Path Loss Prediction based on Machine Learning Techniques: Principal Component Analysis, Artificial Neural Network and Gaussian Process.

Although various linear log-distance path loss models have been developed for wireless sensor networks, advanced models are required to more accurately and flexibly represent the path loss for complex environments. This paper proposes a machine learning framework for modeling path loss using a combination of three key techniques: artificial neural network (ANN)-based multi-dimensional regression, Gaussian process-based variance analysis, and principle component analysis (PCA)-aided feature selection. In general, the measured path loss dataset comprises multiple features such as distance, antenna height, etc. First, PCA is adopted to reduce the number of features of the dataset and simplify the learning model accordingly. ANN then learns the path loss structure from the dataset with reduced dimension, and Gaussian process learns the shadowing effect. Path loss data measured in a suburban area in Korea are employed. We observe that the proposed combined path loss and shadowing model is more accurate and flexible compared to the conventional linear path loss plus log-normal shadowing model.

Development of software sensors for determining total phosphorus and total nitrogen in waters.

Total nitrogen (TN) and total phosphorus (TP) concentrations are important parameters to assess the quality of water bodies and are used as criteria to regulate the water quality of the effluent from a wastewater treatment plant (WWTP) in Korea. Therefore, continuous monitoring of TN and TP using in situ instruments is conducted nationwide in Korea. However, most in situ instruments in the market are expensive and require a time-consuming sample pretreatment step, which hinders the widespread use of in situ TN and TP monitoring. In this study, therefore, software sensors based on multiple-regression with a few easily in situ measurable water quality parameters were applied to estimate the TN and TP concentrations in a stream, a lake, combined sewer overflows (CSOs), and WWTP effluent. In general, the developed software sensors predicted TN and TP concentrations of the WWTP effluent and CSOs reasonably well. However, they showed relatively lower predictability for TN and TP concentrations of stream and lake waters, possibly because the water quality of stream and lake waters is more variable than that of WWTP effluent or CSOs.