Utilizing Deep Neural Networks for Forecasting Prices of Meat-Based Products

Description

The impact of price adjustments on the sales performance of food products has been reported worldwide. This study aims to define how price predictions assist price decision-makers in formulating optimal prices and, subsequently, forecasting sales performance. To achieve this aim, we employed predictive deep neural networks (DNN) as a machine learning (ML) model, specifically designed for predicting meat-based product prices. A dataset from a leading food retailer in a Nordic European country records almost six hundred thousand sales performances associated, particularly with meat-based food products. Having looked through this dataset, three primary input categories are introduced: 1) demographic data, 2) product details, and 3) sales information. The deep predictor detects non-linear and complex relationships between these input categories and product prices, ultimately helping in system precision enhancement. The result of the predictive DNN model can be used as a Business Intelligence (BI) supportive tool, indicating that this predictor performed accurately with lower error rates (0.8731) and high explanatory power (0.9484). Finally, an R-squared (R²) of 0.9484% indicates that approximately 0.9484% of the variance in product prices is explained by inputs.

Period	15 Aug 2024
Held at	Nordic Academy of Management (Nordiska företagsekonomiska föreningen, NFF)
Degree of Recognition	International