sz50_df2=sz50_df.set_index('code')
# print(sz50_df2)
for i in range(n_labels+1):
# print('Cluster: {}----> stocks: {}'.format(i,','.join(np.array(selected_stocks)[labels==i]))) # 这个只有股票代码而不是股票名称
# 下面打印出股票名称，便于观察
stocks=np.array(selected_stocks)[labels==i].tolist()
names=sz50_df2.loc[stocks,:].name.tolist()
print('Cluster: {}----> stocks: {}'.format(i,','.join(names)))
from sklearn import manifold
from matplotlib.collections import LineCollection
# 对这些结果进行可视化
def visual_stock_relationship(dataset,edge_model,labels,stock_names):
node_position_model = manifold.LocallyLinearEmbedding(
n_components=2, eigen_solver='dense', n_neighbors=6)
embedding = node_position_model.fit_transform(dataset.T).T
plt.figure(1, facecolor='w', figsize=(10, 8))
plt.clf()
ax = plt.axes([0., 0., 1., 1.])
plt.axis('off')
# Display a graph of the partial correlations
partial_correlations = edge_model.precision_.copy()
d = 1 / np.sqrt(np.diag(partial_correlations))
partial_correlations *= d
partial_correlations *= d[:, np.newaxis]
non_zero = (np.abs(np.triu(partial_correlations, k=1)) > 0.02)
# Plot the nodes using the coordinates of our embedding
plt.scatter(embedding[0], embedding[1], s=100 * d ** 2, c=labels,
cmap=plt.cm.nipy_spectral)
# Plot the edges
start_idx, end_idx = np.where(non_zero)
# a sequence of (*line0*, *line1*, *line2*), where::
# linen = (x0, y0), (x1, y1), ... (xm, ym)
segments = [[embedding[:, start], embedding[:, stop]]
for start, stop in zip(start_idx, end_idx)]
values = np.abs(partial_correlations[non_zero])
lc = LineCollection(segments,
zorder=0, cmap=plt.cm.hot_r,
norm=plt.Normalize(0, .7 * values.max()))
lc.set_array(values)
lc.set_linewidths(15 * values)
ax.add_collection(lc)
# Add a label to each node. The challenge here is that we want to
# position the labels to avoid overlap with other labels
n_labels=max(labels)
for index, (name, label, (x, y)) in enumerate(
zip(stock_names, labels, embedding.T)):
dx = x - embedding[0]
dx[index] = 1
dy = y - embedding[1]
dy[index] = 1
this_dx = dx[np.argmin(np.abs(dy))]
this_dy = dy[np.argmin(np.abs(dx))]
if this_dx > 0:
horizontalalignment = 'left'
x = x + .001
else:
horizontalalignment = 'right'
x = x - .001
if this_dy > 0:
verticalalignment = 'bottom'
y = y + .001
else:
verticalalignment = 'top'
y = y - .001
plt.text(x, y, name, size=10,fontproperties = 'SimHei',
horizontalalignment=horizontalalignment,
verticalalignment=verticalalignment,
bbox=dict(facecolor='w',
edgecolor=plt.cm.nipy_spectral(label / float(n_labels)),
alpha=.6))
plt.xlim(embedding[0].min() - .15 * embedding[0].ptp(),
embedding[0].max() + .10 * embedding[0].ptp(),)
plt.ylim(embedding[1].min() - .03 * embedding[1].ptp(),
embedding[1].max() + .03 * embedding[1].ptp())
plt.show()
stock_names=sz50_df2.loc[selected_stocks,:].name.tolist()
visual_stock_relationship(stock_dataset,edge_model,labels,stock_names)the end of demo
