سعد صيدي نوره

The sales newspaper problem, considered a fundamental challenge in the economic market, aims to maximize profit by identifying the most profitable order quantities for products. This problem involves a decision maker who places inventory orders before the single-period sales season with uncertain demand. Excess inventory orders lead to leftover stock at the end of the period, while very low orders result in lost sales. The optimal order quantity for maximizing expected profit is widely known, but there is limited understanding of the decision-making process used by managers in practice. In this study, using the dataset from Horizon Kourosh chain stores, we focused on predicting sales quantities for two product groups: yogurt and milk. We executed five models, including multivariable linear regression, ridge and lasso regression, random forest regression, and gradient boosting regression. These models were eva‎luated using metrics such as coefficient of determination, mean absolute error, mean squared error, and root mean squared error. The random forest regression model performed best with an average absolute error of 94.86, a determination coefficient of 88.57%, and an average squared error of 22,382.59. For the yogurt product, the best performance was achieved by a model with an average absolute error of 151.858, a determination coefficient of 94.78%, and an average squared error of 76,254.952. This model was identified as the top performer.

مسئله روزنامه فروش , داده محور , خرده فروشي , يادگيري ماشين , رگرسيون

Newsvendor Problem , Data-Driven , retail , Machine Learning , regression

Saad Saidi

Mahdi Heydari