Particulate matter (PM) concentration in southern Thailand has increased significantly due to open crop burning within the southeast Asian sub-region. This study aimed to explore the trend and relationship between PM and meteorological features in southern Thailand. Also, estimate the PM2.5 when monitoring sites measure PM10. Data on PM concentration and meteorological features were taken from air monitoring stations within southern Thailand from 2012 to 2021. Descriptive statistics were used to explore the data, and then a spline model was used to examine the trends and seasonal patterns of PM concentration and meteorological features. A scatter plot matrix and correlation analysis were used to assess the relation between PM and meteorological features. Machine learning models were used to predict PM2.5 concentration. The highest annual average concentration of PM2.5 and PM10 in southern Thailand was 18.9±8.24 μg/m3 and 36.3±14.2 μg/m3 in Songkhla Province, and the lowest concentration of PM2.5 and PM10 was 13.9±7.65 μg/m3 and 27.5±12.2 μg/m3 at Phuket. The Multiple Linear Regression (MLR) and Artifical Neural Network (ANN) almost perform best for the prediction of PM2.5 at each station, with 13.6% average Mean Absoulte Percentage Error (MAPE). Songkhla and Phuket need significant attention from local government officials and policymakers. PM2.5 can be better predicted using the MLR model when it has missing values at some stations. The results help scientists and policymakers to better understand the condition and find the best possible solution to overcome the health issue that arises due to exposure.