Előadás címe: Improving inventory management using innovative market demand prediction
Helyszín: Online https://meet.google.com/pfp-yzpu-cbe?authuser=0
Időpont: 2020.06.30., 18:10 – 18:30
The goal of the presented project is to provide state-of-the-art prediction methodologies, using advanced machine learning technology, that can be used to predict demand of certain goods for a foreseeable amount of time. Such predictions can serve as input for inventory planning strategies of retail companies.
The business model of the industrial partner is that certain products are kept on inventory to ensure high service rate of demand for them. Planning the inventory level for these products rely heavily on a prediction about the future demand for them.
The complex nature of the real-life dataset requires machine learning methods to be tailored to incorporate and utilize this complexity. To mention some of the complicating factors:
- there are items that are always sold together (complementary parts of a whole installation system, like radiators and tubes),
- there are items that can replace each other and
- items whose sales are linked for some external reason (e.g. weather condition enabling construction).
The expected results of the research are criteria based on which attribute selection and prediction algorithms can be evaluated in relation to demand prediction in the presented setup. Based on the type of data available for the task, and their temporal behavior, certain algorithms are expected to perform better whilst others less so. We aim at identifying quantifiable criteria for discriminating between these.