Proof of concept of the potential of a machine learning algorithm to extract new information from conventional SARS-CoV-2 rRT-PCR results.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has been and remains one of the major challenges modern society has faced thus far. Over the past few months, large amounts of information have been collected that are only now beginning to be assimilated. In the present work, the existence of residual information in the massive numbers of rRT-PCRs that tested positive out of the almost half a million tests that were performed during the pandemic is investigated. This residual information is believed to be highly related to a pattern in the number of cycles that are necessary to detect positive samples as such. Thus, a database of more than 20,000 positive samples was collected, and two supervised classification algorithms (a support vector machine and a neural network) were trained to temporally locate each sample based solely and exclusively on the number of cycles determined in the rRT-PCR of each individual. Overall, this study suggests that there is valuable residual information in the rRT-PCR positive samples that can be used to identify patterns in the development of the SARS-CoV-2 pandemic. The successful application of supervised classification algorithms to detect these patterns demonstrates the potential of machine learning techniques to aid in understanding the spread of the virus and its variants.

Optimisation of thermal comfort and indoor air quality estimations applied to in-use buildings combining NSGA-III and XGBoost

Indoor environmental quality (IEQ) monitoring of in-use buildings has become essential in recent years due to the COVID-19 pandemic, as it significantly affects the well-being, health and productivity of building users. Nevertheless, knowing in real time the environmental conditions in large multi-zone areas is a difficult issue. Thus, the use of machine learning techniques to estimate indoor conditions has increased considerably. The aim of this paper is to present an interpolation model, based on an optimised extreme gradient boosting algorithm, to estimate every minute the indoor temperature, relative humidity and CO2 concentration inside buildings. These estimations are obtained without requiring permanent monitoring in the occupied zone. The optimisation, focused on finding the minimun number of monitoring devices needed to provide accurate interpolations, is performed using the multi-objective genetic algorithm NSGA-III. This methodology was applied in a research centre in the north-western Spain. The results show that the optimised or reduced model is capable of estimating indoor temperatures and relative humidity with relative errors below 6% and CO2 levels below 10%.