Uso de IA para estimar desempenho de uma rede Wi-Fi

Abstract

This study presents the development and evaluation of a predictive model of IA aimed at estimating the performance of Wi‑Fi networks in domestic and institutional environments. The research is justified by the growing demand for efficient, low‑latency, and cost‑effective wireless connections, which are essential for applications such as video conferencing, online gaming, and Internet of Things devices. The main objective was to build a system capable of predicting network throughput (Mbps), considering variables such as the distance from the access point and physical barriers. The methodology comprised three main stages: automated data collection in a controlled environment, processing and organization of measurements using the Python programming language and statistical libraries. The results showed that the model can represent the real behavior of signal attenuation even under different environmental conditions, standing out for its simplicity of implementation and low computational cost. The technical analysis revealed that outlier treatment was decisive for achieving near‑ideal performance (R² ≈ 0.91), and that the Power Law model reproduced the asymptotic behavior of the signal more realistically at greater distances. It is concluded that the proposed approach contributes to the optimization of wireless networks, supporting decision‑making regarding equipment positioning and infrastructure, while opening perspectives for future research involving hybrid models and more advanced machine learning techniques