Keeping Prices Cool: How Machine Learning Transforms Refrigerator Pricing Strategies

The demand prediction of product pricing is in retail and a key challenge to balancing profitability and market competitiveness. This study explores the potential of machine learning algorithms for predicting refrigerator prices in the household appliance industry, filling a gap between classical statistical approaches and sophisticated regression models. Existing literature reflects a trend towards algorithmic price modelling, yet few specifically focus on the consumer durables sector, with studies observing mostly real estate or commodities rather than differentiated retail products. The research uses eight different machine learning methods to analyze a rich dataset of refrigerators from tree-based ensemble methods to support vector machines and neural networks. The models use the various features of a product, including technical specifications (volume, energy efficiency), aesthetics (colour, design), and brand positioning to predict market prices. Results show that ensemble models have a significantly better prediction power than traditional approaches, with the Random Forest algorithm achieving significantly lower error metrics than simple regression models. Clustering analysis shows that algorithms have different performance tiers, with tree-based methods being the highest-performing cluster. Finally, our findings provide meaningful implications for the development of retail pricing strategies by identifying economically significant product attributes in various market segments. The study’s core contribution is in building a methodological framework for well-defined price prediction in markets with differentiated products that support equity pricing in retail and can be translated across product categories to change the way pricing strategies are set beyond the household appliance industry.