MACHINE LEARNING FOR CONCRETE SUSTAINABILITY IMPROVEMENT: SMART FLEET MANAGEMENT

Abstract

In the dynamic landscape of modern business operations, ensuring economic security through efficient and intelligent fleet management is imperative. This necessitates a dual focus on safeguarding revenue streams and optimizing operational costs. The aim of this study centers on two main objectives: first, to identify driving behaviors that have a substantial impact on vehicle maintenance costs; second, to ensure the sustainability of the fleet is managed effectively. To achieve these objectives, the research employs unsupervised Machine Learning (ML) techniques for segmenting driving styles based on diverse parameters collected from Internet of Things (IoT) devices. Furthermore, the Long Short-Term Memory (LSTM) algorithm is used for forecasting fuel consumption, offering a predictive glance into future expenditures. The methodology is based on the analysis of data gathered from sensors installed on the vehicle's Controller Area Network (CAN), collected over a span of five months. The findings spotlight a subset of drivers whose aggressive driving significantly influences maintenance costs and highlight optimal indicators for drivers to monitor to minimize CO2 emissions. Additionally, the study identifies key performance indicators that drivers should monitor to reduce CO2 emissions, contributing to the environmental sustainability of the fleet. This investigation not only elucidates the financial and environmental implications of driving behaviors but also showcases the transformative potential of ML technologies in enhancing the strategic management of vehicle fleets. Through this exploration, the research advocates for the integration of advanced analytics and sustainable practices as foundational elements for businesses striving to achieve economic security and operational resilience.

Keywords

• driver behavior
• fleet sustainability
• unsupervised machine learning
• clustering analysis
• co2 forecasting