Project Tokyo – Blending Machine Learning and Human Expertise

Location: Adelaide, SA

Duration: 5 months

Proposed start date: 12 August 2019

Project Background

Data driven models for time series forecasting are proving to give reasonable predictions and forecasts. In combination with expert human judgement we expect that time sensitive and expert knowledge is critical in terms of influencing a final prediction.

Project Tokyo is an innovative blend of Machine Learning and Human Expertise which aims to bring together the best of both to enable unprecedented accuracy in sales and production forecasting.

The leap in capability is achieved by marrying machine learning experts’ knowledge and insights to bring predictions to a super-forecaster level of accuracy, timeliness, and trustworthiness.

Project Tokyo aims to take into consideration specific human configurable parameters that an operator can tune to provide a final prediction that is the result of combining a data driven model and an expert human judgement (biological neural network model).

Research to be Conducted

Research and design a consistent way of modelling human configurable human input parameters in combination with robust data-driven machine learning models for time series forecasting.

For example, a pharmaceutical company may have a historical sales dataset which can drive the data driven sales forecasting model. However, the data driven model does not take into account recent temporal events that would influence the outcome of sales, e.g. tweets from politicians or an outbreak of swine flu.

Objective 1 – Single tuneable risk slider.

Objective 2 – A configurable set of risk sliders each with configurable weightings to influence the model.

Skills Required

We are looking for a PhD student with the following:

ESSENTIAL

  • Python Programming
  • Strong foundation in Mathematics and Statistics
  • Machine Learning experience

DESIRABLE

  • Jupyter notebooks
  • Angular

Expected Outcomes

  1. Demonstrate an algorithmic combination of machine learning output and human tuning based on a judgement of risk.
  2. Demonstrate an algorithmic combination of machine learning output and human tuning based on a set of configurable parameters with configurable parameters.
  3. Conduct a case study with a real customer and real historical sales data.

Additional Details

The intern will receive $3,000 per month of the internship, usually in the form of stipend payments.

It is expected that the intern will primarily undertake this research project during regular business hours, spending at least 80% of their time on-site with the industry partner.  The intern will be expected to maintain contact with their academic mentor throughout the internship either through face-to-face or phone meetings as appropriate.

The intern and their academic mentor will have the opportunity to negotiate the project’s scope, milestones and timeline during the project planning stage.

Applications Close

 24 July 2019

Reference

APR – 1023