Optimizing Autonomous Systems through Reinforcement Learning: The Role of Linear Regression, Random Forest, and Support Vector Machines in Decision Making
International Journal of Computer Science and Data Engineering · 2025
DOI: 10.55124/csdb.v2i2.249 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Computer Science and Data Engineering, 2025. Verfasst von Akhilesh Reddy Eppa.
Abstract (Originalsprache)
Reinforcement Learning (RL), a powerful paradigm for decision-making in autonomous systems, has enabled agents to learn the optimum policies through trial and error in dynamic environments. This paper explores the integration of RL within autonomous systems, emphasizing how various machine learning algorithms Linear Regression (LR), Random Forest Regression (RFR), and Support Vector Machines (SVM) can assist in predicting agent performance and optimizing training processes. These algorithms are employed to analyze key input parameters, including sensor accuracy (%), processing power (GHz), and the number of training episodes (#), to determine their influence on the agent’s learning efficiency and overall performance. The effectiveness of an RL agent is often measured using an evaluation metric such as the average reward (R), which quantifies the long-term benefits obtained by following a learned policy. By leveraging predictive modeling techniques, we aim to establish correlations between input parameters and RL performance, helping to refine system design and resource allocation. Sensor accuracy has a direct influence on decision-making processes and is essential in assessing the dependability of state information.. Processing power influences the speed and complexity of model updates, affecting convergence rates. The number of training episodes determines the agent’s exposure to various environmental states, influencing its ability to generalize learned behaviors. This study employs a hybrid approach where RL agents are trained in simulated environments, and machine learning models are used to analyze performance trends. LR provides a simple yet interpretable linear relationship between parameters and average reward, while RFR captures complex nonlinear interactions and enhances prediction robustness. SVM, known for its strong generalization capabilities, further refines decision boundaries in high-dimensional spaces. By comparing these approaches, we derive insights into which factors most significantly impact RL performance and how predictive models can be leveraged to improve autonomous system efficiency. The findings show that the average reward is significantly impacted nonlinearly by sensor accuracy., highlighting the need for high-fidelity sensing in autonomous applications. Processing power influences real-time adaptability, while an optimal number of training episodes ensures sufficient learning without excessive computational overhead. The findings demonstrate that integrating supervised learning techniques with RL not only aids in understanding system behavior but also provides a foundation for adaptive optimization strategies in real-world applications. Future research will concentrate on extending these techniques to more complex, multi-agent environments and exploring meta-learning approaches for enhanced adaptability.
Zitieren
APA
Eppa, A. R. (2025). Optimizing Autonomous Systems through Reinforcement Learning: The Role of Linear Regression, Random Forest, and Support Vector Machines in Decision Making. International Journal of Computer Science and Data Engineering. https://doi.org/10.55124/csdb.v2i2.249
Vancouver
Eppa AR. Optimizing Autonomous Systems through Reinforcement Learning: The Role of Linear Regression, Random Forest, and Support Vector Machines in Decision Making. International Journal of Computer Science and Data Engineering. 2025. doi:10.55124/csdb.v2i2.249
BibTeX
@article{Eppa2025,
author = {Eppa, Akhilesh Reddy},
title = {Optimizing Autonomous Systems through Reinforcement Learning: The Role of Linear Regression, Random Forest, and Support Vector Machines in Decision Making},
journal = {International Journal of Computer Science and Data Engineering},
year = {2025},
doi = {10.55124/csdb.v2i2.249}
}
EUVIMED