Artificial intelligence-supported cost management: an analysis with statistical and machine learning models

The modern business world is undergoing continuous transformation due to technological advancements and globalization. Increasing competition, particularly in the manufacturing sector, compels businesses to adopt new methods to reduce costs, utilize resources more efficiently, and quickly adapt to customer demands. Traditional management and production planning approaches may be insufficient in simultaneously addressing all these complex and dynamic needs. In this context, modern methods such as artificial intelligence and data analytics enable businesses to optimize costs while enhancing operational efficiency. This thesis examines the integration of AI-supported data analytics and cost optimization. The study analyzes the differences between multiple linear regression and machine learning models for sales and cost forecasting of the X product used in the automotive industry. It also explores how specific variables influence sales and cost predictions. The findings demonstrate how artificial intelligence and data analytics technologies ca The multiple linear regression analysis identified key independent variables affecting sales volumes, including the Number of Motor Vehicles (NMV), the Industrial Production Index (IPI), Consumer Price Expectation (CPE), and Wage Change Expectation (WCE). The Adjusted R² value of the sales volume model was 0.916, indicating high accuracy in explaining sales quantities. Meanwhile, the Adjusted R² value for the cost estimation model was calculated as 0.974, demonstrating its strong explanatory power. These results emphasize the critical role of certain variables in sales and cost predictions and highlight the importance of businesses monitoring these factors closely. Further analyses using Gradient Boosted Regression Trees (GBRT) and Artificial Neural Networks (ANN) showed that ANN models provided the best performance for both dependent variables. ANN models successfully captured nonlinear relationships, yielding more accurate predictions with lower error rates and higher precision than traditional statistical methods. Specifically, the R² value for X product sales forecasts was 0.98, with a Mean Absolute Percentage Error (MAPE) of 0.08, while for cost forecasts, the R² value was 0.997, with a MAPE of 0.02. The multiple linear regression model results reported an Adjusted R² value of 0.916 for X product sales and 0.974 for X product costs. When comparing these findings, machine learning models demonstrated superior predictive capabilities over traditional statistical methods, contributing more effectively to business decision-making processes. The results revealed that NMV and IPI significantly impact sales and cost forecasts. In this regard, businesses should regularly monitor these variables in their production and cost planning to develop more accurate and efficient strategies. In conclusion, this thesis demonstrates how both traditional statistical methods and modern machine learning approaches can be effectively utilized in cost optimization processes. AI and machine learning-based models provide more reliable predictions, supporting businesses' strategic decision-making processes. The findings help companies to make more accurate forecasts, enabling efficient planning and gaining a competitive advantage.