Buevara的博客

来源于： Kaggle Lending Club Loan Data数据可视化分析与不良贷款预测#对特征缺失值的处理计算特征缺失值比例的函数:def draw_missing_data_table(data): total = data.isnull().sum().sort_values(ascending=False) percent = (data.isnull().sum() / data.shape[0]).sort_values(ascending=False) m

sns.set_style(“whitegrid”)必须注释或删除 否则改任何配置都是错的(都是血泪的教训)import seaborn as snssns.set_style({'font.sans-serif':['SimHei']})plt.rcParams ['font.sans-serif'] = ['SimHei'] #Used to display Chinese labels normallyplt.rcParams ['axes.unicode_minus'] = Fal.

# XX为df格式 train特征表feature_names = XX.columns.tolist()feature_names = np.array(feature_names)feature_importances = clf.feature_importances_indices = np.argsort(feature_importances)[::-1]x = feature_importances[show_indices],y=feature_names[indices]

import mathdef scale(x): if x > 2: x = int(math.log(float(x))**2) return xdata_value['I5']=data_value['I5'].apply(scale)原始数据data_value.head(10)进行函数变换后：data_value.head(10)

转载自：https://www.it1352.com/1689965.htmlimport numpy as npimport pandas as pdarr = np.array( [[1,0,1,1,1,5], [0,0,0,0,1,3], [1,0,0,0,1,1], [1,0,0,1,1,1], [1,0,0,0,1,1], [1,1,0,0,1,1]] )df = pd.DataFrame( np.repeat(

numpy.dtype.kinddtype.kindA character code (one of ‘biufcmMOSUV’) identifying the general kind of data.b booleani signed integeru unsigned integerf floating-pointc complex floating-pointm timedeltaM datetimeO objectS (byte-)stringU Un

import numpy as nplist1 = np.array([4,1,5,8,7,9,2])list2 = np.array(["44","11","55","88","77","99","22"])# 需要从大到小 加上[::-1] sort_1 = np.argsort(list1)[::-1]res = list2[sort_1]print(res)结果：['99' '88' '77' '55' '44' '22' '11']...

1、两个列表转换list1 = [‘key1’,‘key2’,‘key3’]list2 = [‘1’,‘2’,‘3’]把他们转为这样的字典：{‘key1’:‘1’,‘key2’:‘2’,‘key3’:‘3’}ist1 = ['key1','key2','key3']list2 = ['1','2','3']dict(zip(list1,list2)){'key1':'1','key2':'2','key3':'3'}2、嵌套列表转换有两种方法new_list= [['key1',

def calculateIV(train_data,label_columnName): ''' @description @param label_columnName: column name of label @return train_data: pd.DataFrame，include label test_data: pd.DataFrame ''' import math print("calcul

import pandas as pd### d 传入是浅拷贝 只拷贝数据地址 而不是数据 当运行完d[cname] = d[cname].map(dic) 输入的d已经改变 就算没有返回值，函数外的d也发生了改变def string2index(d, cname_list): #空值不进行编号 for cname in cname_list: counts = d[cname].value_counts() s = counts.index.toli

pearson相关系数计算函数data.corr()该方法支持空值：np.nanimport seaborn as snsimport numpy as npimport matplotlib.pyplot as pltdata = pd.DataFrame({"A":[np.nan,2,9], "B":[4,14,6], "c":[987,8,9]})f, ax= plt.subplots(figsize = (14, 10))corr = data.corr()# print(

计算WoE指标时：import numpy as npimport mathimport pandas as pddf_min_max_bin["ok"] = ((df_min_max_bin["nums_label_1"]+1)/sum_label_1)/((df_min_max_bin["nums_label_0"]+1)/sum_label_0) df_min_max_bin["WoE"] = math.log(df_min_max_bin["ok"])结果报错：T

1. mn = ab.append(cd)ab = pd.DataFrame({"aa":[1,2,3,np.nan,np.nan],"bb":[11,22,33,44,55]})print(ab)cd = pd.DataFrame({"aa":[1,2,5,123,4546],"bb":[11,22,55,111,222]})print(cd)mn = ab.append(cd, ignore_index = True)print(mn)一定要写等式，因为append后不是覆盖ab 而是

pandas判断df中某个值是否为空正确使用：pd.notnull(s.iloc[1,1])错误方法：1. s.iloc[1,1].empty2.s.iloc[1,1].isnull()3.s.iloc[1,1]==np.nanpandas中c列值取决于 a,b两列例子中 bb列取决于 cc和a这两列deo = pd.DataFrame({"a":[1,2,3,np.nan,5],"bb":[0,0,0,0,0], "cc":[22,22,22,22,22]})deo.head()

# 天数>=15 label 为1# 天数<15 label 为0f = lambda s: 1 if s["天数"]>=15 else 0## 如果只想设置 大于阈值 小于阈值 取本身值f = lambda s: 1 if s["天数"]>=15 else s["天数"]result["label"] = result.apply(f, axis=1)result["label"].value_counts()...

整表保留两位小数方法import numpy as npimport pandas as pdformat="{0:.02f}".format# 或者#format = lambda x:'%.2f' % xx = x.applymap(format)x.head()注意：***通过此步骤后，所有列属性均变为object (str类型) ***

错误做法：lists = [[]] * 3lists[0].append(3)[[3], [3], [3]]正确做法：lists = [[] for i in range(3)]lists[0].append(3)lists[1].append(5)lists[2].append(7)[[3], [5], [7]]https://www.cnblogs.com/PyLe...

1. 不包含某一个特征#取p_feature中不包含列名为index 的所有列p_feature = p_feature.iloc[:, p_feature.columns != "index"]2.不包含多个特征拓展： 选取某一列中等于或不等于某个值的dfisin()参考1：https://www.cnblogs.com/nxf-rabbit75/p/10105271.h...

ValueError: Input contains NaN, infinity or a value too large for dtype(‘float64’).检查：https://blog.youkuaiyun.com/yi976263092/article/details/87878112

#等频分箱def frequencybox(demo, name, new_name, n): demo["tmp"] = pd.qcut(demo[name],n) group_by_age_bin = demo.groupby(["tmp"],as_index=True) df_min_max_bin = pd.DataFrame()#用来记录每个箱体的最大...

#利用泰坦尼克数据data.head() PassengerId Survived Pclass Name Sex Age SibSp Parch Ticket Fare Cabin Embarked0 1 0 3 Braund, Mr. Owen Harris male 22.0 1 0 A/5 21171 7.2500 NaN S1 ...

1.pandas修改列类型使用astype如下：df[[column]] = df[[column]].astype(type)-2.

在处理数据时NAN值是非常常见的，但是NaN值你真的了解吗？下面让楓尘君带大家了解一下这个神奇的NAN在csv表格中长这样：在python里展示为这样：我们创建这个值时是这样：import numpy as npnp.nan但是想判断是否是np.nan还真是有点难度：a=np.nanprint(a==np.nan)结果是Falseprint(a.equal(np.na...

pandas从dataframe中获取值可以用:import numpy as npfrom pandas import Series, DataFrameimport pandas as pddf = DataFrame({'CST_NO': [11, 22, 11, 22, 11, 22], 'AMOUNT': [10, 20, 30, np.nan...

假设有两个lista = [2,3,1,5]b = [“b”, “c”, “a”, “e”]想为a生序排列变为：sorta=[1,2,3,5]希望b对应变为：sortb=[“a”,“b”, “c”, “e”]处理方案：list1, list2 = (list(t) for t in zip(*sorted(zip(list1, list2))))或者：from opera...

今天遇到一个问题：pos_data = session_data.loc[(session_data['CST_NO'].isin(pos_list))&(session_data['BIZ_CODE']=='TRANSFER')]报错：查阅资料后发现，是isin的左右类型不一致，都转化为str则不会出现这种问题：改为：添加 astype(str)pos_data = ses...