What Would Make YOU Use a London Bike Share?

An overview of the London Bike Share usage from 2015 and 2017; the important factors and what it would mean for the future.

Junyi

Published in

The Startup

7 min readNov 9, 2020

Introduction

London, with a population of over 9 million¹ people, it’s a city of its own. To a city of this magnitude, transportation is a vital component. The London Underground as well as the red double decker buses have no doubt, become icons of London.

London Bike Share, (or more popularly known as Boris Bikes or Santander Bikes²) is a younger icon of London. It was launched on 30th July 2010³ and we have just celebrated its 10th birthday. There are more than 750 docking stations and 11,500 bikes in circulation across London⁴, making it one of the largest bike share schemes in Europe.

Living in London, I enjoy walking from places to places, if I have to get to somewhere afar, we are always just a short walk away from a tube/bus stop. Therefore, I have never really paid much attention to the bike share schemes. However, the scheme has been going for over 10 years now, and regardless of my personal opinions, it must be important (or at least useful) to others.

Table 1. Bike shares in London per day between 2015 and 2017

Between the year of 2015 and 2017, on average, each day, there were over 27 thousand times a bike was used, with over 72 thousand times on the busiest day — that’s a lot of journeys!!

The above results comes from a dataset⁵ from Kaggle (with the raw source containing TFL data), I will be using this dataset to try to provide an unbiased macro-view of the usage as well as trying to predict bike usages for the future. I will be mainly using this dataset as it contains hourly London share bike usage between 2015 and 2017, as well as information such as weather, temperature, holidays season etc. It is perfect for providing a good overview for the London bikes.

Part I: How does the bike share usage vary through out the years?

My first thought was to look into the general trend to see if people are using the bikes throughout the year, and when?

It can been seen that between 2015 and 2017, there is a clear trend in overall seasonality, where there are generally more bikes being used in the times around August than the times around January-February. This fits the common sense that as the weather becomes warmer and drier, more people would use bikes (and vice versa).

Graph 1. Daily as well as 14 Day rolling average of the bike usage between 2015 and 2017. There are 2 possible anomalies detected.

However, the blue line that shows daily movements indicates regardless of the season, there seems to be another trend at a daily level. Also there are 2 points on the graph at 9th of July 2015 and 6th of August 2015 that are much higher than seasonal trend (even taking into account the oscillation of daily trends). These will be discussed further.

Part II: Do seasons affect bike usage?

First things first, does the data provide quantitative evidence that seasons affects bike usage? — My guess is YES! What is your guess?

Graph 2. Shows the count of bike shares by season, weather condition, and temperature (to the nearest 2 degrees)

To a certain extent, seasons (by extension, temperature and weather too) do correlate with the bike usage, however, the correlation seems to be somewhat weaker than I had expected as shown in the plots above. Whilst the averages do trend in the direction as I thought they would, however, the range of the values are far bigger than I had anticipated.

Interesting… so does that mean a lot of people use the bikes regardless of the weather? Hmmm… (the plot thickens…)

Graph 3. Plot weekends and weekday against season.

From graph 3, we can see bike usages at weekends are usually lower than the 14 day rolling average for cooler seasons due to many weekends (green plots) are below the rolling average (orange) line. However, they are much closer (if not higher) than the rolling average for warmer seasons! This means many people use the bikes for work commutes, but would use them more for leisure as weather gets better! This would explain why season, weather and temperature only have small correlation with bike usage: there is a more important factor at play here! — WORK!

Part III: How does the trend differ between weekdays and weekends?

Graph 4. Weekends vs weekdays hourly bike usage comparison.

By breaking it down by the hour and looking at weekends vs weekdays, we can see very distinct patterns between weekends and weekdays. For weekdays, it appears that most bikes are being used at around 8am and 5 (and 6) pm; these coincide perfectly with the rush hours when most people go to and from work. For weekends, the pattern is more smooth, usage peak at around early afternoon (perhaps due to more people are out and about around these times?).

Regardless, it is obvious that many people use the bikes for work commutes during weekdays as seen by the sharp contrasts between the two.

Part IV: How well can the bike usage be predicted?

Lastly, I wanted to see if we could predict the future bike usage using features (information) such as hour of the day, temperature, season etc. So I ran the data through a machine learning model (here are my workings in code for those that are interested) and here are the results.

Graph 5. It shows that given the features (information) provided, the general trend for 7 day rolling averages fits the true values.

Overall, it seems that we can make a good high level prediction by predicting by hour, then averaging over 7 days using factors provided such as time of the day, temperature, weekday/weekend etc.

On the Anomalies:

Anomalies! — Don’t worry, I have not forgotten about them!

There is one more thing we have yet to discuss, on graph 1, we have identified 2 anomalies that showed substantial bike usage. On the 9th of July and 6th of August 2015, over 72K and 63K times the bikes were used respectively, when the average from the dataset is just over 27K. Even taking into account that they had happened in the middle of the summer, the values were still much higher than expected (around 5 the standard deviations over the mean!).

Can you guess what it is??

TUBE STRIKES! By looking at what happened in London on those days, I found out that on both 9th of July 2015 ⁶ and 6th of August 2015 ⁷, there were tube strikes happening on each of those days. This also provides evidence that a significant proportion of bike usage was due to work commute, hence why a tube strike on a working day would mean many people had to find alternatives to work. Therefore, causing a significant increase of bike usage.

I bet the bus usage on those 2 dates had also gone through the roof!

Conclusion:

In this article, we provided an overview of the London Bike Share usage from 2015 and 2017.

We looked at the general trend of the bike usage and found a few interesting factors that drive the usage:

Warm weathers definitely encourage the use of bikes as data showed more bikes were used summer than winter.
There is a distinctly different usage patterns for weekdays and weekends as shown from the hourly breakdown graphs. This proves that a significant proportion of bikes are being used for work related travels. This hypothesis is further strengthened by the discover of anomalies caused by Tube Strikes.
The factors discussed (as well as other factors from the dataset) were used to provide a good macro-prediction of the bike usages.

To go even further, if I could separate bike usages into work commutes and leisure/sightseeing, how differently would hour of the day and weather condition affect these? I bet that would provide some interesting insights!

What conclusions will YOU draw?

Final Thoughts

This article was written in early November 2020, just as London is starting its 2nd lockdown due to Corona Virus. Covid-19 has changed many things including the way we work.

As social distancing takes place, tubes are not as busy as before, however, less people are expected to go to work as well. Therefore, would there be a higher or a lower demand for these London Bikes?

Again, going back to the anomalies caused by Tube Strikes, do they provide evidence that many people do need to be physically present at work? Otherwise, if people could, why would many chose to cycle, rather than simply working from home for a day?

Are we just waiting for this temporary setback to pass and we will be back to the old normal with Covid being nothing but a chapter in our history books? Or, are we experiencing a pivotal moment in our history and time where we are heading into a whole new direction?