Gas Values Prediction and Forecast Techniques Based on DL Models

Analyzing, monitoring, and predicting air quality trends is essential for our health and environmental sustainability. Artificial intelligence can help us achieve this purpose in a timely and effective manner, contributing to improving air quality for people's health. This study proposes the creation of a dataset containing various types of environmental gases (
$CO,\ CH_{2}O,\ C_{6}H_{6}$
, and
$H_{2}S$
) in outdoor environments and the application of a Recurrent Predictor (Long Short-Term Memory - LSTM) for data prediction. After a phase of network training and validation, a comparison is made between the actually recorded data and those reconstructed by the neural network. The performance in terms of Mean Squared Error (MSE) obtained by the model for each type of gas is evaluated and commented upon.