In [57]:

    

import matplotlib.pyplot as plt
from datetime import datetime
from collections import Counter
import numpy as np
import json
import random
%matplotlib inline


    

In [58]:

    

def time(time):
    try:
        return datetime.strptime(time[:19], '%Y-%m-%dT%H:%M:%S')
    except ValueError:
        pass


    

In [59]:

    

logs_count = 400
logs = open('').read()
logs = logs.split('\n')
logs_new = []


    

In [60]:

    

logs = random.sample(logs, 30000)


    

In [61]:

    

logs = list(map(lambda a: a.split('\t'), logs))


    

In [62]:

    

a = 0
for log in logs:
    if time(log[0]):
        log[0] = time(log[0])
    logs[logs.index(log)] = log
    if a % 100 == 0: 
        print a
    a+=1


    

    




0

In [63]:

    

logs[:10]


    

    
Out[63]:





[[datetime.datetime(2016, 5, 13, 23, 49, 2),

In [64]:

    

first_time = {}
for log in logs:
    new= log[0].hour + log[0].day*24 + log[0].month*24*7
    try:
        if first_time[log[1]] > new:
            first_time[log[1]] = new
    except:
        first_time[log[1]] = new


    

In [65]:

    

import pandas as pd
data = {}
for a in range(0,101):
    data[a] = [0]*len(Counter(list(map(lambda a: a[1], logs))))


    

In [66]:

    

data = pd.DataFrame.from_dict(data)


    

In [67]:

    

data.index = Counter(list(map(lambda a: a[1], logs))).keys()


    

In [68]:

    

for log in logs:
    old = log[0].hour + log[0].day*24 + log[0].month*24*7
    new = first_time[log[1]]
    if not old-new > 99:
        data[old-new][log[1]]+= 1


    

In [89]:

    

k = 100


    

In [90]:

    

y = []
for i in range(0,k):
    y.append(np.mean(data[i].values))


    

In [91]:

    

x = range(0,k)


    

In [92]:

    

plt.figure(figsize=(40,20))
plt.grid(True)
plt.plot(x,y)


    

    
Out[92]:





[<matplotlib.lines.Line2D at 0x7f9dcf6fa490>]