进一步有进一步的欢喜~

1、填充特征值为’//N’的所有列为None data[i][data[i] == '\\N'] = None 2、labelencoder from sklearn.preprocessing import LabelEncoder le = LabelEncoder() data[cat] = le.fit_transform(data[cat])

import numpy as np a=np.load('ip_explain_by_geoip2_china.npy',allow_pickle=True) ip_exp=a.item() temp = pd.DataFrame(list(ip_exp.items()), columns=['ip', 'ip_exp']) temp[['country','province_exp','c...

1、GBDT，xgboost对比 添加链接描述 https://wenku.baidu.com/view/f3da60b4951ea76e58fafab069dc5022aaea463e.html 2、xgboost论文 https://arxiv.org/pdf/1603.02754.pdf 3、lightgbm论文 http://papers.nips.cc/paper/6907-lightg...

def ctr_fea(train,test,feature): for fea in feature: print(fea) temp = train[['label',fea]].groupby(fea)['label'].agg({fea+'_sum':sum, ...

在def cnt_fea(data,feature,train_num): data['flag'] = '-' for fea in feature: print(fea) data[fea] = data[fea].map(data[fea].value_counts()) for i in range(len(feature)-1): ...

差异性编码快速写法 1、取set（） 2、建pd.dataframe格式 3、merge（） arrs = ['adidmd5', 'imeimd5', 'macmd5', 'openudidmd5', 'ip'] val = [] for i in range(len(arrs)): val.append(list(set(train[arrs[i]].unique()) & s...

# 众数 def get_mode(arr): mode = [] arr_appear = dict((a, arr.count(a)) for a in arr) # 统计各个元素出现的次数 if max(arr_appear.values()) == 1: # 如果最大的出现为1 return # 则没有众数 else: ...

for col in obj_cols: data[col].fillna('-1', inplace = True) data[col] = data[col].map(dict(zip(data[col].unique(),list(range(data[col].nunique()))))) print(col+' over...')

1、华为比赛总结 1、top2选手：EDA探索 比赛第一步，先做EDA，发现强特具体来说，如观察某个变量对于label的分布 2、top1选手：比赛理论 3、自我总结 理论深挖一下，如lgb模型原理，nn原理，避免侥幸。 比赛不能犯懒，理论补充不能犯懒 不能有依赖心理，不能仅靠依赖队友 做技术需要静下心来 不能有畏难心理 2、Ctr总结 1、EDA 观察特征，比如观察uid_value_cou...

import geoip2.database import sys # ip = input() ip = '210.32.149.0' reader = geoip2.database.Reader('./GeoLite2-City.mmdb') data = reader.city(ip) def ip_explain(ip): data = reader.city(ip) ...

# train = data[data.label!=-1] # test = data[data.label==-1] # train = train.dropna() # test = test.dropna() # # for i in data.columns: # for i in ['city','lan', 'os', 'osv', 'ver', 'orientation', 'ca...

# train_pos = data[data['label']==1] # train_neg = data[data['label']==0] # train = train.dropna() # test = test.dropna() # for i in ['city','lan', 'os', 'osv', 'ver', 'orientation', 'carrier', 'ntt',...

cols = ['area','location', 'pv/uv', 'totalFloor', 'pv', 'shi'] cols_kmeans = [] for i in cols: data[i+'_kmeans'] = (data[i]- data[i].min())/(data[i].max() - data[i].min()) cols_kmeans.append(i...

import seaborn as sns import matplotlib.pyplot as plt for i in train.columns: try: g = sns.kdeplot(train[i], color="Red", shade = True) g = sns.kdeplot(test[i], ax =g, color="Blue"...

#分层抽样 gbr = data.groupby("area") gbr.groups typicalFracDict = { 1: 0.2, 2: 0.4, 3: 0.6 } def typicalSampling(group, typicalFracDict): name = group.name frac = typicalFracDi...

根据ip获取信息 import requests import IPy def get_location(ip): url = 'https://sp0.baidu.com/8aQDcjqpAAV3otqbppnN2DJv/api.php?co=&resource_id=6006&t=1529895387942&ie=utf8&oe=gbk&c...

# def cut_col(data, col_name, cut_list): # print('cutting', col_name) # def _trans(array): # count = array['box_counts'] # for box in cut_list: # if count <= bo...

1、排序 train_new.sort_values(by='imeimd5') train_new.sort_values(by='imeimd5')['imeimd5'].max() train_ime = train_new['imeimd5'].unique() 2、迭代器进度条：tqdm tqdm cnt = 0 for i in tqdm.tqdm_notebook(test_ne...

import numpy as np # a=np.load('ip_dict.npy',allow_pickle=True) # data=a.item() temp = train[['ip','label']].groupby('ip')['label'].agg({'mean_label':'mean','count_label':'count','sum_label':'sum'}...

# 特征重要性 import matplotlib.pyplot as plt import seaborn as sns cols = (feature_importance_data[["feature", "importance"]] .groupby("feature") .mean() .sort_values(by="importance...

folds = KFold(n_splits=5, shuffle=True, random_state=1333) oof_lgb = np.zeros(len(train)) predictions_lgb = np.zeros(len(test)) feature_importance_data = pd.DataFrame() best_score = 0 learning_rate ...