Can LLM-assisted regularization increase forecast accuracy for migration flows in low data regimes?

2026-10-05

AI Summary

This study addresses the limitation of traditional gravity models in predicting human migration under data-scarce conditions, where contextual signals are often lacking. To overcome this, it proposes a hierarchical large language model (LLM) reasoning pipeline that extracts push-pull factors from unstructured news reports and integrates them as feature-specific regularization penalties within a weighted Lasso framework to enhance multi-corridor migration forecasting. This work represents the first effort to structurally embed unstructured textual context into classical migration models. Experimental evaluations on migration corridors in Mexico, Syria, and Ukraine demonstrate that the optimal configuration achieves mean absolute percentage errors of 17.15%, 29.29%, and 41.05%, respectively. These results validate both the effectiveness and the generalization potential of LLM-augmented regularization strategies for population mobility prediction in low-data environments.

Abstract

Predicting migration flows remains a significant challenge for traditional gravity-based forecasting models, which primarily rely on structured socio-economic indicators such as economic disparity, political stability, and geographic distance. This work investigates whether Large Language Models (LLMs) can improve migration forecasting by extracting contextual migration-related signals from news articles and incorporating them into a weighted Lasso forecasting framework through feature-specific regularization penalties. The proposed framework uses hierarchical LLM inference pipelines to classify migration-related push--pull signals from news data and evaluates the resulting forecasting performance across multiple migration corridors between November 2021 and November 2022, including Mexico--United States, Ukraine--Poland, and Syria--Turkey. Experimental results showed mixed performance across migration corridors and modeling strategies, and no single regularization approach consistently outperformed the others across all experiments. The best-performing Mexico configuration, which consisted of a gravity-based model augmented with the proposed push--pull ratios, achieved a Mean Absolute Percentage Error (MAPE) of 17.15%, while the strongest Syria configuration achieved a MAPE of 29.29% using Direct LLM-Lasso. For Ukraine, the best-performing configuration used LLM-Assisted Regularization (AR) and achieved a MAPE of 41.05%. Overall, the results suggest that contextual article-derived features and LLM-guided regularization can improve migration forecasting under certain conditions, although migration corridor characteristics, article volume, and hyperparameter configuration strongly influenced performance.

Problem

• migration forecasting
• low data regimes
• Large Language Models
• gravity-based models
• regularization

Innovation

• Large Language Models
• LLM-assisted regularization
• migration forecasting
• weighted Lasso
• push-pull signals