Immigration to countries, including legal and illegal migrants, are generally not easy to predicted. In fact, standard models have been criticized as being too general or not useful. There are alternative approaches, however, which include using online searches that can be georeferenced. The presumption is that migrants may conduct online searches as they prepare to leave their country of origin. Additionally, web services also provide a similar way to capture login patterns for users, also aiding in migration estimates.
One recent study used online Google searches, from 2004-2015, in the form of Google Trends Index data, and searched terms in English, French, Spanish, looked at search data from non-OECD countries, that is countries that are relatively poor, where these results are normalized to make them comparable. Then, a gravity model is used to compare to migrant data and searches, with the results generally showing that searches demonstrate or have a positive relationship to intent and migration. This enables online searches that can be georeferenced to be used as forecasts of where and when migration might vary between countries.
Such research is similar to other work that has used repeated login to the same website for different accounts, in this case using Yahoo! services, where repeated use allows a global-scale pattern to be developed. By capturing where people log into accounts across a timespan covering a year, migration flows can be developed, showing where people have decided to move or go to for different periods. For many cases, this includes short-term visits rather than long-term migration, but longer use of the accounts can differentiate migration beyond short trips. This enables a temporal and migration map to be developed, although the predictive capability might be best used with other supportive data, such as longer-term collected data or economic indicators.
Another work used blogging, specifically micro-blogging in Brazil, to demonstrate migration flows there. In this case, more than 13 million blog posts were used to create a network display of where migrants were likely to go where the forecast was created using a linear regression model. The effective assumption is that outside migrants will blog or communicate more as they move, resulting in changing internet activity as cities attract more migrants. This assumption, in fact, does not replace traditional migrant data collected but it does demonstrate that online tracking might be a useful way to relatively accurately estimate the number of migrants coming to different cities.
Not only can international migration best apply forecasts using geocoded searches, but internal migration within countries can be useful for economists attempting to determine where population is likely to go in coming years. This was the case in the United States, where housing and and employment queries obtained from Bing.com allowed a study to demonstrate where movement within the United States has been trending in recent years by forecasting migration and searches using regression models. Additionally, the benefit of extracting online searches is that the intent of migrants can be better mapped, as searches also demonstrate why migration is occurring and what resources migrants may require, such as schools or if they want to buy or rent. The potential in this type of approach is it can be used by economists and researchers that are interested in obtaining faster data on migration trends that may have long-term social consequences that can affect infrastructure or other resources.
What these works demonstrate is that geolocated data, particularly in the realm of big data problems where millions of searches or login information is present, can be used to track and forecast migration inside countries and at an international scale. This demonstrates we can now create better tools to forecast migration by using the ubiquitous nature of the Internet and its use by different socio-economic classes.
 For more on the study looking at migration and applying Google searches and search index data from countries, see: Böhme, M.H., Gröger, A., Stöhr, T., 2019. Searching for a better life: Predicting international migration with online search keywords. Journal of Development Economics S0304387819304900. https://doi.org/10.1016/j.jdeveco.2019.04.002.
 For more on using Yahoo1 login data to produce a map of migration, see: State, B., Weber, I., Zagheni, E., 2013. Studying inter-national mobility through IP geolocation, in: Proceedings of the Sixth ACM International Conference on Web Search and Data Mining– WSDM ’13. Presented at the the sixth ACM international conference, ACM Press, Rome, Italy, p. 265. https://doi.org/10.1145/2433396.2433432.
 For more on forecasting migrants using micro-blogging, see: Vaca-Ruiz, C., Quercia, D., Aiello, L.M., Fraternali, P., 2014. Tracking Human Migration from Online Attention, in: Nin, J., Villatoro, D. (Eds.), Citizen in Sensor Networks. Springer International Publishing, Cham, pp. 73–83. https://doi.org/10.1007/978-3-319-04178-0_7.
 For more on the study using Bing.com searches, see: Lin, Allen Yilun and Cranshaw, Justin and Counts, Scott, Forecasting U.S. Domestic Migration Using Internet Search Queries (February 26, 2019). Proceedings of the 2019 World Wide Web Conference (WWW’19), May 13–17, 2019, San Francisco, CA, USA. ACM, New York, NY, USA, 12 pages.. Available at SSRN: https://ssrn.com/abstract=3341776.
Radiant Earth has launched Radiant MLHub, a cloud-based open library for training geospatial data used by machine learning algorithms.
Researchers have compiled an enormous global dataset with over four trillion satellite-based measurements of sea surface temperature.