Comparing Machine Learning and Conventional Statistical Models for Business Decision-Making Precision: Proof from Predictive Analytics

Abstract

In the era of data-driven decision-making, selecting the appropriate analytical model is critical for improving business outcomes. This study aims to compare the effectiveness of machine learning (ML) techniques and conventional statistical models in terms of predictive accuracy, robustness, and decision-making precision. Using the Telco Customer Churn dataset comprising 7,043 observations, a quantitative comparative research design was adopted. Logistic Regression was used to represent traditional statistical modelling, while Decision Trees, Random Forests, and Support Vector Machines (SVM) represented machine learning approaches. Model performance was evaluated using Accuracy, Precision, Recall, F1-score, and RMSE, and a paired t-test was conducted to examine the statistical significance of performance differences. The results indicate that machine learning models outperform the traditional Logistic Regression model in predictive accuracy, with Random Forest achieving the highest accuracy (91.3%), followed by SVM (89.5%) and Decision Tree (84.9%), compared to Logistic Regression (78.6%). However, Logistic Regression demonstrated greater interpretability, providing clear insights through model coefficients. The study is limited to a single classification dataset and relies on secondary data, suggesting the need for cross-industry validation in future research. The findings highlight that while machine learning models are more suitable for complex predictive tasks, traditional statistical models remain valuable in contexts requiring transparency and explainability. This study contributes by offering a balanced framework for selecting appropriate analytical models in business decision-making.