London operates a bike rental service under its Transport for London (TfL) network, where citizens can participate in a bike-sharing system and rent out bikes for transport around the city. TfL has collected detailed historical data on this service and its users and posted it publicly on Kaggle. In this scenario we want to examine the data on this service and its users between the years 2015 and 2016. Specifically, we want to build a dashboard to visualize the historical data relating the number of bike rides to temperature, humidity, wind speed, weather, and time of year. The questions we want to answer are:
What times of year are the most popular to rent a bike?
What impact does the weather have on the number of bike rides?
What times of day are citizens most likely to rent a bike?
METHODS
The open-access London bike sharing dataset was downloaded from the website Kaggle, containing bike rental data from the TfL service from January 1, 2015 to December 31, 2016. Using Python, the dataset was then imported from a CSV file into a Pandas DataFrame to conduct exploratory analysis and data cleaning.
I examined the dataset's structure and content, including the number of rows and columns. Unique values in the 'weather_code' and 'season' columns were counted. I then renamed several columns for clarity, such as changing 'timestamp' to 'time' and 'cnt' to 'count'. Additionally, the humidity values were normalized to a scale between 0 and 1, and numerical values were mapped to descriptive labels for seasons and weather conditions. Finally, the modified DataFrame was saved as an Excel file for use in Tableau visualizations.
In Tableau, the dashboard features a total count of bike rides, a timeline showing the moving average of bike rides, and a heatmap showing the impact of temperature and wind speed on bike rides. Interacting with the charts displays a tooltip giving additional data on the bike rides for weather and time. Dragging and holding across a section of the timeline filters the total count and heatmap on the selected range. Additional filters can adjust the moving average's duration and period, and a timeline filter allows for customization of the displayed time period.
RESULTS
Most Popular Bike Rental Times
Summer has the highest total count of bike rides of all seasons, with over 6 million total rides for 2015 and 2016 combined. Fall has the second highest total count with just over 5 million rides, Spring is third with just below 5 million rides, and Winter has the lowest with below 4 million rides.
In 2015, the average count of bike rides are at their highest from July to August, averaging between 32,000 to 35,000 per day, or 185,000 to 205,000 per week.
In 2016, the average count of bike rides are at their highest from mid-July to mid-October, averaging 32,00 to 39,000 per day, or 185,000 to 225,000. per week.
Best Weather for Bike Rentals
The top 3 weather conditions with the highest total bike rentals during the 2-year period are ranked are clear skies at over 7 million rides, scattered clouds at over 6 million rides, and broken clouds at over 4 million rides. Understandably, weather conditions with more hazardous weather such as rain, snowfall, and thunderstorms have significantly lower total bike rides.
During the most popular bike rental periods of Summer and Fall, the most frequented weather conditions with the highest average bike rentals alternate between clear skies and scattered clouds.
When comparing temperature to wind speed over the 2-year period, bikers tend to rent bikes during wind speeds between 10.7 and 17.9 Kph, with a temperature between 5.0 and 19.9 C.
During the most popular bike rental periods of Summer and Fall, the preferred temperatures stay between 14.9 C and 24.9 C with wind speeds between 7.1 and 17.9 Kph.
Best Time of Day for Bike-riding
The times of day with the highest total and average bike rentals over the 2-year period are at 8 AM and 5-6 PM, with 8 AM and 5 PM having 2 million total rides each and 6 PM have just below 2 million total rides.
Average bike rentals decrease somewhat after 8 AM and increase as 5 PM approaches. The average bike rentals significantly decrease after 5PM and are at they're lowest around 2 to 5 AM
SUMMARY
The analysis of the London Bike Rides dataset and visualization reveals several key insights into bike rental patterns in London. Regarding the time of year, summer emerges as the most popular season for bike rentals, with the highest total count of rides, indicating a strong seasonal influence on biking habits. This preference for warmer months is further supported by the dataset showing that clear skies and scattered clouds have the highest total count of bike rides. These findings suggest that clear and mildly cloudy days significantly boost bike rental activity.
In terms of daily patterns, 8 AM is identified as the peak time for bike rentals, both in total and average counts, likely reflecting morning work commutes. The late afternoon and early evening hours, particularly 5 PM and 6 PM, also see high rental activity, reflecting with the evening work commutes. Interestingly, this indicates a direct relationship between weather and biking trends; milder, clearer weather conditions tend to encourage more bike usage.
The data also indicates the biking is at its highest during times of light and mild wind speeds, as well as cool to warm temperatures. However, this does not provide explicit insights into how wind speed and temperature independently affect bike rentals, since these patterns of wind and temperature are common during the Summer and Fall seasons. This suggests a potential area for further detailed analysis to understand these specific impacts. Overall, the dataset offers valuable insights into how seasonal, weather, and time-of-day factors influence bike rental patterns in London, which could be crucial for urban planning, resource allocation, and promoting sustainable transportation in the city.
Below is the Python script I used to conduct exploratory analysis and data cleaning:
Click on the GitHub and Tableau icons below to view both the Python script and interactive Tableau visualizations respectively.
Comments