Adam Fletcher

PREDICTING THE RESALE PRICE OF USED MACHINERY: PROOF OF CONCEPT

The client wanted to determine factors that influenced the resale price of machines to better optimise listed prices.Used random forest regression to predict the price of an asset and how that price changes over time. The output of which was imported into Power BI via a bespoke API.Developed a pricing index to show market trends at variable resolutions, utilising regression to normalise auction sales to reference values.The business is now able to sell assets faster and at a more appropriate price. They are also able to predict depreciation price on new assets and factor in price of selling assets in project costings.

PREDICTING THE RESALE PRICE OF USED MACHINERY: WEB APP DELIVERY

Based on feedback, we changed both the model to a multidimensional non-linear regression model and the UI was changed from Power BI react to better incorporate new features.This enabled more explainability and continuity of predictions and ultimately more customisable model fitting.Worked with DevOps Engineers to develop data prep, model training and deployment pipelines in production.Liaising with Full-Stack developers to create Flask APIs using the regression models in different ways from single predictions, to predictions over the lifespan of an asset.Conducted a plethora of validation activities to build user-confidence in predictions as the industry was typically not data-driven.

MODELLING THE COST OF DEBT FOR DELAYED PAYMENT TERMS IN COMMERCIAL PROJECTS:

The client wanted to understand how the debt resulting from delayed payments terms affected final profit of a project it affects company liquidity.Using incoming and outgoing costs for each project I created a mixed-phase model for running projects, including the amount of debt accrued and how increased payment terms affected debt the cost of credit.With this information, the client restructured their pricing system and was better able to forecast their liquidity at key times of the year.