Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study

A Master of Science thesis in Engineering Systems Management by Khaled Essam Hosny Ahmed entitled, “Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study”, submitted in February 2026. Thesis advisor is Dr. Hussam AlShraideh and thesis co-advisor is Dr. Anis Samet. Soft...

وصف كامل

محفوظ في:
التفاصيل البيبلوغرافية
المؤلف الرئيسي: Ahmed, Khaled Essam Hosny (author)
التنسيق: doctoralThesis
منشور في: 2026
الموضوعات:
الوصول للمادة أونلاين:https://hdl.handle.net/11073/33523
الوسوم: إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
_version_ 1870676404235927552
author Ahmed, Khaled Essam Hosny
author_facet Ahmed, Khaled Essam Hosny
author_role author
dc.contributor.none.fl_str_mv Alshraideh, Hussam
Samet, Anis
dc.creator.none.fl_str_mv Ahmed, Khaled Essam Hosny
dc.date.none.fl_str_mv 2026-06-22T07:52:56Z
2026-06-22T07:52:56Z
2026-02
dc.format.none.fl_str_mv application/pdf
dc.identifier.none.fl_str_mv 35.232-2026.04
https://hdl.handle.net/11073/33523
dc.language.none.fl_str_mv en_US
dc.relation.none.fl_str_mv Master of Science in Engineering Systems Management (MSESM)
dc.subject.none.fl_str_mv Negative Coefficient of Skewness
Down-to-Up Volatility
Expected Shortfall
Stock-Price measures
Stock-Liquidity measures
Machine Learning
dc.title.none.fl_str_mv Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study
dc.type.none.fl_str_mv info:eu-repo/semantics/publishedVersion
info:eu-repo/semantics/doctoralThesis
description A Master of Science thesis in Engineering Systems Management by Khaled Essam Hosny Ahmed entitled, “Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study”, submitted in February 2026. Thesis advisor is Dr. Hussam AlShraideh and thesis co-advisor is Dr. Anis Samet. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form).
format doctoralThesis
id aus_bb985271a0cee1c2d251c9954434977d
identifier_str_mv 35.232-2026.04
language_invalid_str_mv en_US
network_acronym_str aus
network_name_str aus
oai_identifier_str oai:repository.aus.edu:11073/33523
publishDate 2026
repository.mail.fl_str_mv
repository.name.fl_str_mv
repository_id_str
spelling Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning StudyAhmed, Khaled Essam HosnyNegative Coefficient of SkewnessDown-to-Up VolatilityExpected ShortfallStock-Price measuresStock-Liquidity measuresMachine LearningA Master of Science thesis in Engineering Systems Management by Khaled Essam Hosny Ahmed entitled, “Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study”, submitted in February 2026. Thesis advisor is Dr. Hussam AlShraideh and thesis co-advisor is Dr. Anis Samet. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form).The stability and predictability of stock markets are crucial for shaping global economic dynamics, particularly in emerging markets where volatility and risk are elevated. This research conducts a rigorous study to predict stock crashes caused by price volatility and periods of insufficient liquidity using machine learning techniques. Using novel firm-specific stock-price and stock-liquidity measures, specifically Negative Coefficient of Skewness (NCSKEW), Down-to-Up Volatility (DUVOL), and Expected Shortfall (ES), among others, we engineered 261 predictive features across multiple time windows. To define an output crash event, we employ a consensus-based ensemble approach, flagging crashes when at least five of the twenty complementary risk indicators simultaneously fall below their rolling 5th percentile thresholds. Our analysis leverages an exceptionally comprehensive dataset: 23 million observations spanning approximately 37 years (1985 – 2022) across 24,197 stocks from 57 emerging markets. This geographic and temporal breadth significantly exceeds that of the existing literature. Evaluating four machine learning algorithms: Classification and Regression Trees (CART), Logistic Regression (GLM), XGBoost, and Random Forest, we find that Random Forest achieves superior performance: 98.81% accuracy, Cohen’s Kappa of 0.815, and a Receiver Operating Characteristic – Area Under the Curve (ROC AUC) of 0.995, demonstrating exceptional prediction performance despite severe class imbalance ~3.96% (25:1 ratio). Notably, the model maintains 99.9% sensitivity while keeping the false alarm rate at 0.11%. Through SHAP analysis of Random Forest, we identify Expected Shortfall (ES) as the dominant crash predictor, with ES measures across multiple confidence levels (90%, 95%, 97.5%, 99%) and time windows dominating feature importance rankings. The SHAP beeswarm analysis reveals a clear relationship: higher tail risk values directly increase stock crash probability, with 504-day ES windows proving most predictive. Our results demonstrate that ensemble machine learning methods effectively capture complex, non-linear relationships between stock crashes and various price and liquidity measures, providing actionable insights for investment decision-makers.College of EngineeringDepartment of Industrial EngineeringMaster of Science in Engineering Systems Management (MSESM)Alshraideh, HussamSamet, Anis2026-06-22T07:52:56Z2026-06-22T07:52:56Z2026-02info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/doctoralThesisapplication/pdf35.232-2026.04https://hdl.handle.net/11073/33523en_USMaster of Science in Engineering Systems Management (MSESM)oai:repository.aus.edu:11073/335232026-06-23T06:52:41Z
spellingShingle Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study
Ahmed, Khaled Essam Hosny
Negative Coefficient of Skewness
Down-to-Up Volatility
Expected Shortfall
Stock-Price measures
Stock-Liquidity measures
Machine Learning
status_str publishedVersion
title Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study
title_full Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study
title_fullStr Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study
title_full_unstemmed Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study
title_short Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study
title_sort Forecasting Stock Crashes in Emerging Markets: A Comparative Machine Learning Study
topic Negative Coefficient of Skewness
Down-to-Up Volatility
Expected Shortfall
Stock-Price measures
Stock-Liquidity measures
Machine Learning
url https://hdl.handle.net/11073/33523