Abstract
How do you know you select enough tuning dataset from measurements to guarantee model prediction accuracy? Tuning datasets are often selected based on simple random sampling with predefined rates. Usually, these rates are determined as <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">a</i> / <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b</i> , where <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">a</i> % of the data goes to training and the remaining <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">b</i> % goes to testing. But it is not clear to what extent tuning dataset in order to minimize the estimation path loss errors. It is, thus, required to analyze the performance of channel modeling by selecting—among all measurement samples—appropriate tuning dataset. Using radio measurements and deterministic Ray Launching techniques to collect enough reliable samples, this letter analyzes the impact of tuning dataset selection—expressed in terms of the mean absolute error and cost—on a novel Kriging-aided in-building measurement-based path loss prediction model.
Highlights
Accuracy and efficacy are fundamental channel modeling features
In [10] the authors analyze the following approaches: what happens when the method selection of tuning dataset varies in order to select the right one to get the most out of Kriging, and what percentage of data should be selected for tuning dataset to obtain the best goodness of fit; choosing a suitable approach for optimal tuning selection involves a deeper study that includes the analysis of the separation sampling distance and the cost function that minimizes the mean absolute error (MAE), the separation sampling distance, and the tuning dataset of the system
Towards choose an optimal tuning dataset to be representative of the extent of the target area, which is reasonable to perform in a practical situation, it is helpful to analyze and estimate the error when a different size of tuning dataset is selected to predict path loss in testing locations
Summary
Accuracy and efficacy are fundamental channel modeling features. it is important to strike a balance between them. Empirical models are useful for practical radio designs, but the similarity of the measured and predicted environment is a fundamental consideration to achieve accurate estimations; deterministic models are complex in terms of both computational resources and rigorous building details for channel modeling; but hybrid models represent a good trade-off between empirical and deterministic models by including rigorous fitting and suitable validations [1]–[3]. Measurement-based models employ radio measurements to perform model fitting and model validation through tuning and testing datasets, respectively In most studies, these datasets are selected from the corresponding measurements according to a desired ratio a/b that is based on random sampling [8], [9], i.e., an a% for tuning the model and a b% for testing the model. In [10] the authors analyze the following approaches: what happens when the method selection of tuning dataset varies in order to select the right one to get the most out of Kriging, and what percentage of data should be selected for tuning dataset to obtain the best goodness of fit; choosing a suitable approach for optimal tuning selection involves a deeper study that includes the analysis of the separation sampling distance and the cost function that minimizes the mean absolute error (MAE), the separation sampling distance, and the tuning dataset of the system
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
