Prediction of daily maximum ozone levels using Lasso sparse modeling method

TL;DR

This study employs Lasso sparse modeling to predict next-day maximum ozone levels using a large set of features, achieving good accuracy and outperforming other methods.

Contribution

The paper introduces a Lasso-based approach for high-dimensional ozone prediction, effectively reducing features and improving accuracy over existing methods.

Findings

Lasso model achieved RMSE of 5.63 ppb for maximum ozone prediction.

The approach demonstrated superior accuracy compared to other recent methods.

The model effectively selected relevant features from a large candidate set.

Abstract

This paper applies modern statistical methods in the prediction of the next-day maximum ozone concentration, as well as the maximum 8-hour-mean ozone concentration of the next day. The model uses a large number of candidate features, including the present day's hourly concentration level of various pollutants, as well as the meteorological variables of the present day's observation and the future day's forecast values. In order to solve such an ultra-high dimensional problem, the least absolute shrinkage and selection operator (Lasso) was applied. The $L_1$ nature of this methodology enables the automatic feature dimension reduction, and a resultant sparse model. The model trained by 3-years data demonstrates relatively good prediction accuracy, with RMSE= 5.63 ppb, MAE= 4.42 ppb for predicting the next-day's maximum $O_3$ concentration, and RMSE= 5.68 ppb, MAE= 4.52 ppb for predicting the next-day's maximum 8-hour-mean $O_3$ concentration. Our modeling approach is also compared with several other methods recently applied in the field and demonstrates superiority in the prediction accuracy.