Yandex.Praktikum 🍂

sprint3_1.py

import pandas as pd

data = pd.read_csv("/datasets/visits.csv", sep="\t")
data['local_time'] = (
    pd.to_datetime(data['date_time'], format='%Y-%m-%dT%H:%M:%S')
    + pd.Timedelta(hours=3)
)
data['date_hour'] = data['local_time'].dt.round('1H')
data['too_fast'] = data['time_spent'] < 60
data['too_slow'] = data['time_spent'] > 1000

too_fast_stat = data.pivot_table(index='id', values='too_fast')
good_ids = too_fast_stat.query('too_fast < 0.5')
good_data = data.query('id in @good_ids.index')
good_data = good_data.query('60 <= time_spent <= 1000')

station_stat = data.pivot_table(index="id", values="time_spent", aggfunc="median")
good_station_stat = good_data.pivot_table(index="id", values="time_spent", aggfunc="median")

stat = data.pivot_table(index='name', values='time_spent')
good_stat = good_data.pivot_table(index='name', values='time_spent', aggfunc='median')
stat['good_time_spent'] = good_stat['time_spent']

name_stat = data.pivot_table(index='name', values='time_spent')
good_name_stat = good_data.pivot_table(index='name', values='time_spent', aggfunc='median')
name_stat['good_time_spent'] = good_name_stat['time_spent']

id_name = good_data.pivot_table(index='id', values='name', aggfunc=['first', 'count'])
id_name.columns = ['name', 'count']
station_stat_full = id_name.join(good_station_stat)

good_stat2 = (
    station_stat_full
    .query('count > 30')
    .pivot_table(index='name', values='time_spent', aggfunc=['median', 'count'])
)
good_stat2.columns = ['median_time', 'stations']
final_stat = stat.join(good_stat2)

station_stat_multi = pd.pivot_table(data, index='id', values=['time_spent', 'too_fast', 'too_slow'])
print(station_stat_multi.corr())

pd.plotting.scatter_matrix(station_stat_multi, figsize=(9, 9))

sprint3_1.txt

Результат
            time_spent  too_fast  too_slow
time_spent    1.000000 -0.640658  0.802247
too_fast     -0.640658  1.000000 -0.255876
too_slow      0.802247 -0.255876  1.000000

Если долго стоять, то можно простоять слишком долго. А если быстро сбежать, то не больно-то и задержишься.
Как ни смешно, это окончательное подтверждение того, что вы отрезали ненужное там, где надо.
Если бы не получались результаты «капитана очевидность», пришлось бы вернуться и пришивать отрезанное на место.

sprint3_2.py

import pandas as pd

data = pd.read_csv("/datasets/visits.csv", sep="\t")
data['local_time'] = (
    pd.to_datetime(data['date_time'], yearfirst=True)
    + pd.Timedelta(hours=3)
)
data['date_hour'] = data['local_time'].dt.round('1H')
data['too_fast'] = data['time_spent'] < 60
data['too_slow'] = data['time_spent'] > 1000

too_fast_stat = data.pivot_table(index='id', values='too_fast')
good_ids = too_fast_stat.query('too_fast < 0.5')
good_data = data.query('id in @good_ids.index')
good_data = good_data.query('60 <= time_spent <= 1000')

station_stat = data.pivot_table(index="id", values="time_spent", aggfunc="median")
good_station_stat = good_data.pivot_table(index="id", values="time_spent", aggfunc="median")

stat = data.pivot_table(index='name', values='time_spent')
good_stat = good_data.pivot_table(index='name', values='time_spent', aggfunc='median')
stat['good_time_spent'] = good_stat['time_spent']

id_name = good_data.pivot_table(index='id', values='name', aggfunc=['first', 'count'])
id_name.columns = ['name', 'count']
station_stat_full = id_name.join(good_station_stat)

good_stat2 = (
    station_stat_full
    .query('count > 30')
    .pivot_table(index='name', values='time_spent', aggfunc=['median', 'count'])
)
good_stat2.columns = ['median_time', 'stations']
final_stat = stat.join(good_stat2)

big_nets_stat = final_stat.query('stations > 10')

station_stat_full['group_name'] = (
    station_stat_full['name']
    .where(station_stat_full['name'].isin(big_nets_stat.index), 'Другие')
)

stat_grouped = (
    station_stat_full
    .query('count > 30')
    .pivot_table(index='group_name', values='time_spent', aggfunc=['median', 'count'])
)
stat_grouped.columns = ['time_spent', 'count']

good_data['group_name'] = (
    good_data['name']
    .where(good_data['name'].isin(big_nets_stat.index), 'Другие')
)

for group_name, group_data in good_data.groupby('group_name'):
    group_data.plot(kind='hist', y='time_spent', bins=50, title=group_name)