Abstract:Since the emergence of wireless communication networks, quality aspects of wireless links have been studied for several technologies. By analyzing the rich body of existing literature on link quality estimation that uses models built from data traces, we have noticed that the techniques used for modelling are becoming increasingly complex. Several recent estimators use machine learning techniques, however it is sometimes difficult to understand the reported results as each model that relies on machine learning requires a complex design and development process. Each step of this process has the potential to significantly impact the final performance. The aim of this paper is to provide an in depth study of how each step in the process of designing and developing a link quality estimator using machine learning affects its performance. Based on the analysis of the state of the art, we selected a representative subset of machine learning models used in the literature and a representative publicly available dataset and performed a systematic study on the influence of the design decisions taken in each step of the machine learning process on the performance of machine learning based link quality estimators. Our results show that measurement data pre-processing and feature engineering have a higher influence on the performance of the model than the choice of the algorithm.