pyspark:稀疏向量到稀疏矩阵

问题描述

我有一个spark数据框，其中有一列短句子，以及一列带有分类变量.我想对句子执行tf-idf，对分类变量执行one-hot-encoding，然后将其输出到驱动程序上的稀疏矩阵中，一旦它的大小变得小得多(对于scikit-learn模型).

I have a spark dataframe with a column of short sentences, and a column with a categorical variable. I'd like to perform tf-idf on the sentences, one-hot-encoding on the categorical variable and then output it to a sparse matrix on my driver once it's much smaller in size (for a scikit-learn model).

以稀疏形式获取数据的最佳方法是什么?稀疏向量上似乎只有一个toArray()方法，可以输出numpy数组.但是，文档确实说scipy稀疏数组可以代替火花稀疏数组.

What is the best way to get the data out of spark in sparse form? It seems like there is only a toArray() method on sparse vectors, which outputs numpy arrays. However, the docs do say that scipy sparse arrays can be used in the place of spark sparse arrays.

还请记住，tf_idf值实际上是一列稀疏数组.理想情况下，将所有这些功能整合到一个大的稀疏矩阵中会很好.

Keep in mind also that the tf_idf values are in fact a column of sparse arrays. Ideally it would be nice to get all these features into one large sparse matrix.

推荐答案

一种可能的解决方案可以表示为:

One possible solution can be expressed as follows:

• 将特征转换为RDD并提取向量:

from pyspark.ml.linalg import SparseVector
from operator import attrgetter

df = sc.parallelize([
    (SparseVector(3, [0, 2], [1.0, 3.0]), ),
    (SparseVector(3, [1], [4.0]), )
]).toDF(["features"])

features = df.rdd.map(attrgetter("features"))

添加行索引:

add row indices:

indexed_features = features.zipWithIndex()

展平为元组(i, j, value)的RDD:

flatten to RDD of tuples (i, j, value):

def explode(row):
    vec, i = row
    for j, v in zip(vec.indices, vec.values):
        yield i, j, v

entries = indexed_features.flatMap(explode)

收集并重塑:

collect and reshape:

row_indices, col_indices, data = zip(*entries.collect())

计算形状:

compute shape:

shape = (
    df.count(),
    df.rdd.map(attrgetter("features")).first().size
)

创建稀疏矩阵:

create sparse matrix:

from scipy.sparse import csr_matrix

mat = csr_matrix((data, (row_indices, col_indices)), shape=shape)

快速健全性检查:

quick sanity check:

mat.todense()

预期结果:

matrix([[ 1.,  0.,  3.],
        [ 0.,  4.,  0.]])

另一个:

• 将features的每一行转换为矩阵:

• convert each row of features to matrix:

import numpy as np

def as_matrix(vec):
    data, indices = vec.values, vec.indices
    shape = 1, vec.size
    return csr_matrix((data, indices, np.array([0, vec.values.size])), shape)

mats = features.map(as_matrix)

并用vstack减小:

from scipy.sparse import vstack

mat = mats.reduce(lambda x, y: vstack([x, y]))

或collect和vstack

mat = vstack(mats.collect())

这篇关于pyspark:稀疏向量到稀疏矩阵的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！