数据可视化

import pandas as pd#import zipfileimport matplotlib.pyplot as pl#z = zipfile.ZipFile('../input/train.csv.zip')#print(z.namelist())train = pd.read_csv(open(r'C:\Users\lujinyu\Desktop\atae-lstm\atae-lstm\data\train.csv'), parse_dates=['Dates'])train['Year'] = train['Dates'].map(lambda x: x.year)train['Week'] = train['Dates'].map(lambda x: x.week)train['Hour'] = train['Dates'].map(lambda x: x.hour)print(train.head())train.PdDistrict.value_counts().plot(kind='bar', figsize=(8,10))pl.savefig('district_counts.png')train['event']=1weekly_events = train[['Week','Year','event']].groupby(['Year','Week']).count().reset_index()weekly_events_years = weekly_events.pivot(index='Week', columns='Year', values='event').fillna(method='ffill')#%matplotlib inlineax = weekly_events_years.interpolate().plot(title='number of cases every 2 weeks', figsize=(10,6))pl.savefig('events_every_two_weeks.png')hourly_events = train[['Hour','event']].groupby(['Hour']).count().reset_index()hourly_events.plot(kind='bar', figsize=(6, 6))pl.savefig('hourly_events.png')hourly_district_events = train[['PdDistrict','Hour','event']].groupby(['PdDistrict','Hour']).count().reset_index()hourly_district_events_pivot = hourly_district_events.pivot(index='Hour', columns='PdDistrict', values='event').fillna(method='ffill')hourly_district_events_pivot.interpolate().plot(title='number of cases hourly by district', figsize=(10,6))pl.savefig('hourly_events_by_district.png')