E-Commerce Product Demand Modelling Using Machine Learning Algorithm Case Study of Rice Trading Products in PT XYZ

E-commerce XYZ is an Indonesian commerce company that have 3 types of products in its B2C business line: trading, consignment, and marketplace. From January 2021 until October 2022, the company’s trading rice category product sales generated a negative profit. Even though for the last several years e-commerce has been focused on growth instead of profitability, the current economic environment is forcing e-commerce companies to focus on profitability as well. For trading products, maximum profit can be achieved in two ways: selling products with a very high margin but with less quantities or selling in large quantities but with a sub-optimal margin. Hence, the company needs to find a demand function model that can be used to generate maximum profit. To find the best model, the researcher first created a baseline model by using median for every product group which is already grouped based on their Unit of Measurement. Next, to find the best model, the researcher will create a demand function using 4 other models. It is found that Gradient Boosted is the best algorithm to model the demand function. Although this model successfully models a demand function for a product category in e-commerce, business context still needs to be added before this model can be implemented in real life as well as finding other features that might affect the demand function.