The Problem

UK public transport is difficult to navigate, due to having to use multiple different apps just to view schedules and plan routes.

I built this app, CommuteClever, to connect between the train and bus APIs, and provide live times and notifications.

The live website is not available due to the cost of maintaining the API connections, however there is a YouTube video below showing the system working instead!

App features:

• Live times: Down to 30s accurate bus and train times, pulled directly from their APIs, and linked with the schedule data.
• Schedules: Schedules pulled directly from the transport APIs, and cached in a database locally for quick access.
• Route Planning: Using Djikstra’s algorithm, routes can be planned easily between multiple modes of public transport.
• Notifications: Time to leave, delay and cancellation notifications via SMS in real-time.
• Delay Predictions: Filling in the gaps still left by the railway APIs, I could predict delays on the trains before they happened. For example, the arriving service is delayed 10 mins, meaning the next service will be delayed by the same amount.

Challenges

• Managing a large database: The amount of schedule data was gigabytes in size. I re-formatted this using Python into an SQL databse, allowing it to be accessed much quicker than by iterating through SOAP and JSON files. However, I ran into issues with concurrency and poor read speeds with SQL - so in future I would consider a database able to handle larger files.
• Optimised System: Flask was resource intensive and slow while running. In future, a compiled language would be more suitable, as Python is inherently slow for web applications. I would build an app next time where the client is directly connected to the API feeds. This would also reduce the load for the backend database.