HTML5ページオーディオの自動再生実装

import pandas as pd
import numpy as np
from pandas import DataFrame,Series
import matplotlib.pyplot as plt
import sys

import seaborn as sns
#import lightgbm as lgb
import gc
import time
from sklearn import preprocessing
from sklearn.metrics import precision_recall_curve
from sklearn.model_selection import train_test_split
from sklearn.feature_selection import VarianceThreshold

# Read the data content and save it in dataframe format
data = pd.read_csv('. /input/atec_anti_fraud_train.csv')
data['id'] = data['id'].astype('category')
data['label'] = data['label'].astype('category')
# print(data.describe())
# View the distribution of the labels of the sample
fre = data.label.value_counts()
# Observe the amount of data per day
date_info = data['date'].value_counts().sort_index()

data.describe()
               date f1 ...           f296 f297
count 9.947310e+05 994731.000000 ...  787146.000000 787146.000000
mean 2.017098e+07 0.726283 ...      41.244121 99.023500
std 6.011735e+01 0.651947 ...      45.658241 96.240483
min 2.017090e+07 0.000000 ...       1.000000 1.000000
25% 2.017092e+07 0.000000 ...       2.000000 2.000000
50% 2.017101e+07 1.000000 ...      17.000000 115.000000
75% 2.017102e+07 1.000000 ...      79.000000 194.000000
max 2.017110e+07 2.000000 ...     280.000000 299.000000

[8 rows x 298 columns]
----------------------------------------
fre = data.label.value_counts()
 0 977884
 1 12122
-1 4725
Name: label, dtype: int64
----------------------------------------
# print(date_info) # Average daily record size, from 20170905 to 20171105
# 20171029 15592
# 20171030 16455
# 20171031 16596
# 20171101 17884
# 20171102 17305
# 20171103 16406
# 20171104 15639
# 20171105 15891
# Name: date, Length: 62, dtype: int64

# Check for missing features, the first column is the number of missing features, the second column is the number of feature columns
data_lost=data.isnull().sum().reset_index(name='count')
data_lost.columns=['feature','counts']
lost_count=data_lost.groupby('counts',as_index='False').count()

print(data.isnull().sum())
id 0
label 0
date 0
f1 0
f2 0
f3 0
f4 0
f5 199825
f6 0
...
f288 207585
f289 207585
f290 207585
f291 207585
f292 207585
(f293 207585)
f294 207585
f295 207585
f296 207585
f297 207585
Length: 300, dtype: int64

#print(lost_count)
# feature
# counts
# 0 21
# 919 48
# 136021 69
# 199825 1 # Only one feature with missing number 199825
# 207448 6
# 207585 24
# 211036 4
# 270515 78
# 273904 14
# 286285 5
# 341661 14
# 341887 4
# 925781 12