'scipy.sparse.csr.csr_matrix‘的散点图

Question

我有一个来自TfidfVectorizer的矩阵，那么能不能画出一个图来呢？

(0,6164) 0.019338613120625153

(0,8791) 0.030431754891299245

(0,13418) 0.019338613120625153

(0,12251) 0.040303890966260525

(0,2896) 0.017899943021264794

(0,2172) 0.12091167289878157

(0,3413) 0.040303890966260525

(0,6571) 0.014970123315314715

(0,13039) 0.015495789594635422

(0,11488) 0.03593540116095009

(0,12423) 0.030431754891299245

(0,11803) 0.017899943021264794

(0,14555) 0.017899943021264794

Answer 1

可以按如下方式使用plt.scatter：

import matplotlib.pyplot as plt
from scipy.sparse import csr_matrix
import numpy as np

# create a 15000x15000 sparse matrix with some toy data
N = 15000
npnts = 1000
# mtrx = csr_matrix(([y for (x1,x2),y in data], ([x1 for (x1,x2),y in data], [x2 for (x1,x2),y in data])), shape=(N, N))
mtrx = csr_matrix((np.random.uniform(0, 0.2, npnts),
                   (np.random.randint(0, N, npnts), np.random.randint(0, N, npnts))),
                  shape=(N, N))

# convert the dense matrix to dictionary format, get an array of xy-coordinates and an array of values
mtrx_dict = mtrx.todok()
xy = np.array(list(mtrx_dict.keys()))
vals = np.array(list(mtrx_dict.values()))

# create a scatter plot
plt.scatter(xy[:,0], xy[:,1], s=5, c=vals, cmap='inferno')
plt.colorbar()
plt.show()

​

​

根据数据的含义和您想要显示的内容，您可以尝试着色、alpha、大小等。