Top Secret - Machine Learning Project

More and more people are beginning to comprehend the opportunities that exist with machine learning and Artificial intelligence. Hunted Hive has been working on a very interesting private project for event prediction.

The objective of this project was to utilise current academic advances in state of the art machine learning algorithms and models to predict the outcome of an event occurring given past historical data. This can be applied to any statistical event where ample historical data is known. For example:

  • Sporting event
  • Stock market event
  • Weather event
  • Scientific event

 

How we delivered value

Hunted hive developed a machine learning framework which can be deployed locally or on an external cluster of servers or all of the above for a distributed cloud solution utilising both CPU and GPU processing power. The data was acquired and cleaned,  target variables were identified, feature engineering and evolutionary strategies were used to ensure the feature set was optimised for the prediction capability, then finally the machine learning models were developed and used to make accurate predictions. 

The framework can be used to make predictions in short periods of time while remaining accurate and private. The Customer is happy with the result and we continue to provide our services to constantly improve and provide more value.

Our Role

Our primary roles included:

  • Data exploration with our in-house data scientists
  • Research and development of the most appropriate theory to be applied
  • Research and development of the most appropriate technology architecture to be applied
  • Feature engineering to ensure optimal features were being used
  • Optimisation of framework given hardware constraints
  • Regular reporting on progress
  • Reporting framework to display analytics and sights on simulation data

 

A flexible framework to crunch some serious numbers - Roger Chappel, Data Scientist
View Website