Machine Learning in Business Analytics: Advancing Statistical Methods for Data-Driven Innovation

Abstract

Machine learning has disrupted enterprises and business analytics. This has brought a shift from reliance on standard statistical techniques to a scientific approach where models depend on the available data. This paper assesses how ML packages could be used in diverse business settings with more focus on Random Forests and Neural Networks. The outlined models are assessed against linear and logistic regression models, and the distinction is made using accuracy, ROC curves, and precision-recall evaluative metrics. The study shows that even at 88% accuracy rate Neural Networks clearly outperform traditional methods and performing this task in the American business environment. It was also found that Random Forests can outperform 85% of the simple methods. The results also show that these metrics can be modified to achieve further efficiency. The application of those models showed improved performance metrics specifically in ROC and precision-recall curves. The study findings relevant to the domains of ML explain the effectiveness of combining batch size and learning rate optimally to achieve high accuracy rates in Neural Networks, e.g., 90%. Some suggestions for future work describe work needed to improve the explainability and ethics of the model while making it as usable as possible for businesses.