Using Constrained Optimisations in Hierarchical Time Series Forecasting

Project Level: Winter

Project Duration:

4 weeks, Full time 36 hours per week. Applicant can work online or on-campus.

Description:

Time series forecasting plays a crucial role in many business decision makings. Whether it is sales, temperature, or price, we want to know what will happen in the future and accordingly make decisions today. This project is about forecasting a set of sales time series. Data has hierarchical format, meaning that there is a hierarchical structure in data. For example, consider hierarchical sales data for a product in which we have total sales for Australia at the top level, sales for each state at the middle level, and sales at each store at the bottom level. There is a logical structure that needs to be considered.

In this project, you will aim to use linear and mixed integer optimisation methods to develop an algorithm for forecasting hierarchical sales data.

Expected Outcomes:

In this project you will have the opportunity to learn state-of-the-art forecasting models, data science modelling, data science communication, and your thinking skills. There is an opportunity present in national or international conferences and write article for top-tier international journals.

Suitable for:

This project is open to applications from postgraduate students with a background in data science/mathematics/machine learning and closely related fields. It is essential to be familiar with linear and mixed integer programming.

Further Information:

Please contact Dr Mahdi Abolghasemi and send your CV/Transcript, and ideally one reference before applying for this project.

It is not compulsory to know time series, but it would be nice if you are familiar with it.