1. Summary visualization recap
With just 3 summary plots, we've already learned a great deal about our data. Let's quickly recap some insights and get ready to dig a bit deeper.
2. Daily rides
From our simple first plot of daily rides with points colored by weekday or weekend, we learned that
1- It appears that bikes are only rented from mid-April to mid-November.
2- Ridership peaks in the summer, around July or August.
3- The number of rides doesn't appear to be very differentiated on weekends vs. weekdays.
3. Hourly rides over time
Breaking the data up weekly by hour of day and weekday/weekend gives us some additional insights.
1- Regardless of day of week, there is a strong diurnal component, as we would expect.
2- There are clear differences between weekdays and weekends. Weekend ridership on average follows a unimodal pattern, peaking around 3pm, while workweek ridership is bimodal with peaks at 8am and 5pm.
3- In the very early hours of the day, weekend and weekday ridership follows roughly the same pattern of number of rides, even though there are 5 days in the workweek and 2 days in the weekend.
4. Hourly rides over time + membership
Adding the dimension of membership helps us realize that the number of rides from non-members is roughly the same for the workweek and weekend, and the bimodal workday pattern is only present with members, suggesting that non-members are probably mostly tourists or using bikes for leisure, and members are probably more likely to be commuters.
5. Diving deeper
All of this insight would play a major role in how we would approach building a model to predict the number of rides at any given time. We know that time of day, day of week, and time of year all have an effect on ridership, as does whether the rider is a member or not. We may have been able to guess beforehand that these would all be important variables to include in a model, but our insights from visualization help us better understand *how* a model can best leverage these variables. Now that we have some interesting high-level insights, let's start diving a little deeper.