DEEP LEARNING (DL) BASED APPLICATION FOR FRESH PRODUCE YIELD AND PRICE FORECASTING BASED ON SATELLITE IMAGES AND STATION-BASED PARAMETERS

Loblaw Companies Limited (LCL) supplies all fresh produce (FP) to South Western Ontario stores from Waterloo Distribution Center (DC). DC decides prices and quantities to meet FP demand. Timed fair priced orders minimize waste, bring prosperity to growers, consumers and FP trades. Factors affecting prices are highly uncertain due to environmental and socio-economic effects such as income, labor, trade, globalization and climate change which makes price prediction challenging. Immediate produce past prices are used to predict future prices in univariant models, other multivariant models consider price most influential factors as input attributes to predict future prices. Machine learning models deployed for price forecasting can currently be outperformed by univariant models. In this project multivariant Machine learning models finetuned by deploying online learning are tested against mathematical univariant models and their effectiveness is assessed using accuracy measures; micro cent improvement in each transaction save hundreds of millions of dollars for Canada.

Faculty Supervisor:

Fakhri Karray

Student:

Lobna Nassar

Partner:

Loblaw Company Limited

Discipline:

Engineering - computer / electrical

Sector:

Service industry

University:

University of Waterloo