Visualizing Eight Years of Strava Activity
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from datetime import datetime
import calendar
from matplotlib.lines import Line2D
Strava is an activity tracking application that allows users to see their distance and route, and a whole host of other interesting activity data. Since I have been using this application for over 7 years now, I have built up a large cache of personal data in the app. To see how my activity in general and my activity of choice in particular have changed over the years, I created the following data visualizations.
View this data visualization as a web app
Import and view data
df = pd.read_csv ('~/Desktop/Code/StravaData/activities.csv')
df.head()
| Activity ID | Activity Date | Activity Name | Activity Type | Activity Description | Elapsed Time | Distance | Relative Effort | Commute | Activity Gear | ... | Cloud Cover | Weather Visibility | UV Index | Weather Ozone | <span class="translation_missing" title="translation missing: en-US.lib.export.portability_exporter.activities.horton_values.jump_count">Jump Count</span> | <span class="translation_missing" title="translation missing: en-US.lib.export.portability_exporter.activities.horton_values.total_grit">Total Grit</span> | <span class="translation_missing" title="translation missing: en-US.lib.export.portability_exporter.activities.horton_values.avg_flow">Avg Flow</span> | <span class="translation_missing" title="translation missing: en-US.lib.export.portability_exporter.activities.horton_values.flagged">Flagged</span> | <span class="translation_missing" title="translation missing: en-US.lib.export.portability_exporter.activities.horton_values.avg_elapsed_speed">Avg Elapsed Speed</span> | <span class="translation_missing" title="translation missing: en-US.lib.export.portability_exporter.activities.horton_values.dirt_distance">Dirt Distance</span> | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 159187244 | Jun 28, 2014, 4:48:31 PM | Horsetooth | Ride | NaN | 4641 | 14.67 | NaN | False | NaN | ... | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN |
| 1 | 159314142 | Jun 28, 2014, 10:41:53 PM | Antelope trail | Ride | NaN | 5455 | 8.79 | NaN | False | NaN | ... | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN |
| 2 | 160874659 | Jul 2, 2014, 4:07:30 PM | North Fruita Desert | Ride | NaN | 7846 | 21.89 | NaN | False | NaN | ... | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN |
| 3 | 160971473 | Jul 2, 2014, 8:00:21 PM | Horsethief loop | Ride | NaN | 5257 | 13.34 | NaN | False | NaN | ... | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN |
| 4 | 161316446 | Jul 3, 2014, 3:48:59 PM | The Whole Enchilada (before phone died) | Ride | NaN | 10706 | 16.55 | NaN | False | NaN | ... | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN |
5 rows × 80 columns
Format activity date column as datetime object
df['Activity Date'] = list(map(lambda x: datetime.strptime(x,'%b %d, %Y, %H:%M:%S %p'),df['Activity Date']))
df['Year'] = list(map(lambda x: x.year, df['Activity Date']))
df['Weekday'] = list(map(lambda x: x.weekday(), df['Activity Date'])) # weekdays labeled 0-6 meaning Monday-Sunday
daylist = ['Monday','Tuesday','Wednesday','Thursday','Friday','Saturday','Sunday']
df['weekday'] = list(map(lambda x: daylist[x], df['Weekday'])) # weekdays labeled Monday-Sunday
df['Distance mi'] = list(map(lambda x: x*0.621371, df['Distance']))
years = np.unique(np.array(df['Year']))
array = np.zeros((years.size,7,53))
Create activity arrays for visualization
yearcount = 0
for i in range(1,df['Activity Date'].size):
lastdate = df['Activity Date'][i-1]
date = df['Activity Date'][i]
x = (date.weekday(),date.week,date.year)
if date.year != lastdate.year:
yearcount+=1
if df['Activity Type'][i] == 'Run':
array[yearcount,date.weekday(),date.week%53] = 1
if df['Activity Type'][i] == 'Ride':
array[yearcount,date.weekday(),date.week%53] = 2
if df['Activity Type'][i] == 'Nordic Ski':
array[yearcount,date.weekday(),date.week%53] = 3
if df['Activity Type'][i] == 'Hike':
array[yearcount,date.weekday(),date.week%53] = 4
if df['Activity Type'][i] == 'Alpine Ski':
array[yearcount,date.weekday(),date.week%53] = 5
if df['Activity Type'][i] == 'Canoe':
array[yearcount,date.weekday(),date.week%53] = 6
Define visualization colors
none = 'black'
run = 'maroon'
ride = 'darkorange'
nordicski = 'darkcyan'
hike = 'yellow'
alpineski = 'azure'
canoe = 'midnightblue'
Create visualization
plt.style.use('fivethirtyeight')
fig, axs = plt.subplots(8,figsize=(24,28))
plt.subplots_adjust(hspace=0.1)
for j in range(0,years.size):
if j == years.size-1:
axs[j].set_xlabel('Week of the Year',fontsize=26, color = 'w')
if j == 0:
axs[j].set_title('Strava Activities',fontsize=26, color = 'w')
axs[j].set_facecolor('black')
axs[j].grid(False)
axs[j].set_ylabel(f"{years[j]}",fontsize=26, color = 'w')
fig.patch.set_facecolor('black')
activities = array[j]
X,Y = np.meshgrid(np.arange(activities.shape[1]), np.arange(activities.shape[0]))
colors = {0.0:none,1.0:run, 2.0:ride, 3.0:nordicski,
4.0:hike, 5.0:alpineski, 6.0:canoe}
axs[j].scatter(X.flatten(), abs(Y.flatten()-6), c=pd.Series(activities.flatten()).map(colors), s = 500)
axs[j].set_xlim(-1,53)
axs[j].set_ylim(-1,7)
axs[j].set_yticks(ticks = [6,5,4,3,2,1,0])
axs[j].set_yticklabels(['Monday','Tuesday','Wednesday','Thursday','Friday','Saturday','Sunday'],
fontsize=20, color = 'w')
axs[j].spines['top'].set_visible(False)
axs[j].spines['right'].set_visible(False)
axs[j].spines['bottom'].set_visible(False)
axs[j].spines['left'].set_visible(False)
axs[j].set_xticks(ticks = np.linspace(0,52,27))
axs[j].set_xticklabels(labels = (np.linspace(0,52,27,dtype=int)),fontsize=16, color = 'w')
custom_markers = [Line2D([0], [0], marker = "o", ms=22 , color=run, lw=0),
Line2D([0], [0], marker = "o", ms=22 , color=ride, lw=0),
Line2D([0], [0], marker = "o", ms=22 , color=hike, lw=0),
Line2D([0], [0], marker = "o", ms=22 , color=alpineski, lw=0),
Line2D([0], [0], marker = "o", ms=22 , color=nordicski, lw=0),
Line2D([0], [0], marker = "o", ms=22 , color=canoe, lw=0)]
plt.legend(custom_markers, ['Run', 'Ride', 'Hike','Alpine Ski','Nordic Ski','Canoe'],
loc=(0.03,7.82),fontsize=20,labelcolor='w',facecolor='black')
plt.show()
Additional visualizations
def group_and_count(colname):
count_df = df.groupby([colname]).count()['Activity ID']
count_df = count_df.sort_values(ascending=False)
x_labs = np.array(count_df.index)
y_vals = np.array(count_df)
return x_labs,y_vals
def charter(xlabs,y,title):
plt.figure(figsize=(10,6))
xpoints = np.linspace(start = 0, stop = len(y)-1,num = len(y))
plt.bar(x = xpoints, height = y, width = 0.9, tick_label = xlabs, color = 'teal')
plt.title(title)
plt.xticks(fontsize=16)
plt.grid(None,axis='x')
for count, yval in enumerate(y):
plt.annotate(text = f"{yval}", xy = (count-0.15,yval+5),fontsize=14)
plt.show()
results = group_and_count('Activity Type')
charter(results[0],results[1],'Activity Count by Type (2014-2021)')
results = group_and_count('Year')
charter(results[0],results[1],'Activity Count by Year (2014-2021)')
results = group_and_count('weekday')
charter(results[0],results[1],'Activity Count by Weekday (2014-2021)')
avg_dist = df.groupby(['Activity Type']).mean()['Distance mi'].sort_values(ascending=False)
activity_count_dict = df.groupby(['Activity Type']).count()['Activity ID'].to_dict() # dict for annotations
y = avg_dist
xlabs = avg_dist.index
plt.figure(figsize=(10,6))
xpoints = np.linspace(start = 0, stop = len(y)-1,num = len(y))
plt.bar(x = xpoints, height = y, width = 0.9, tick_label = xlabs, color = 'teal')
plt.title("Average Distance by Activity Type (2014-2021)")
plt.ylabel("Distance (mi)")
plt.xticks(fontsize=16)
plt.grid(None,axis='x')
for count, yval in enumerate(y):
plt.annotate(text = f"{yval:.2f}", xy = (count-0.15,yval+0.5),fontsize=14)
plt.annotate(text = f"n = {activity_count_dict[xlabs[count]]}", xy = (count-0.3,yval-1),fontsize=14)
plt.show()
def specific_df(activity):
specific_df = df.where(df['Activity Type']==activity,np.nan)
return specific_df.dropna(axis=0,how='all').reset_index()
ride_df = specific_df('Ride')
run_df = specific_df('Run')
ski_df = specific_df('Nordic Ski')
hike_df = specific_df('Hike')
plt.figure(figsize=(10,6))
plt.boxplot([ride_df['Distance mi'],run_df['Distance mi'],ski_df['Distance mi'],hike_df['Distance mi']],
labels = ['Ride','Run','Nordic Ski','Hike'])
plt.ylabel("Distance (mi)")
plt.title("Average Distance by Activity Type (2014-2021)")
plt.show()
plt.figure(figsize=(10,8))
plt.hist([ride_df['Distance mi'],run_df['Distance mi'],ski_df['Distance mi'],hike_df['Distance mi']],
histtype = 'step', alpha = 1.0, range = (0,60), bins=60,
label = ["Ride","Run",'Nordic Ski','Hike'],color = ['blue','red','teal','orange'],lw=1.0)
plt.xlim(-1,30)
plt.title("Distance Distribution by Activity Type (2014-2021)")
plt.xlabel("Distance (mi)")
plt.ylabel("Occurrences")
plt.legend()
plt.show()
plt.figure(figsize=(10,8))
plt.hist([ride_df['Distance mi'],run_df['Distance mi'],ski_df['Distance mi'],hike_df['Distance mi']],
histtype = 'step', alpha = 1.0, range = (0,60), bins=120, cumulative = True, density = True,
label = ["Ride","Run",'Nordic Ski','Hike'],color = ['blue','red','teal','orange'],lw=1.0)
plt.xlim(-1,20)
plt.title("Empirical Cumulative Distribution Function\nby Activity Type (2014-2021)")
plt.xlabel("Distance (mi)")
plt.ylabel("Probability")
plt.legend(loc='lower right')
plt.show()
Just for interest, I’ll create a simple model to calculate the probability that any of my rides is greater than a certain distance.
from statsmodels.distributions.empirical_distribution import ECDF
ecdf = ECDF(ride_df['Distance mi']) # create model with ride data
print(f"*Based on my personal riding history*")
print(f"Distance: 5 miles \t Probability of Riding Farther: {1-ecdf(5):.4f}")
print(f"Distance: 10 miles \t Probability of Riding Farther: {1-ecdf(10):.4f}")
print(f"Distance: 15 miles \t Probability of Riding Farther: {1-ecdf(15):.4f}")
print(f"Distance: 20 miles \t Probability of Riding Farther: {1-ecdf(20):.4f}")
print(f"Distance: 30 miles \t Probability of Riding Farther: {1-ecdf(30):.4f}")
*Based on my personal riding history*
Distance: 5 miles Probability of Riding Farther: 0.8468
Distance: 10 miles Probability of Riding Farther: 0.4393
Distance: 15 miles Probability of Riding Farther: 0.1908
Distance: 20 miles Probability of Riding Farther: 0.0838
Distance: 30 miles Probability of Riding Farther: 0.0260