Investigation of Chiang Rai irregular migration and its environmental factors using Machine Learning

Abstract

Irregular migration constitutes a multifaceted challenge in Southeast Asia. Thailand serves as a significant transit and destination for migrant workers from neighboring countries, especially Myanmar, Laos, and Cambodia. Chiang Rai Province is located in the uppermost region of Thailand, which is joint between Myanmar and Laos. Its geographical and socioeconomic conditions attract a lot of migrant workers, both legal and illegal. This study investigates the environmental factors influencing irregular migration along Thailand-Myanmar and Laos borders in Chiang Rai province. The study utilized geospatial data, including Remote Sensing (RS) and Geographic Information System (GIS) data, combined with arrest point data of irregular migration along the border from 2019 to 2023; three machine learning algorithms consisting of XGBoost, Random Forest, and LightGBM were applied to analyze and predict irregular migration patterns.

The primary objectives are threefold: (1) to analyze the variables associated with irregular migration in Chiang Rai Province between 2019 and 2023 (2) to evaluate and compare the predictive performance of machine learning algorithms for accurately predicting irregular migration (3) to create a risk map visualizing the spatial distribution of irregular migration incidents within the study area.

The study focuses on four districts of Chiang Rai province which consist of Mae Sai, Mae Chan, Chiang Saen, and Mae Fa Luang. Their different geographic characteristics present challenges for irregular migration management. Mae Sai represents the highest irregular migration incidents. Meanwhile, Mae Fa Luang and Mae Chan reported fewer cases. The study shows that the topography of the study area influences migration patterns. Remote sensing data, including the Normalized Difference Vegetation Index (NDVI) and Bare Soil Index (BSI), were utilized to assess land cover and vegetation density. Moreover, to understand human activity and the physical landscape, Nighttime light data and Digital Elevation Models (DEM) were applied in this study.

The results show that irregular migration activity was highest in April or during the summer season and lowest during the dry/winter months of October and December. Temporal analysis also marked a preference for nighttime crossings to avoid detection. The highest number of irregular migration cases was recorded in 2022, with 1,882 cases, while in 2019, 56 incidents were recorded, the lowest number out of all five years.

Myanmar nationals make up the majority of irregular migrants (1,991 cases), while Thai and Chinese nationals were recorded in smaller proportions. The study indicated that 55% of migrants were male, while 45% were female; this shows that migration pressures affect both genders almost equally.

XGBoost exhibited the highest predictive accuracy among the other machine learning algorithms, with an R-squared value of 0.91 and the lowest Root Mean Square Error (RMSE). Feature importance analysis across all models consistently identified road networks as the most significant predictor of irregular migration. The transportation infrastructure is facilitating movement across borders. Elevation and proximity to rivers are also key factors that light the physical environment's impact on irregular migration routes. In addition, The Random Forest model, which is less accurate than XGBoost, also highlighted Nighttime light as a crucial factor, indicating that irregular migrants prefer areas with the dance of human activity.

An irregular migration risk map was generated using the best machine learning algorithm. The map presents the high-risk area along the border, especially near Mae Sai. Mae Sai district is in a critical environment; its road networks and challenging terrain converge to create migration hotspots. The risk map from this study is very beneficial, especially for border authorities to have more focus and effective monitoring and intervention efforts.

This study illustrates the application of geospatial techniques and machine learning to analyze the irregular migration patterns in Chiang Rai Province along Thai-Myanmar and Laos border. Environmental factors, geographic data, and irregular migration arrest data are important tools for investigating migration patterns in border areas. Developing more accurate predictive models will be able to support policymakers and border control authorities to understand essential insights. As irregular migration remains a critical issue in Thailand and Southeast Asia, applying predictive analytics and geospatial technologies will help in effective border management and immigration control strategies.

-