pandas编程技巧
-
drop_duplicates 删除重复行
from pandas import DataFrame, Series info = [ {"order_id":1001, "name":"wangwu"}, {"order_id":1002, "name":"zhangsan"}, {"order_id":1003, "name":"lisi"}, {"order_id":1001, "name":"wang wu"} ] df = DataFrame(info) print(df) ''' #drop_duplicates 删除重复行.参数如下： subset:删除某一列下面的重复行 keep : {‘first’, ‘last’, False}, default ‘first’ ，默认：删除重复项并保留第一次出现的项 inplace : boolean, default False ，是直接在原来数据上修改还是保留一个副本 ''' df = df.drop_duplicates(subset="order_id", keep="first") print("删除重复行后:", "*"*30) print(df) -------------------------------------------结果------------ order_id name 0 1001 wangwu 1 1002 zhangsan 2 1003 lisi 3 1001 wang wu 删除重复行后: ****************************** order_id name 0 1001 wangwu 1 1002 zhangsan 2 1003 lisi -
将DataFrame转化为字典dict
from pandas import DataFrame, Series info = [ {"order_id":1001, "name":"wangwu"}, {"order_id":1002, "name":"zhangsan"}, {"order_id":1003, "name":"lisi"}, {"order_id":1001, "name":"wang wu"} ] df = DataFrame(info) print(df) print("orient=index", "*"*20) dic_1 = df.to_dict(orient = "index") dic_2 = df.to_dict(orient = "records") print("orient=index:","-"*10) print(dic_1) print("orient=records:", "-"*10) print(dic_2) dic_3 = df.to_dict(orient="dict") print("dic_3:", "-"*10) print(dic_3) dic_4 = df.to_dict(orient = "series") print("series:", "-"*10) print(dic_4) print("+"*5) print(dic_4["order_id"]) --------------------------------------结果----------------- order_id name 0 1001 wangwu 1 1002 zhangsan 2 1003 lisi 3 1001 wang wu orient=index ******************** orient=index: ---------- {0: {'order_id': 1001, 'name': 'wangwu'}, 1: {'order_id': 1002, 'name': 'zhangsan'}, 2: {'order_id': 1003, 'name': 'lisi'}, 3: {'order_id': 1001, 'name': 'wang wu'}} orient=records: ---------- [{'order_id': 1001, 'name': 'wangwu'}, {'order_id': 1002, 'name': 'zhangsan'}, {'order_id': 1003, 'name': 'lisi'}, {'order_id': 1001, 'name': 'wang wu'}] dic_3: ---------- {'order_id': {0: 1001, 1: 1002, 2: 1003, 3: 1001}, 'name': {0: 'wangwu', 1: 'zhangsan', 2: 'lisi', 3: 'wang wu'}} series: ---------- {'order_id': 0 1001 1 1002 2 1003 3 1001 Name: order_id, dtype: int64, 'name': 0 wangwu 1 zhangsan 2 lisi 3 wang wu Name: name, dtype: object} +++++ 0 1001 1 1002 2 1003 3 1001 Name: order_id, dtype: int64 <class 'pandas.core.series.Series'> list: ---------- {'order_id': [1001, 1002, 1003, 1001], 'name': ['wangwu', 'zhangsan', 'lisi', 'wang wu']} -
apply函数
axis=0/1 默认=0， 1：以行为单位； 0：以列为单位
from pandas import DataFrame, Series info = [ {"order_id":1001, "name":"wangwu", "score_1":90, "score_2":85}, {"order_id":1002, "name":"zhangsan", "score_1":60, "score_2":75}, {"order_id":1003, "name":"lisi", "score_1":96, "score_2":95}, {"order_id":1001, "name":"wang wu", "score_1":89, "score_2":90} ] df = DataFrame(info) df["score1_is_good"] = df.apply(lambda x: 1 if x.score_1 > 80 else 0, axis = 1) df["score2_is_good"] = df.apply(lambda x: 1 if x.score_2 > 80 else 0, axis = 1) df["score_total"] = df.apply(lambda x: x.score_1 + x.score_2, axis = 1) print(df) -----------------------结果----------------- order_id name score_1 score_2 score1_is_good score2_is_good score_total 0 1001 wangwu 90 85 1 1 175 1 1002 zhangsan 60 75 0 0 135 2 1003 lisi 96 95 1 1 191 3 1001 wang wu 89 90 1 1 179
例子2：对特征进行归一化处理
import numpy as np
from pandas import Series, DataFrame
arr = np.random.randn(5,6)
df = DataFrame(arr)
print(df)
def feature_normal(arr):
return (arr - arr.min())/(arr.max() - arr.min())
print("*"*20,“归一化后的结果”)
print(df.apply(feature_normal , axis = 0))
---------------------------------result------------------------
0 1 2 3 4 5
0 1.079595 -0.351661 0.309417 2.155845 -2.012087 0.835636
1 1.594347 0.524309 0.642305 0.138198 -0.075382 -0.791217
2 1.849154 0.651603 0.574854 0.741077 -0.788106 -0.115997
3 0.686670 -0.787141 0.048679 0.520909 1.597393 1.850610
4 -0.391876 -0.536599 -1.522485 -0.463636 0.478463 -0.351089
********************
0 1 2 3 4 5
0 0.656605 0.302681 0.846226 1.000000 0.000000 0.615806
1 0.886299 0.911524 1.000000 0.229753 0.536561 0.000000
2 1.000000 1.000000 0.968842 0.459905 0.339102 0.255588
3 0.481272 0.000000 0.725781 0.375855 1.000000 1.000000
4 0.000000 0.174139 0.000000 0.000000 0.690003 0.166600
-
groupby()：根据某个字段进行分组
agg()：对分组后的结果进行聚合运算
reset_index():对分组后的数据重新编排顺序
merge():将两个dataframe根据某个字段进行列融合
from pandas import DataFrame, Series import numpy as np info = [ {"order_id":1001, "name":"wangwu", "score_1":90, "score_2":85, "created_time": "2019-10-10"}, {"order_id":1002, "name":"zhangsan", "score_1":60, "score_2":75, "created_time": "2019-10-10"}, {"order_id":1002, "name":"zhangsan", "score_1":96, "score_2":95, "created_time": "2019-08-10"}, {"order_id":1001, "name":"wangwu", "score_1":89, "score_2":90, "created_time": "2019-08-10"} ] df = DataFrame(info) gp1 = df.groupby("order_id").agg({"score_1":["max", "min", np.mean, "count"]}) gp1 = gp1.reset_index() gp1.columns = ["order_id", "score_1_max", "score_1_min", "score_1_avg", "score_1_count"] print(gp1) gp2 = df.groupby("order_id").agg({"score_2":["max", "min", np.mean, "count"]}) gp2 = gp2.reset_index() gp2.columns = ["order_id", "score_2_max", "score_2_min", "score_2_avg", "score_2_count"] print(gp2) df2 = df.merge(gp1, on="order_id", how="inner").merge(gp2, on="order_id", how="inner") print(df2) df3 = df2[["order_id", "name", "score_1_max", "score_1_min", "score_1_avg", "score_1_count", "score_2_max", "score_2_min", "score_2_avg", "score_2_count"]] df3 = df3.drop_duplicates(subset = "order_id", keep = "first") print(df3) --------------------------------------结果------------------ order_id score_1_max score_1_min score_1_avg score_1_count 0 1001 90 89 89.5 2 1 1002 96 60 78.0 2 order_id score_2_max score_2_min score_2_avg score_2_count 0 1001 90 85 87.5 2 1 1002 95 75 85.0 2 order_id name score_1 score_2 ... score_2_max score_2_min score_2_avg score_2_count 0 1001 wangwu 90 85 ... 90 85 87.5 2 1 1001 wangwu 89 90 ... 90 85 87.5 2 2 1002 zhangsan 60 75 ... 95 75 85.0 2 3 1002 zhangsan 96 95 ... 95 75 85.0 2 [4 rows x 13 columns] order_id name score_1_max score_1_min ... score_2_max score_2_min score_2_avg score_2_count 0 1001 wangwu 90 89 ... 90 85 87.5 2 2 1002 zhangsan 96 60 ... 95 75 85.0 2 [2 rows x 10 columns] -
对Series进行去重unique()操作
from pandas import DataFrame, Series info = [ {"order_id":1001, "name":"wangwu", "score_1":90, "score_2":85, "created_time": "2019-10-10"}, {"order_id":1002, "name":"zhangsan", "score_1":60, "score_2":75, "created_time": "2019-10-10"}, {"order_id":1002, "name":"zhangsan", "score_1":96, "score_2":95, "created_time": "2019-08-10"}, {"order_id":1001, "name":"wangwu", "score_1":89, "score_2":90, "created_time": "2019-08-10"} ] df = DataFrame(info) order_list = df["order_id"] print(order_list, type(order_list)) order_list = df["order_id"].unique() print(order_list, type(order_list)) -------------------------------result-------------------- 0 1001 1 1002 2 1002 3 1001 Name: order_id, dtype: int64 <class 'pandas.core.series.Series'> [1001 1002] <class 'numpy.ndarray'> -
concat()
from pandas import DataFrame, Series import pandas as pd df_1 = DataFrame([ {"order_id": 1001, "name": "wangwu"}, {"order_id": 1002, "name": "lisi"}, {"order_id": 1003, "name": "zhangsan"}, ]) df_2 = DataFrame([ {"order_id": 1004, "name": "wangqi"}, {"order_id": 1005, "name": "wangfang"}, {"order_id": 1006, "name": "zhouliu"}, ]) df_new = pd.concat([df_1, df_2], axis =0) #行合并 （沿着垂直的方向） df_new2 = pd.concat([df_1, df_2], axis =1) #列合并 （沿着水平的方向） print(df_new) print(df_new2) ---------------------------result------------------- order_id name 0 1001 wangwu 1 1002 lisi 2 1003 zhangsan 0 1004 wangqi 1 1005 wangfang 2 1006 zhouliu order_id name order_id name 0 1001 wangwu 1004 wangqi 1 1002 lisi 1005 wangfang 2 1003 zhangsan 1006 zhouliu
273
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
