@article{Sotiroudis2021b,

title = {Enhancing Machine Learning Models for Path Loss Prediction Using Image Texture Techniques},

author = {S. Sotiroudis and K. Siakavara and G. Koudouridis and P. Sarigiannidis and S. Goudos},

url = {https://www.researchgate.net/publication/352111245_Enhancing_Machine_Learning_Models_for_Path_Loss_Prediction_Using_Image_Texture_Techniques},

doi = {10.1109/LAWP.2021.3086180},

year  = {2021},

date = {2021-06-03},

journal = {IEEE Antennas and Wireless Propagation Letters},

volume = {(Early Access)},

abstract = {The performance of machine learning-based path loss models relies heavily on the data they use at their inputs. Feature engineering is therefore essential for the models success. In the work at hand, we extract a new set of input features, based on image texture. The image that we use represents the footprint of the urban built-up area, where the gray scale values of the building blocks are analogue to their heights. We extract texture information by applying the Segmentation-based Fractal Texture Analysis algorithm on the orthogonal area that is bounded between the transmitter and the receiver. To the best of our knowledge this is the first time that such a technique is applied to a path loss modeling problem in electromagnetics. The algorithm thus delivers a new set of features, based on the images texture, which eventually reveal the built-up profile of the area. These new features are injected to an already existing feature set. Comparative analysis shows that the addition of texture-based features leads to enhanced predictions, for a diverse set of transmitter heights, machine learning algorithms, and performance metrics.},

keywords = {image texture, machine learning, mobile communications, pathloss prediction},

pubstate = {published},

tppubtype = {article}

}