Ernestynne Walsh

Challenge: Over 2 million personal injury claims are manually processed by staff each year. Notification of these claims being accepted is done via postal mail and is too slow.

Solution: Building two services that utilise a series of predictive models to automate this process so that text alerts and emails can notify clients of cover within seconds of the claim being submitted.

Value: Clients will know sooner that their claims will be covered and can focus on rehabilitation rather than worrying about paying for their treatment.

Challenge: Limited number of people with cross sector evaluation experience who can use the IDI.

Solution: Standardised processing code to use in the IDI that is publicly available on Github - https://github.com/nz-social-investment-agency. Also worked on developing use cases to illustrate how an evaluation could be carried out using such tools.

Value: Each project that uses these tools saves over 6 weeks in time to get a dataset ready for analysis. Currently six other organisations are making use of this code and the tools.

Challenge: Shortage of people with statistics and machine learning skills.

Solution: SAS training programs for which I teach the open source courses, machine learning and more on the cloud platform SAS Viya in New Zealand and Australia.

Value: More skilled analysts in the workforce. Also, I add additional topics to the course material like ethics, explainable algorithms and how the tools could be used at work.

Challenge: Systems focused view on child welfare that operated in silos.

Solution: A client centred view that consisted of several predictive models relating to safety, behaviour and future earning ability of a child using SAS EG and SAS Enterprise Miner.

Value: Parts of models were used as input to conduct a Randomised Control Trial at the Contact Centre. Models help construct more robust quantitative investment cases to help allocate resources to groups of children where interventions have been successful.

Challenge: Asset replacement is not done in the most efficient manner. For example roads are replaced then shortly after the road is removed to replace the pipes beneath.

Solution: Run a network simulation using igraph that forecasts growth in suburbs and identifies future bottlenecks in the 3 waters network. Overlay this with forecasts of when roads need replacing to project the optimised depreciated replacement costs over the next 100 years complete with heatmap and cost visualisation.

Value: Enables a work programme to be created that replaces assets in close proximity at a similar time to reduce cost.