Examining relative variable importance (RVI) and strength of interaction effects (SIE) of particulate matter (PM10) concentrations in moving vehicles at selected route in east coast of Malaysia

Abstract

The poor indoor air quality inside the cabin of a coach can be due to indoor air pollutants or to inadequate ventilation. Exposure to surrounding airborne particulate matter inside a long hour journey may increase health risk of the coach passengers. In this study, a one-minute gaseous (carbon monoxide (CO), carbon dioxide (CO2)), physical parameters (relative humidity, airflow, temperature), coach speeds, and number of occupants' data were gathered two days trips for each route travelled coaches via motorways and local routes. The Route 1 journey started from Terminal Kota Bharu (TKB) to Terminal Bersepadu Selatan (TBS) and Route 2 link between Terminal Kuala Terengganu (TKT) and Terminal Bersepadu Selatan (TBS). Data were analysed to understand variability of the variables by using R software and its packages were applied temporally and statistically. An advance and a computational intelligence technique named the Boosted Regression Trees (BRT) was used to examine Relative Variable Importance (RVI) and estimation the Strength of Interaction Effects (SIE) inside cabin coaches. Results demonstrates significant variation in PM10 are distance (39.19%), elevation (15.19%) temperature (10.56%) humidity (6.65%) for Route 1 and 46.7% for CO (13.77 %), humidity (8.92 %), and elevation (8.01 %) for Route 2. It was found that strength of interactions (SIE) index was determined for humidity and temperature for both routes which range between 0.23 to 0.32, followed by humidity and airflow interactions (0.05) and CO and airflow (0.18) for Route 2. Since SIE or H-values fall within range from 0 to 1 which value of zero indicates that there are no interactions between the variables, the interaction between variables will be stronger if the value is closer to 1. It appears and promises that this machine learning method in particular BRT is able to examine in cabin pollutants and physical environment data.