In [1]:
import os, json
import numpy as np
import pandas as pd
import datetime as dt
import matplotlib.pyplot as plt

Reading and cleaning the data

To extract the data from the Endomondo export, I first create a list of all workout files.

In [2]:
path_to_json = 'workouts/'
json_files = [pos_json for pos_json in os.listdir(path_to_json) if pos_json.endswith('.json')]

Then, from each file, I extract the data I need and append it to a list, so every list item is a json.

In [3]:
data = []

for f in json_files:
    with open('workouts/'+f) as json_file:
        json_data = json.load(json_file)

        d = {}
        for j in json_data:
            d.update(dict(j))

        if d['sport'] == 'WALKING':
            data.append(d)

Then I can use the list of jsons to create a data frame.

In [4]:
df = pd.DataFrame(data)
df = df[['start_time', 'distance_km', 'duration_s', 'speed_avg_kmh']]

Fixing some data type and unit tings.

In [5]:
df['start_time'] = pd.to_datetime(df['start_time'], format='%Y/%m/%d').dt.normalize()
df['duration_min'] = df['duration_s'] / 60
df.head()
Out[5]:
start_time distance_km duration_s speed_avg_kmh duration_min
0 2020-03-13 1.179 681.77 6.225560 11.362833
1 2017-04-29 6.504 3686.43 6.351511 61.440500
2 2017-05-15 2.195 1458.99 5.416075 24.316500
3 2017-10-09 3.599 2351.28 5.510360 39.188000
4 2017-10-11 2.388 1517.30 5.665854 25.288333

Since I only need data from Feb and March 2020, I filter out the rest.

In [6]:
df = df[df['start_time']>='2020-02-01'].sort_values(by='start_time')
df.reset_index(inplace=True, drop=True)
print(f'number of datapoints: {len(df)}')

number of datapoints: 38

Analysis

Since I'm mainly interested in average walking speed, let's see its descriptive statistics.

In [7]:
df[['speed_avg_kmh']].describe()
Out[7]:
speed_avg_kmh
count 38.000000
mean 6.026441
std 0.297371
min 5.535161
25% 5.780796
50% 6.038148
75% 6.219591
max 6.892989

Then, let's try to visualise the relationship between distance and duration in a scatter plot. We should see a nice linear correlation.

In order to draw the regression line, we need to calculate its formula.

In [8]:
from sklearn.linear_model import LinearRegression
In [9]:
model = LinearRegression()
In [10]:
x = df[['distance_km']]
y = df[['duration_min']]
model.fit(x, y)
model = LinearRegression().fit(x, y)

Now we can extract the y-intercept and the slop of the line.

In [11]:
intercept = model.intercept_[0]
slope = model.coef_[0][0]

print('intercept:', intercept)
print('slope:', slope)

intercept: 0.3217424666737756
slope: 9.779546814599154

And now finally to the visualisation.

In [12]:
fig,ax = plt.subplots(figsize = (7, 7))

# plot the dots for scatter
X = df[['distance_km']]
Y = df[['duration_min']]
plt.scatter(X, Y, s=100, alpha=.5, zorder=3)

# plot the regression line
x = np.linspace(np.min(X)-1, np.max(X)+1, 1000)
y = intercept + slope * x
plt.plot(x, y, linestyle='--', label='Regression Line', zorder=2)

# tweaks
ax.set_xlabel('distance km')
ax.set_ylabel('duration min')
ax.set_xlim(0, 3.5)
ax.set_ylim(0, 35)
plt.grid(which='major', zorder=0, color='#E8E8E8')

# plt.savefig('scatter.png', bbox_inches='tight', dpi=200)

plt.show()

We already learned before that my average walking speed was 6.03 km/h. According to Google, the average walking speed is some 5.2 km/h. So how do I compare? Am I significantly faster than an average person? Let's do a t-test.

In [13]:
import scipy.stats
In [14]:
pop_mean = 5.2
scipy.stats.ttest_1samp(df['speed_avg_kmh'], pop_mean)
Out[14]:
Ttest_1sampResult(statistic=17.1318870129101, pvalue=3.518159029717419e-19)

The p value is very very small, so we can conclude that I am faster than an averge person.