Background:
Project Geminae redefines how businesses view and solve their portfolio optimization problems with a current focus on oil and gas production. The project involves the development of a new tool utilising machine learning to predict oil, gas, and water production based on historical data.
Before:
- Model performance was inadequate, failing to meet customer expectations and contractual obligations.
- The existing model was unable to make accurate predictions for oil, gas, and water production.
- Lack of model interpretability hindered user trust and understanding.
- The system lacked a user-friendly visualization tool for explaining model predictions.
After:
- Significantly improved model accuracy: predictions now closely match actual production values for oil, water, and gas.
- The improved model demonstrated the potential to increase earnings by over a billion dollars over a 10-year period.
- Successfully implemented a visualization tool for explaining model predictions, increasing user understanding and trust.
- The new model is interpretable and traceable, allowing users to understand the logic behind predictions.
- Enhancements in data quality and preprocessing contribute to the model’s improved accuracy.
- Integrated accurate predictions for oil, water, and gas production into a single, comprehensive model.
- Increased customer satisfaction due to improved reliability and value of predictive services.
Process:
- Comprehensive review and enhancement of the model’s architecture, algorithms, and performance, as well as a thorough analysis of the underlying data.
- Development of new features to increase model interpretability and user-friendliness, including a visualization tool for explaining model predictions.
- Enhancement of the machine learning model to improve accuracy and reliability in predicting oil, water, and gas production.
- Code optimization for better efficiency and maintainability.
- Provision of expert suggestions for further model improvements and implementation of best practices in energy sector predictive modeling.
- Documentation and knowledge transfer to ensure the client’s team could maintain and further develop the enhanced model.
The improved model demonstrated the potential to increase earnings by over a billion dollars over a 10-year period.
