Machine learning models that are able to incorporate any number of exogenous features, to predict origin and destination human migration flows. These machine learning models outperform traditional human mobility models on a variety of evaluation metrics, both in the task of predicting migrations between US counties as well as international migrations. In general, predictive machine learning models of human migration will provide a flexible base with which to model human migration under different what-if conditions, such as potential sea level rise or population growth scenarios.

This paper uses Google Trends data for migration forecasting models. It demonstrates significant improvements by replacing the gravity model with a Long short-term memory (LSTM) artificial recurrent neural network (RNN) architecture, outperforming standard artificial neural networks (ANNs). However, limitations include not testing the model on unknown origin-destination country pairs and potential improvements in computation methods. The study suggests enhancing models by incorporating factors like catastrophic events, unemployment rates, and internet usage. Additionally, utilizing categorical labels differently could capture more complex factors like distance and language similarities between countries. Despite losing interpretability with machine learning approaches, future work aims to employ interpretability techniques to identify key features for better migration forecasts.

Migration Predictions in Austalian Population Projections 2017-2066. Australian population projections, sourced from the ABS, span 2017 to 2066 with three series based on fertility, mortality, and migration assumptions. 

This paper aims to enhance migration predictions by combining qualitative and quantitative methods through Bayesian models. It reveals the unpredictability of migration flows, emphasizing the importance of embracing uncertainty. Expert knowledge aids parameter estimation but has limited impact on model selection. Forecasts beyond 5-10 years are deemed unreliable due to increasing uncertainty. Decision-makers must weigh the costs of underestimating or overestimating migration flows based on uncertain forecasts.

This article presents a new method for predicting international migration rates aims to provide clearer insights into global movement patterns. Despite an apparent increase in migration relative to population size, actual migration rates haven't risen significantly. While the model assumes independence of prediction errors across countries and time, real-world correlations may exist. Outliers, like wars, can affect predictions, favoring simpler models over complex ones. Nonetheless, the chosen model might underestimate uncertainties in long-term forecasts. Enhancements, such as incorporating population projections and considering economic and political factors, are key to refining predictions for better future planning.

The article examines a game-theoretic model of undocumented immigration, highlighting the strategic interaction among firms, native labor, elected officials, and undocumented immigrants, revealing that border enforcements impact is diminished due to strategic interactions and discussing the effects of labor market uncertainty on migration policy decisions made before or after market conditions are known. Undocumented immigration, a complex political issue, involves conflicting interests among politicians, firms, and workers in both destination and source countries. The game theory model reveals that changes affecting only some players may have minimal impact due to strategic interactions. For instance, lower wages in source countries can prompt more family members to attempt undocumented immigration, but increased border enforcement in destination countries may counterintuitively reduce successful crossings. Future research could explore the interplay between economic and political incentives in shaping immigration policies and their long-term effects.

The paper evaluates the use of Google Trends (GT) in forecasting migration patterns, highlighting its varying predictive power and the contextual factors influencing its efficacy. Unlike previous studies focusing on GT's usability, this research emphasizes model complexity and contextual influences on GT's predictive ability. By analyzing EUROSTAT asylum application data and push-pull indicators, three classes of gravity models are trained to predict refugees' EU destination choices, with GT's inclusion examined. Results indicate heterogeneous effects of GT inclusion, challenging the notion of GT outperforming established predictors. This nuanced perspective underscores the need for comprehensive analysis of big data's strengths and limitations in migration forecasting, encouraging further research in diverse contexts and with alternative data sources beyond GT.

This paper proposes a statistical framework to combine social media data with traditional survey data to produce timely nowcasts of migrant stocks by state in the United States. The authors developed a statistical framework  to produce short-term migrant stock projections in the US, combining Facebook and survey data. Results showed improved accuracy, especially with Bayesian hierarchical time series modeling. Despite limitations, the approach offers real-time insights into migration trends, aiding in identifying unexpected changes. Future research could address evolving biases in social media data and expand the model to incorporate additional demographic factors. This methodological approach, blending traditional and novel data sources, can be adapted to analyze various demographic indicators beyond migrant stocks.

This paper was presented at the Joint Eurostat/UNECE Work Session on Demographic Projections in April 2010. An econometric model was used to estimate migration flows to and from Norway, incorporating economic forecasts. The resulting projection depicts a notable shift: from a decline due to the financial crisis, to recovery-driven growth, and ultimately a gradual decline reflecting Norway's changing economic landscape, notably reduced petroleum production. This projection model's ability to capture such a complete turnaround in trend is uncommon. The predicted decline aligns with other studies and recent migration data, showing a peak in immigration in 2008 followed by increased outmigration. Public opinion surveys suggest a majority in several countries, including Norway, favoring more restrictive immigration policies, although attitudes towards refugees and asylum seekers vary within Norway itself.

The project develops a forecasting model for total forced displacement from a country 1-3 years ahead. Total forced displacement is a combination of refugees and asylum seekers outside a given country and the internally displaced in the samecountry.

This is a report by the JRC building scenarios for demographics of the EU and its related challenges up to 2060. The report delves into the factors shaping European demographics, including migration, fertility, mortality, education levels, and labor force participation rates. It addresses challenges such as population aging, a shrinking workforce, and high emigration levels in some EU Member States. Looking ahead to 2060, the report examines various scenarios to understand the long-term implications of these trends and how to mitigate negative consequences.

This study analyzes how remittances and migration affect poverty rates in Kosovo using data from 2011. By simulating a scenario without remittances, it predicts household consumption and compares poverty rates across regions. Results show remittances reduce poverty, especially in rural areas. To sustain this effect, policymakers must encourage investment-oriented use of remittances for long-term economic development.

This paper proposes a method in which different potential change scenarios within a predetermined set of six key drivers. They subsequently involve migration scholars in assessing how these scenarios might impact both migration flows and the remaining five drivers. The study reveals potential future migration trends based on various drivers but the impact of these drivers may differ between sending and receiving countries. The study acknowledges limitations in simplifying complex migration dynamics and the need to consider disparities within countries and regions. It also highlights the non-linear nature of migration changes and the influence of unforeseen events like the COVID-19 pandemic. This approach is applied to analyze the case of migration pressure and demand originating from less developed countries to Europe up to the year 2050.

Migration emerges as a top political priority for the European Union (EU), requiring comprehensive data for effective management. The report outlines a methodology for estimating expatriate numbers across 17 EU nations, utilizing Facebook Network data. By adjusting for the representativeness of Facebook users, considering demographics and gender ratios, the aim is not to replicate migration statistics but to provide supplementary expatriate estimates. Despite its advantages in timeliness and accessibility, using ocial media data presents notable methodological challenges. 

This paper argues that goresight methodology, whether used alone or alongside traditional approaches, holds significant promise in augmenting empirical data by offering potential future scenarios. In this paper, several scenarios were used as narratives connecting present actions to future outcomes. The SEEMIG foresight exercise in Slovenia exemplifies stakeholder engagement in envisioning migration-related futures, yielding policy insights essential for enhancing migration policies, fostering integration, and addressing specific societal needs. 

FEWS NET employs scenario development as a methodology. This process enables the organization to construct a "most likely" scenario of the future, facilitating the fulfillment of its core mission: providing early warnings about food security crises to decision-makers.

This study introduces an adaptive machine learning system integrating data from origin countries, European border crossings, and asylum decisions. It forecasts mixed migration flows 1-4 weeks ahead, outperforming benchmarks. Unlike static models, it adapts to diverse migration processes, addressing temporal and spatial complexities. By highlighting causal factors, it enhances understanding and policymaking, potentially aiding international protection efforts. Rooted in migration theory and data science, this approach offers a robust early warning system. Its adaptability suggests broader applicability to forecasting complex systems with sparse data.

The paper assesses the practicality of constructing a refugee flow forecasting model, leveraging high-dimensional data and machine learning techniques. Evaluating out-of-sample forecasting models against a Random Walk baseline, the study incorporates Google Trends time series alongside asylum seeker flows and classical predictor variables. The results show that the Random Forest and XGBoost models are the most effective. It also finds that while Google Trends predictors offer marginal performance gains overall, they show promise in forecasting specific corridors with substantial refugee flows. The study recommends customizing forecasting models based on corridor-specific performance and shows the inherent limitations in predicting sudden migration shocks like the COVID-19 pandemic. 

This paper compares predictions against reality using nationally representative data. The results emphasize the importance of cognitive abilities and social networks in forecasting accuracy and find that homogeneous networks create blind spots, highlighting the cognitive gains obtained from diverse networks. The authors finds a need for  interdisciplinary data collection and call for integrating psychological and sociological insights for nuanced understanding and improved forecasting models. It show that social diversity can improve predictions particularly when using methodologies such as the Delphi methodology.

The paper addresses the challenge of forecasting and understanding large-scale migration patterns, crucial for both academia and government planning. It explores the use of internet search queries as a predictor for domestic migration trends, showing that these queries correlate with and even precede official migration metrics. The study demonstrates that incorporating search query data improves migration prediction models, offering insights into migration motives such as housing and employment. This method provides real-time, leading indicators of migration trends.

This report evaluates the long term impact of receving refugees from the Ukraine-Russia war in the population of Estonia. Initially, the Foresight Centre projected varying numbers of war refugees settling in Estonia, expecting a positive impact on population growth due to higher birth rates among immigrants, particularly women. However, updated data on gender and age distribution suggests this impact may diminish sooner. Integration and sustained support are crucial for societal cohesion. Despite high initial support, economic fluctuations could affect refugees' employment opportunities.

Subchapter on Migration in German population predictions. Some of the reports produced with this data include Nowcast, persons seeking protection, immigration history, naturalisations, among other statistics.

The Hilando Futuros project seeks to invite people to reflect on the present and future of migration processes in Montevideo, by opening up dialogue through the creation of an embroidered textile piece based on the narratives and stories of individuals who have chosen to migrate from other countries to Uruguay

The report proposes an augmentation of an integrated stochastic population and labor force participation forecasting framework by a gravity-equation component to model future immigration and emigration, their interaction, and their determinants more appropriately. By conducting a stochastic forecast, the authors find that until 2060 the potential labor supply in Germany is declining by 11.7 percent, strongly driven by the even more distinct decline of the working-age population and only partially cushioned by rising participation rates.

In this article, the articles propose a Bayesian model to overcome the limitations of the various data sources. The focus is on estimating recent international migration flows among 31 countries in the European Union and European Free Trade Association from 2002 to 2008, using data collated by Eurostat. The authors also incorporate covariate information and information provided by experts on the effects of undercount, measurement, and accuracy of data collection systems.

The research underscores the significance of demographic factors in predicting international migration patterns. It suggests that while economic and political factors influence migration, demographic changes, such as population growth, play a crucial role. Practical implications include anticipating future migration pressures based on demographic trends. For instance, the United States may see a decline in immigration due to slow population growth, while Europe may face sustained immigration pressures. Effective management requires coordinated immigration policies, especially within entities like the European Union. Additionally, this study suggests that historical migration patterns may not continue in the same way, with rapidly growing countries experiencing lower emigration rates.

Island World, is an online game that serves as a small world simulation. Designed to facilitate self-reporting through diverse opportunities for disclosure, including closed and open-ended questions,the objective is to enhance this data by incorporating quantitative behavioral data exhibited by players within the game, alongside the system logs generated by the game itself. There are plans for the next version to include data-mining algorithms to uncover patterns that may be difficult to discern otherwise. However, the focus remains on gathering user feedback to refine the game and ensure accurate correlation between environmental stressors and migration behavior The ultimate goal of the game is to aid policymakers in mitigating the disruptions caused by human migration, including understanding individuals' perceptions of institutional roles in managing migration issues and disaster response. 

By integrating economic performance and demographic trends into functional migration models, this paper identifies two phases in migration trends, influenced by economic development and migration capabilities. The study focuses on emigration models, acknowledging the challenges of linking migratory moves to potential destinations. Additionally, it explores the potential impact of labor market dynamics on migration flows, particularly highlighting the growing migration potential in Africa. Future research could integrate immigration models to establish comprehensive population projection models and exploring various scenarios to improve the migration projections used. 

The estimates in this report carried out by the Office for National Statistics of the United Kingdom are rely on administrative and survey data complemented by innovative statistical modeling.  Some of the main points highlighted in this report are the UK's provisional migration estimates for 2022 indicate 1.2 million long-term immigrants and 557,000 emigrants, resulting in a net migration of 606,000.

This article introduces a simple system dynamic model to simulate migration dynamics, considering economic and social incentives. The model demonstrates how differences in foreign exchange rates and economic downturns, like COVID-19, influence migration patterns. Despite its simplicity, the model offers insights into migration behavior and the role of economic factors (excluding remittances). 

This empirical study examining the cross-impact of socioeconomic development and migration in Russian regions.  The econometric analysis revealed that various types of migration, including international and internal labor migration, do not significantly impact the socioeconomic development of Russian regions. Labor migration, particularly internal movement, is strongly influenced by regional development, driven primarily by economic factors such as job opportunities. Long-term migration, however, is influenced by diverse factors beyond economics, including education, family reasons, and lifestyle preferences. The study emphasizes the multidirectional nature of migration flows and the importance of considering a wider range of socioeconomic and migration indicators, especially disaggregated by age groups. Further research at both regional and municipal levels is necessary to comprehensively assess migration's role in addressing economic challenges in Russia

This article proposes a Flow-Specific Temporal Gravity (FTG) model, informed by the random utility framework, to address these limitations. Using EUROSTAT asylum statistics and various indicators, including climate, conflict, and economic factors, the FTG model demonstrates the ability to account for heterogeneous migration behavior. Results indicate that as flow time-series data lengthens, FTG models' predictions become more accurate, while FE models become less predictive. The FTG approach faces challenges due to limited data availability, which restricts the inclusion of predictors. The paper concludes that while FTG models offer promise in understanding migration dynamics, addressing data constraints and exploring alternative modeling approaches are crucial for improving predictive accuracy.

This research uses nationwide data to forecast large-scale human flows in Spain. By analyzing trip data spanning over nine months, a Graph Neural Network (GNN) is used as a powerful tool for predicting outbound trips within an hour, outperforming other tools such as  Long Short-Term Memory (LSTM) networks and rivaling the Autoregressive Integrated Moving Average (ARIMA) with greater efficiency. Furthermore, this approach simplifies deployment by requiring only one model for the entire dataset, contrasting with ARIMA's need for multiple models. Future research aims to validate this model across diverse datasets and explore synergies with online social network data for enhanced prediction accuracy.

This article improves the understanding of migration aspirations of professionals in Europe by leveraging a previously untapped data source: aggregate-level information on LinkedIn users open to work-related international relocation.  It demonstrates the utility and limitations of leveraging LinkedIn for studying job-related openness to international migration and quantifying the attractiveness of European countries. The authors find that  combining LinkedIn data with sources like the Gallup World Poll could offer deeper insights into migration trends.

This working paper discusses insights from research on strategic foresight, conducted in collaboration between the OECD and the Portuguese government to enhance decision-making and policy development in Portugal. It provides guidance for governments to further promote and disseminate strategic foresight for decision making.

This paper offers insights into Spanish immigration and emigration trends for the 21st century, highlighting two main conclusions. Firstly, migration flows are primarily influenced by dyadic and time-invariant factors, with demographic characteristics at origin playing a significant role. Secondly, while medium-term projections suggest Spain will continue to receive large numbers of immigrants, long-term forecasts vary widely, making it challenging to draw definitive conclusions beyond 2050. 

These projections are part of the Population Projections for Canada (2018 to 2068). The authors argue that forecasting immigration levels presents challenges due to government policies' influence, subject to rapid changes. In 2017, Canada announced a significant increase in planned immigration for three years, extending targets even higher for 2019-2021. This volatility underscores the difficulty in long-term immigration assumptions, highlighting the need for flexible forecasting methodologies. The authors argue that migration's complexity involves economic, cultural, and political factors, making future flows hard to predict. Despite theories attempting to explain migration, practical applications are hindered by data limitations and complexity which are explored in the article.

This paper investigates the "pull factor" claim using discrete time-series counts, focusing on aggregate-level analysis. Employing time-series decomposition and structural change tests, it explores shifts in crossing attempts concerning search-and-rescue policies. The authors also use Bayesian structural time-series (BSTS) modeling generates synthetic counterfactual time-series, considering various factors like trends, seasonality, and covariates. This approach resembles difference-in-differences designs, offering robust predictions and inferring treatment effects. Model validation and covariate selection ensure reliability, with BSTS predicting unobserved crossing attempts under unchanged search-and-rescue policies. This methodological framework contributes to understanding migration dynamics based on specific policy interventions.

In this study, Lake Urmia was selected as the primary case study, and subsequent to the cleaning and normalization of the pertinent data (DOI:10.100), three distinct algorithms were implemented in the final stage of the proposed methodology. The findings reveal that the SVM-based model exhibited the highest performance with an accuracy rate of 88%.

To address the challenge of incomplete and incomparable migration flow data, Sander er al (2014) use statistical methods to estimate movements by linking changes in migrant stock data over time. They assume that individuals are more inclined to stay rather than move, enabling them to estimate the minimum migrant flows needed to reconcile stock differences for each country of birth. This process is replicated for 196 countries, resulting in birthplace-specific flow tables for global migration flows. Adjustments are made for births and deaths during the period to refine country-specific net migration estimates, which closely align with those published by the United Nations. 

This paper challenges traditional frameworks of skilled migration, offering nuanced insights into the diverse patterns of middle-class migration. By highlighting the inadequacy of the 'expats' discourse, it provides a more comprehensive understanding of migration dynamics in contemporary world cities.  The research underscores the importance of contextual analysis in addressing the complexities of skilled migration, offering valuable insights for designing inclusive and responsive policies in diverse urban contexts.

This paper addresses the core insights of the QuantMig project. Highlighting the complexity and uncertainty inherent in migration processes, examines selected theoretical models for understanding migration, analyzes migration drivers in origin, destination, and transit nations, evaluates data and methodologies for building future migration scenarios and presents strategies for creating and communicating these scenarios. The project intends to work as a blueprint for studying future European migration flows, advocating for scenario setting to anticipate migration trends.

The article introduces a novel method using digital records to evaluate global migration and mobility rates, particularly beneficial for countries lacking traditional data systems. It highlights a trend of increasing mobility, especially among females, and addresses methodological challenges like selection bias in digital samples. This paper explores the potential of digital records for studying migration patterns and social networks. It also discusses using Twitter for sentiment analysis related to migration. It also addresses challenges such as selection bias in analyzing email data but it emphasizes the potential for theoretical advancements through innovative data collection methods.