Determining the Likelihood of War in a Country in the Middle East and North Africa Based on Economic and Climate Data via Machine Learning

Abstract

War is a major issue in our current world, with 56 current conflicts around the world, the most since WWII. This is extremely concerning, as most people are not aware of most wars or coups that happen in different countries. A particularly unstable region in this age is the Middle East and North Africa. This study attempts to predict the onset of wars and coups in the MENA region by using economic and weather data and comparing it to other economic and weather data captured during times of conflict in these particular areas. A pipeline classification was built with weather and economic data from various time periods in places that were in a conflict at the time. Articles on events in these regions were also collected to gauge the situation at any given time. The model would then fit the data as well as the article features and output the likelihood of war in a particular area based on the economic and weather data it received. If we can predict wars and other conflicts before they can even happen, we can better prevent them, or decrease the casualty rate. While this is to some extent a report of work done, it is also a proposal for future work.

1 Introduction

Although wars and coups have occurred consistently throughout history, the frequency and intensity of such violent conflicts have notably increased in recent years. In 2021, there were 27 ongoing conflicts, with the number steadily rising every year (Koop 21; Rustard 24). Today, the world is experiencing the most conflicts and wars it has faced since WW2 (Archie 22). Although there are many terrible events happening around the world, affecting different countries everywhere, the general public only knows about a few of them, and most are unaware of how drastic they are.

Unfortunately, this also means that there isn’t too much news coverage and aid for the issues faced by countries experiencing lesser-known conflicts, as sometimes it gets too dangerous in certain areas for people to help (Burchell 2020). This model can be utilized to identify early indicators of unrest and the onset of conflict in specific regions, enabling proactive intervention to mitigate escalation. Ideally, specialists and humanitarian organizations would have the ability to anticipate areas at risk of instability, allowing for preemptive measures that reduce the potential for widespread harm. While there is a possibility that such a model could be misused by authoritarian parties to better monitor and suppress dissent, this risk can be managed with ethical oversight and responsible implementation. This paper demonstrates a program that could allow for the prediction of future conflicts in the MENA region using climate and economic data.

2 Literature Review

The purpose of this literature review is to determine which factors play an important role in causing conflict, and ways we can look for such factors in advance to prevent such events. Nations have declared war on other nations or groups for a variety of reasons. Some may have been more abstract, such as conflicts between religion or culture, while others were also more concrete reasons, such as the need more land, population, or resources. However, because of World War II, war and conflicts in general have changed irrevocably. Now conflicts are characterized by better surveillance technologies, with new inventions such as drones and more accurate weapons. They also tend to involve many civilians now, rather than just soldiers, and often involve various factions, some which aren’t state-affiliated. In this new post-World War II era, conflicts have started to be coined by US analysts as Fourth-Generation Warfare (Holbrook 20). This era has also seen a slow decline in foreign wars, yet the number of civil wars has started to rise, with the number of such conflicts having tripled since the last decade (von Einsiedel 2017).

A civil war, defined by Britannica, is a conflict that involves the clashing of multiple organized non-state actors, which differs from interstate, or foreign wars, which are wars declared on one nation by another. There are two wars that have the world’s attention currently, the Russo-Ukrainian and Israel-Palestinian Conflicts. Although some parties claim otherwise, both happen to be foreign wars, conflicts waged by one functional government against another. Although both are classified as interstate wars, their occurrence represents a deviation from the declining trend of foreign conflicts, and their prominence in global discourse can be attributed to their relative rarity compared to the numerous civil wars that commonly affect failed states.

Despite the fact that it seems as though there are a plethora of differences between the two types of conflicts, in Lemke and Cunningham’s 2009 paper, they come to the conclusion that there isn’t a point to making distinctions between the two. Despite this, distinguishing between the two types could provide more insight, and is something to look into. For now, though, research should instead prioritize evaluating their overall effectiveness as catalysts for conflict. Although there are a lot of different factors that could influence the decision to declare war, there isn’t a definitive determinant that inevitably leads to war. Since there isn’t one deciding factor as to why all conflicts start, many researchers attempt to search for one factor or a set of factors that have the biggest impact as to whether wars start or not.

According to Stewart (2002), war is strongly influenced by the economic is sues of a nation and its surrounding area at a given time. His research paper, ”Root causes of violent conflict in developing countries”, focuses on 4 big economic hypotheses, and how these factors could possibly contribute to the start of intra-state wars in the modern era: Group Motivation Hypothesis, Private Motivation Hypothesis, Failure of the Social Contract, and Green War Hypothesis. While the paper had very in depth information on how economics could cause such conflicts, it excludes many other factors, and doesn’t go into detail concerning political, religious, cultural, resource, geography, or climate related factors. Although economics undeniably plays a major part in the inner and foreign situation of a state, it shouldn’t be the only factor. Another thing that has changed between 2002 and the present day is the US’ involvement and policies regarding the Middle East region after 9/11 (Esfandiary 2021). Although this article was written post-9/11, it ignores big changes in the political landscape of the Middle East and before the true consequences of the US’ actions had been realized. Despite this, it still is applicable to some degree as the economic consequences mentioned in Stewart’s writings would still hold major influence in a nation’s stability and tendency to declare war.

Coccia (2019) explains many different theories that have been used to explain how and why wars happened. She elaborates on both historical and modern theories on the cause of war and conflict. Modern theories could explain or uncover a big part of why such conflicts happen, and the inclusion of past theories allows insight into how modern theories could have developed. The contents of the paper are mostly theoretical, finding reasons that countries would declare war and other theories relating to the nature of conflict. Although this paper focuses on foreign wars, and doesn’t mention the case of a civil war, it has still provided valuable information on why conflicts could happen, and the logic behind how they could start. Even after someone has determined what cause of war they want to further research, there are a variety of ways for them to find, categorize, and store the data they use.

In order to find reasons for involvement in wars, Makarov (2015) uses statistics. He takes data from times when a certain nation is either at war or in a time of peace. The paper considers political (e.g. diplomacy), economic, and religious factors that could influence the likelihood a country would be in a state of warfare. Once Makarov obtains the data, statistical models and distributions are used to then gather the data needed to build graphs and make projections of the previously mentioned factors and their effects. However, the way statistics are used in Makarov’s case differs from this paper’s approach to utilizing statistics and data. While Makarov mainly focuses his paper on why certain nations could be involved in wars, this paper attempts to facilitate prevention of such conflicts by picking up on trends that have been seen before and predicting the likelihood of conflict in a certain area.

3 Methods

This paper primarily examines the economic and climatic factors that contribute to the onset of war, analyzing their correlation with reported events in news articles. The model does this by utilizing features extracted from articles to assess the severity of conflict in a given region at a given time and integrates economic and climate data to evaluate their potential influence on the situation. Research from the UN shows that climate is important to consider as it could determine whether or not a country could face civil unrest, or even wage war against other nations. If a country has been facing climate that isn’t typical for the region over an extended period of time, it could cause an assortment of problems. In one report from the UN in 2021, they state that extreme weather and climate changes, along with other factors such as disease, have caused many droughts and subsequent food shortages for places such as Latin America, Asia, and Africa (UNFCCC 2021). There have been similar trends before and during the outbreak of war as well, with droughts and food shortages being caused by drastic climate change in a region. This could cause leaders to fight for precious resources or cause parts of the country to revolt against the central government. Economics could also be an important part to consider, as harsher economic times could cause unrest, and increase the chances of conflict, whether it be from disgruntled civilians or a government desperate to alleviate itself from these issues.

The code for the prediction model first gathers both economic and climate data, as well as articles and events corresponding to different times during the war. These data points are what the model later trains and tests on to gauge the accuracy of its results. This model is trained on data from the Darfur conflict and Syrian civil war occurring in Sudan and Syria respectively, although more conflicts in the region could be added for more complex and accurate predictions. In order to encode the articles into a readable state, qdrant’s FastEmbed, an embedding model was used to turn raw data (articles) into vectors from 1 to 1 using cosine similarity of article feature vectors (qdrant Version 0.6.0). The outputs/events, originally represented as strings (e.g. ’war’ or ’coups’), are transformed into k-hot encoded vectors to facilitate data visualization. This transformation employs a multi-label k-hot encoding approach, allowing for the representation of multiple events occurring within a single time period, thereby enhancing the clarity of the data.

Before training and testing the data, it is necessary to interpolate periods with no events or articles, or other data. The analysis period is defined from the earliest to the most recent data point, with all intervening empty time periods requiring interpolation to ensure data continuity. Grab-and-hold interpolation is used to fill in all dates with economic and climate data from previous data points. The empty time periods without economic or climate data will be interpolated with the values of the most recent time with data. In the event that datetime periods are requested earlier than the earliest economic or climate data point recorded, then all datetime periods before the first data point will be extrapolated with the values of that first data point. After all the datetimes with empty data have been filled in, the article features, economic data, and climate data are stacked and stored into a single array of features.

The classifier works by binning all dates into larger intervals. The original prediction model required daily data from the entire time span to make predictions, which resulted in long processing times and reduced accuracy due to repetitive values and interpolation-induced redundancy. To address this, the classifier utilizes a sliding window approach, aggregating data into one-week intervals. This method consolidates multiple days of data into a single time frame, enhancing both computational efficiency and model performance.

After the events and the features have been interpolated, a sliding window classifier with a Radial Basis Function kernel is trained to predict the possibility of conflicts in certain regions.

4 Results

Climate, and to a smaller extent weather, plays a big factor in the likelihood of conflict, as it could affect resource scarcity (crops and water), and economic prosperity, etc. To understand the effects of changing climate and deviations in the average temperature on the occurrence of conflicts.

The probabilities that were output by the SVC based on article feature values. The article feature values are abstract features of information extracted from the articles. Based on the data presented in Figure 1, an inverse correlation between the two variables is evident. This trend is particularly noticeable from the starting point of the figure through the 2020s. With the exception of early 2023, a significant rise in the average temperature always saw a decrease in the Article Features Value, and vice versa.

Figure 2 illustrates the relationship between a nation’s economic conditions and its propensity to experience foreign conflicts or civil unrest. The graph tends to show an inverse relationship between the two variables for Syria. Although the graph uses only one feature from the articles and can’t be used to draw definitive conclusions, it is an interesting pattern

In contrast, Sudan demonstrated a somewhat different pattern. Its GDP per capita appeared to be inversely correlated with the article feature value, particularly in the earlier years leading up to 2018. Although the article feature values do not directly correlate with the frequency or severity of a conflict, an increase could still be influenced by the expansion of conflicts, as wars can drive factors that contribute to its growth. The localized nature of the Sudanese conflict in the Darfur region could have influenced this input inconsistency. Taking that into consideration, one reason the article feature value increased could be that before 2018, this conflict primarily affected the Darfur area, while the rest of the country experienced minimal conflict. This regional concentration likely explains why Sudan was generally less impacted by fluctuations in the escalation or de-escalation of conflict. However, in 2018, the conflict significantly expanded, reaching surrounding areas such as Kordofan and Blue Nile, which may explain the country’s increasing vulnerability to these events. After this period, the trend began to align more closely with that of Syria, before witnessing a notable surge in GDP per capita during a phase of reduced conflict around early 2023, greatly surpassing Syria’s economic growth during the same time frame.

Figure 3 provides additional context to Figures 1 and 2. It illustrates that the majority of conflicts in the MENA region during the 2010s were civil wars, whereas foreign wars appeared to become more prevalent in the 2020s. While the relationship is not definitive, there seems to be a stronger correlation between the unstable economic conditions in Sudan and Syria and the prevalence of civil wars. Although it is challenging to determine causality, the fluctuating GDP per capita in both countries throughout the 2010s may have contributed to economic instability, potentially triggering civil unrest, coups, and internal conflicts. While it remains difficult to draw firm conclusions, the 2020s appear to be a period of relative economic stability. This, combined with a notable increase in temperatures, may have prompted both nations and their affiliated groups to seek external resources before conditions could escalate further, thus reducing the risk of internal conflict. These factors may help explain the observed decrease in internal conflicts and the corresponding rise in foreign wars.

5 Future Work

Although this prediction model has already made some predictions based on the quantity of data points given to it, some improvements can be implemented to better streamline this process, and show more accurate results. Currently, all data gathered is treated and displayed as one big timeseries, with all events from all different areas of the region and all different time periods being grouped. This makes it harder for the prediction model to make accurate predictions. As importantly, little has been gathered to come to these conclusions, and an increased volume of data could allow for the easier splitting of the time table into smaller episodes, as well as decreased overfitting.

Currently another problem is that the prediction model predicts the probability of events happening on a particular date solely based on information in a preceding sliding window with a binning interval of one week, while disregarding the sliding windows that came before it. The interval the model is currently fixated on may not provide sufficient indications as to conflict, while the previous ones could give more context to help the model predict events. An alternative could be an LSTM or other RNN for this task to allow the model to account for the entire preceding history of the current episode.

6 Conclusion

This paper explored the prediction of wars in the MENA region using statistics and machine learning. It identified causes of war, such as economic and climate factors, and analyzed papers with similar topics. The model used was trained on various climate and economic data points, and predicted outputs with varying degrees of accuracy. In order to obtain more accurate results, more data must be obtained. The successful development of this predictive model has significant implications for conflict prevention efforts. Early prediction of potential conflicts could enable intervention measures that can mitigate their impact or prevent escalation. This could be particularly useful for humanitarian organizations and governments in conflict-prone regions. If more funding and focus were put into this project, accurate results could be achieved, minimizing unrest and conflict, and increasing overall stability in the MENA region.

7 Bibliography

Lindsey, R., & Dahlman, L. (2024, January 18). Climate Change: Global Temperature. Climate.gov; National Oceanic and Atmospheric Administration.https://www.climate.gov/news-features/understanding-climate/climate-change-global-temperature

WMO Climate Normals. (2021, July 8). National Centers for Environmental Information (NCEI). https://www.ncei.noaa.gov/products/wmo-climate-normals

Compilation of Geospatial Data (GIS) for the Mineral Industries and Related Infrastructure of Africa – ScienceBase-Catalog. (n.d.). Www.sciencebase.gov. https://www.sciencebase.gov/catalog/item/607611a9d34e018b3201cbbf

Middle East & North Africa — Data. (n.d.). Data.worldbank.org. https://data.worldbank.org/region/middle-east-and-north-africa?view=chart

World Economic Situation And Prospects: February 2020 Briefing, No. 134 —Department of Economic and Social Affairs. (2020). Un.org. https://www.un.org/development/desa/dpad/publication/world-economic-situation-and-prospects-february-2020-briefing-no-134/

Business In The Middle East: Cultural Differences You Need To Know. (2022, November 2). https://www.milestoneloc.com/business-in-the-middle-east/

Alex. (2015, June 9). Detailed Maps Of The World’s Religions – Vivid Maps. Vivid Maps. https://vividmaps.com/maps-of-worlds-religions/#google vignette

Poverty headcount ratio at $2.15 a day (2017 PPP) (% of population) – World —Data. (n.d.). Data.worldbank.org. https://data.worldbank.org/indicator/SI.POV.DDAY?locations=1W&start=1984&view=chart

Data. (2023). Resource Trade. https://resourcetrade.earth/?year=2022& category=164&units=value&autozoom=1

Gu, D. (2019). Population Division Exposure and vulnerability to natural disasters for world’s cities*. https://www.un.org/en/development/desa/population/publications/pdf/technical/TP2019-4.pdf

Herre, B., & Arriagada, P. (2023). The Human Development Index and relatedindices: what they are and what we can learn from them. Our World in Data.https://ourworldindata.org/human-development-index

Burchell, K. (2020). Reporting, Uncertainty, and the Orchestrated Fog of War: A Practice-Based Lens for Understanding Global Media Events. International Journal of Communication. https://ijoc.org/index.php/ijoc/article/viewFile/11205/3102

World Bank. (2024). World Bank Group – International Development, Poverty and Sustainability. Worldbank.org. https://www.worldbank.org/ext/en/home

IMF. (2024). International Monetary Fund . IMF. https://www.imf.org/en/Home Vision of Humanity — Destination for Peace. (n.d.). Vision of Humanity. https://www.visionofhumanity.org

National and Local Weather Radar, Daily Forecast, Hurricane and information from The Weather Channel and weather.com. (2019, March 7). The Weather Channel. https://weather.com/

Conflict Trends: A Global Overview, 1946–2023 – World — ReliefWeb. (2024, June 10). Reliefweb.int. https://reliefweb.int/report/world/conflict-trends-global-overview-1946-2023

Archie, A. (2022). World is seeing the greatest number of conflicts since the end of WWII, U.N. says. NPR. https://doi.org/1089884798/united-nations-conflict-covid-19-ukraine-myanmar-sudan-syria-yemen

Thegsaljournal. (2020, February 23). How has warfare changed since WWII? The GSAL Journal. https://thegsaljournal.com/2020/02/23/how-has-warfare-changed-since-wwii/

United Nations. (2020). A New Era of Conflict and Violence. United Nations;United Nations. https://www.un.org/en/un75/new-era-conflict-and-violence

von Einsiedel, S. (2017). Civil War Trends and the Changing Nature of Armed Conflict. United Nations University. https://collections.unu.edu/eserv/UNU:6156/Civil war trends UPDATED.pdf

Mackarov, I. (2015). Statistical look at reasons of involvement in wars. arxiv. https://arxiv.org/pdf/1508.06228

Stewart, F. (2002). Root causes of violent conflict in developing countries .BMJ, 324(7333), 342–345. https://doi.org/10.1136/bmj.324.7333.342

Lemke, D., & Cunningham, D. E. (2009, January 1). Distinctions Without Differences?: Comparing Civil and Interstate Wars. https://www.researchgate.net/publication/228191496 Distinctions Without Differences Comparing Civil and Interstate Wars

Esfandiary, Dina. “The Anxiety Effect: How 9/11 and Its Aftermath Changed Gulf Arab States’ Relations with the U.S.” Www.crisisgroup.org, 15 Sept.112021, www.crisisgroup.org/middle-east-north-africa/gulf-and-arabian-peninsula/united-arab-emirates-united-states-saudi-arabia.

Nirant Kasliwal. “FastEmbed.” Github.io, 2025, qdrant.github.io/fastembed/. Accessed 30 Mar. 2025.

---