OS Maps: Automating star ratings with ML solutions
About the product
OS Maps is a mobile and web application that allows users to access standard and green space maps in the UK. The maps are available online and offline by downloading them. The main functionality of OS Maps is related to routes – users can follow thousands of ready-made routes or create their own trails, follow them, and then record them. Routes are recommended based on their user ratings so that the best routes with the highest scores are promoted in the app.
Challenges and business needs
Wanting to meet user expectations, the Ordnance Survey identified the need to implement a machine learning solution to automate the process of assigning star ratings to newly created routes. The solution aimed to facilitate the determination of each route’s quality so that highly reviewed, new trails could be promoted in the app, just like those already rated by users.
Our responsibilities
As part of the cooperation, the Spyrosoft team took responsibility for the comprehensive implementation of the new machine learning-based functionality for the OS Maps application. The scope of our activities included conducting an exploratory analysis of data related to routes, such as name, description, creator, location, points, elevation, and more.
In-depth analysis of existing data allowed us to develop and build machine learning models tailored to predict route ratings based on historical data. They use advanced algorithms to learn patterns from existing route evaluations and extract insights that help predict the quality of new routes with a high degree of accuracy. Furthermore, to ensure the reliability and accuracy of our machine learning models, we conducted a detailed validation process.
Results
The implementation of machine learning for route rating prediction in OS Maps has enabled a method for assessing and promoting new trails within the application. By automatically assigning star ratings to newly created routes, OS Maps can offer users a curated selection of high-quality path suggestions. This approach intends to encourage users to explore various new trails while providing a more personalised user experience.
Contact us