Scholarlink

<p>Illicit investment in residential real estate poses persistent challenges for anti–money‑laundering (AML) enforcement due to ownership opacity, fragmented data, and the scarcity of confirmed criminal cases. This study evaluates supervised machine‑learning and neural‑network models to determine their ability to detect properties linked to criminal activity and to identify the most influential predictors of illicit investment. Using a rare‑event dataset and cross‑validated modelling framework, the analysis shows that meaningful patterns can be detected despite extreme class imbalance. Across logistic regression, Random Forest, XGBoost, CART oversampling experiments, and an artificial neural network, consistent predictors—market_value, owner_legal_person, owner_owns_multiple, land_acres, and out_of_state_owner—emerge as central risk indicators. The findings support Rational Choice Theory by illustrating how offenders exploit structural vulnerabilities to maximize utility. Policy implications include property‑level risk scoring, early‑warning systems, and enhanced support for gatekeepers. Machine‑learning approaches show strong potential to strengthen real‑estate AML frameworks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Detecting Illicit Investment in Real Estate: a Machine‑Learning Approach to Rare‑Event AML Risk

  • 期刊论文

  • 发表日期：

    2026年5月6日

查看全文
  • Mark Lokanan

摘要

Illicit investment in residential real estate poses persistent challenges for anti–money‑laundering (AML) enforcement due to ownership opacity, fragmented data, and the scarcity of confirmed criminal cases. This study evaluates supervised machine‑learning and neural‑network models to determine their ability to detect properties linked to criminal activity and to identify the most influential predictors of illicit investment. Using a rare‑event dataset and cross‑validated modelling framework, the analysis shows that meaningful patterns can be detected despite extreme class imbalance. Across logistic regression, Random Forest, XGBoost, CART oversampling experiments, and an artificial neural network, consistent predictors—market_value, owner_legal_person, owner_owns_multiple, land_acres, and out_of_state_owner—emerge as central risk indicators. The findings support Rational Choice Theory by illustrating how offenders exploit structural vulnerabilities to maximize utility. Policy implications include property‑level risk scoring, early‑warning systems, and enhanced support for gatekeepers. Machine‑learning approaches show strong potential to strengthen real‑estate AML frameworks.