python pd.cut()和pd.qcut()分箱操作
cbright63 人气:51.cut()可以实现类似于对成绩进行优良统计的功能,来看代码示例。
假如我们有一组学生成绩,我们需要将这些成绩分为不及格(0-59)、及格(60-70)、良(71-85)、优(86-100)这几组。这时候可以用到cut()
import numpy as np import pandas as pd # 我们先给 scores传入30个从0到100随机的数 scores = np.random.uniform(0,100,size=30) # 然后使用 np.round()函数控制数据精度 scores = np.round(scores,1) # 指定分箱的区间 grades = [0,59,70,85,100] cuts = pd.cut(scores,grades) print('\nscores:') print(scores) print('\ncuts:') print(cuts) # 我们还可以计算出每个箱子中有多少个数据 print('\ncats.value_counts:') print(pd.value_counts(cuts)) ======output:====== scores: [ 6. 50.8 80.2 22.1 60.1 75.1 30.8 50.8 81.6 17.4 13.4 24.3 67.3 84.4 63.4 21.3 17.2 3.7 40.1 12.4 15.7 23.1 67.4 94.8 72.6 12.8 81. 82. 70.2 54.1] cuts: [(0, 59], (0, 59], (70, 85], (0, 59], (59, 70], ..., (0, 59], (70, 85], (70, 85], (70, 85], (0, 59]] Length: 30 Categories (4, interval[int64]): [(0, 59] < (59, 70] < (70, 85] < (85, 100]] cuts.value_counts: (0, 59] 17 (70, 85] 8 (59, 70] 4 (85, 100] 1 dtype: int64
默认情况下,cat()的区间划分是左开右闭,可以传递right=False来改变哪一边是封闭的
代码示例:
cuts = pd.cut(scores,grades,right=False)
也可以通过向labels选项传递一个列表或数组来传入自定义的箱名
代码示例:
group_names = ['不及格','及格','良','优秀'] cuts = pd.cut(scores,grades,labels=group_names)
当我们不需要自定义划分区间时,而是需要根据数据中最大值和最小值计算出等长的箱子。
代码示例:
# 将成绩均匀的分在四个箱子中,precision=2的选项将精度控制在两位 cuts = pd.cut(scores,4,precision=2)
2.qcut()可以生成指定的箱子数,然后使每个箱子都具有相同数量的数据
代码示例:
import numpy as np import pandas as pd # 正态分布 data = np.random.randn(100) # 分四个箱子 cuts = pd.qcut(data,4) print('\ncuts:') print(cuts) print('\ncuts.value_counts:') print(pd.value_counts(cuts)) ======output:====== cuts: [(-0.745, -0.0723], (0.889, 2.834], (-0.745, -0.0723], (0.889, 2.834], (0.889, 2.834], ..., (-0.745, -0.0723], (-0.0723, 0.889], (-3.1599999999999997, -0.745], (-0.745, -0.0723], (-0.0723, 0.889]] Length: 100 Categories (4, interval[float64]): [(-3.1599999999999997, -0.745] < (-0.745, -0.0723] < (-0.0723, 0.889] < (0.889, 2.834]] cuts.value_counts: (0.889, 2.834] 25 (-0.0723, 0.889] 25 (-0.745, -0.0723] 25 (-3.1599999999999997, -0.745] 25 dtype: int64
加载全部内容