Contrary to the previous observation. there is perhaps an interaction between the time of day of working and nonworking days. A day of week effect is not apparent, but we may need to look in more detail. This idea is easy to explore.
Task: Adding the following code creates box plots for peak demand hours of working and nonworking days. Complete the code to box plot the bike counts for isWorking - [0,1] for these hours: [0900, 1800]
Hint
labels = ["Boxplots of bike demand at 0900 \n\n", "Boxplots of bike demand at 1800 \n\n"]
times = [8, 17]
for lab, tms in zip(labels, times):
temp = bikesData[bikesData.hr == tms]
fig = plt.figure(figsize=(8, 6))
fig.clf()
ax = fig.gca()
temp.boxplot(column = ['cntDeTrended'], by = ['isWorking'], ax = ax)
plt.xlabel('')
plt.ylabel('Number of bikes')
plt.title(lab)
plt.show()
Observations: Now we clearly see what we were missing in the initial set plots. There is a difference in demand between working and nonworking days at peak demand hours.
*Action: *
isWorking along with hr feature may be good to train the model.
xformWorkHr which is a derived feature from hr and isWorking can also be used instead of two features.
Note - Having trouble with the assessment engine? Follow the steps listed here
Loading comments...