Studies investigating the correlation between particulate matter (PM) concentrations measured by a light scattering (LS) device and environmental factors are crucial to identify LS values with significant errors. Herein, the relationship between PM2.5 obtained through beta attenuation monitoring (BAM) and LS was examined with respect to seven environmental factors. Machine learning (ML) and general statistical methods were employed to reveal complex relationships. Data from five cities were initially analyzed to understand the association between BAM measurements and environmental factors. Our findings confirmed that wind direction (WD) had a strong nonlinear impact on short-term measurements, whereas temperature and local pressure had similar effects on long-term PM2.5 measurements. Subsequently, a method was developed using general statistical techniques to establish an environment wherein LS could maintain a relatively high accuracy level. Furthermore, ML techniques were employed to determine that LS was more affected (by 8.2%) by the changes in WD compared with BAM, emphasizing the importance of designing devices capable of responding to WD. Finally, LS was calibrated using four ML algorithms, and through a quantitative evaluation of coefficient of determination, mean absolute error, and root mean square error values, AdaBoost was identified as an effective algorithm for correcting LS measurements. With this understanding of the correlation between PM2.5 and environmental factors, along with an efficient correction method, its widespread adoption in future research concerning real-time PM measurement is anticipated.