29 lines
726 B
Python
29 lines
726 B
Python
import subprocess
|
|
import pandas as pd
|
|
import numpy as np
|
|
# import kaggle
|
|
|
|
# kaggle.api.authenticate()
|
|
# kaggle.api.dataset_download_files('shivamb/real-or-fake-fake-jobposting-prediction', path='fake_job_postings.csv', unzip=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
data=pd.read_csv('data.csv')
|
|
data = data.replace(np.nan, '', regex=True)
|
|
|
|
print("="*20)
|
|
print('Ilość wierszy w zbiorze: ',len(data))
|
|
|
|
print("="*10, ' data["department"].value_counts() ', 10*'=')
|
|
print(data["department"].value_counts())
|
|
|
|
print("="*10, ' data.median() ', 10*'=')
|
|
print(data.median())
|
|
|
|
print("="*10, ' data.describe(include="all") ', 10*'=')
|
|
print(data.describe(include='all'))
|
|
|
|
data.describe(include="all").to_csv(r'stats.txt', header=None, index=None, sep='\t', mode='a') |