Company

Projects

Company

Dŵr Cymru Welsh Water is a not-for-profit company which supplies drinking water and wastewater services to most of Wales and parts of western England that border Wales. In total, it serves around 1.4 million households and businesses, around three million people, and supplies nearly 830 million litres of drinking water per day.

Projects

• End to end machine learning pipelines for predictive modelling.
• Using data and building tools that solve problems and increase the analytical intelligence that goes into large scale decisioning.
• Working with stakeholders to identify problem areas and creating solutions to those problems.
• Following Agile methodologies and the CRISP-DM life cycle.

SRV Bacteria Predictor

An ensemble model with a Gini Coefficient of 70+ was developed to identify service reservoir assets at risk of ecoli failures to direct manual investigations. Water quality samples are regularly taken from each of the 300+ service reservoirs located across Wales. These samples are analysed and the data uploaded to a centralised database. A legacy model was historically used to identify assets but stakeholders felt its performance had degraded to the point of no longer being fit for purpose. This model redevelopment project included:

• Extensive EDA.
• Data drift analysis.
• 10, 000: 1 class imbalance.
• Low capacity for on-site investigation required minimal false positives.
• Interactive Power-BI Dashboard.
• SMOTE & TOMEK link resampling.
• Supervised Discritisation.
• Azure Virtual Machines & DevOps Pipelines.

Other Projects

• CSO Monitoring
• Risk Assessment Risk Index
• Property Risk Profile
• Model Monitoring Suite

Company

One of the UK's fastest growing independent finance companies, helping thousands of people fund the purchases of cars, vans, and bikes. Working with thousands of dealer partners across the UK, they've been in the industry for over 40 years and are now part of the Aldermore Group.

Projects

Building and monitoring credit risk models; predominantly logistic regressions and decision trees, and Customer analytics.

• Loan Application Models
• Behavioural Models (Roll, Recovery etc.)
• Customer Affordability Models
• Significant Increase in Credit Risk (SICR) (Regulatory)
• Propensity to Purchase GAP Insurance

Company

We Fight Any Claim was one of the UK’s most successful Claims Management Companies and proudly helped over 220,000 customers claim back over £600 million in compensation for mis-sold PPI. In August 2019, the Financial Conduct Authority introduced a deadline for making payment protection insurance complaints. From April 2020, We Fight Any Claim ceased claims management services and commenced an extensive exercise to review outstanding accounts and to exchange information with lenders to conclude remaining claim activity for existing customers.

Projects

• Monitored and forecasted business KPI's
• Used 'Google Analytics' to analyse website traffic and product performance.
• Created a data driven target setting system for telephony staff performance targets.