arXiv subjects

Krishna Neupane

Publications and source records attributed to Krishna Neupane.

6 recordsLinked to original sources

arXiv2026

Detecting and Explaining Unlawful Insider Trading: A Shapley Value and Causal Forest Approach to Identifying Key Drivers and Causal Relationships

Corporate insiders trade for diverse reasons, often possessing Material Non-Public Information (MNPI). Determining whether specific trades leverage MNPI is a significant challenge due to inherent complexity. This study focuses on two critical objectives: accurately detecting Unlawful Insider Trading (UIT) and identifying key features explaining classification. The analysis demonstrates how combining Shapley Values (SHAP) and Causal Forest (CF) reveals these explanatory drivers. The findings underscore the necessity of causality in identifying and interpreting UIT, requiring the consideration of alternative scenarios and potential outcomes. Within a high-dimensional feature space, the proposed architecture integrates state-of-the-art techniques to achieve high classification accuracy. The framework provides robust feature rankings via SHAP and causal significance assessments through CF, facilitating the discovery of unique causal relationships. Statistically significant relationships are documented between the outcome and several key features, including director status, price-to-book ratio, return, and market beta. These features significantly influence the likelihood of UIT, suggesting potential links between insider behavior and factors such as information asymmetry, valuation risk, market volatility, and stock performance. The analysis draws attention to the complexities of financial causality, noting that while initial descriptors offer intuitive insights, deeper examination is required to understand nuanced impacts. These findings reaffirm the architectural flexibility of decision tree models. By incorporating heterogeneity during tree construction, these models effectively uncover latent structures within trade, finance, and governance data, characterizing fraudulent behavior while maintaining reliable results.

q-fin.ST↗

arXiv2026

Beyond the Numbers: Causal Effects of Financial Report Sentiment on Bank Profitability

This study establishes the causal effects of market sentiment on firm profitability, moving beyond traditional correlational analyses. It leverages a causal forest machine learning methodology to control for numerous confounding variables, enabling systematic analysis of heterogeneity and non-linearities often overlooked. A key innovation is the use of a pre-trained FinancialBERT to generate sentiment scores from quarterly reports, which are then treated as causal interventions impacting profitability dynamics like returns and volatilities. Utilizing a comprehensive dataset from NEPSE, NRB, and individual financial institutions, the research employs SHAP analysis to identify influential profit predictors. A two-pronged causal analysis further explores how sentiment's impact is conditioned by Loan Portfolio/Asset Composition and Balance Sheet Strength/Leverage. Average Treatment Effect analyses, combined with SHAP insights, reveal statistically significant causal associations between certain balance sheet and expense management variables and profitability. This advanced causal machine learning framework significantly extends existing literature, providing a more robust understanding of how financial sentiment truly impacts firm performance.

q-fin.CP↗

arXiv2026

The Information Dynamics of Insider Intent: How Reporting Inversions (Form 144) Mask Informational Rents in Insider Sales (Form 4)

This study identifies and quantifies a significant informational friction embedded in the SEC Form 144 disclosure regime, characterized as predictive decoupling. Drawing on a theoretical foundation of welfare economics, the article argues that the current reporting inversion -- where trade execution (Form 4) frequently precedes the public notice of intent (Form 144) -- violates the conditions for Pareto efficiency by inducing non-symmetric pricing. Utilizing an event-study framework of intent-to-sell windows, the analysis examines cases where insiders file a notice of proposed sale but fail to execute within the statutory 90-day period. The machine learning audit reveals a persistent 52.4 percent opacity rate, where aborted signals remain statistically indistinguishable from routine executions, creating a structural information ceiling that prevents the market from exhausting the signal's informational content. Contrary to the traditional small-firm effect, the study documents a large-cap significance paradox: while small-cap portfolios yield higher absolute abnormal returns (32.21 bps), statistically significant alpha is concentrated in large-cap firms (14.49 bps, $p = 0.021$). The results suggest that Institutional Salience enables more reliable processing of this negative non-event when reputational costs are maximized. Cross-sectional tests confirm that prior idiosyncratic volatility serves as a signal amplifier, with causal estimators identifying an illiquidity jump of up to 2.63 times. To mitigate this market failure, the study proposes a mandatory execution confirmation (Form 144-A) to transition the regime toward bilateral accountability, converting a predictive blind spot into a verifiable data stream and restoring the informational symmetry requisite for efficient capital allocation.

q-fin.CP↗

arXiv2026

The Strategic Gap: How AI-Driven Timing and Complexity Shape Investor Trust in the Age of Digital Agents

Traditional models of market efficiency assume that equity prices incorporate information based on content alone, often neglecting the structural influence of reporting timing and cadence. This study introduces the Autonomous Disclosure Regulator, a multi-node AI framework designed to audit the intersection of disclosure complexity and filing unpredictability. Analyzing a population of 484,796 regulatory filings, the research identifies a structural Strategic Gap: a state where companies use confusing language and unpredictable timing to slow down how fast the market learns the truth by 60%. The results demonstrate a fundamental computational asymmetry in contemporary capital markets. While investors are now good at spotting "copy-paste" text, they remain vulnerable to strategic timing that obscures structural deterioration. The framework isolates 39 high-priority failures where the convergence of dense text and temporal surprises facilitated significant information rent extraction by insiders. By implementing a recursive agentic audit, the system identifies a cumulative welfare recovery potential of over 360\% and demonstrates near-perfect resilience against technical data interruptions. The study concludes by proposing a transition toward an agentic regulatory state, arguing that as information integration costss rise, infrastructure must evolve from passive data repositories into active auditing nodes capable of real-time synthesis to preserve market integrity.

q-fin.CP↗

arXiv2025

An extreme Gradient Boosting (XGBoost) Trees approach to Detect and Identify Unlawful Insider Trading (UIT) Transactions

Corporate insiders have control of material non-public preferential information (MNPI). Occasionally, the insiders strategically bypass legal and regulatory safeguards to exploit MNPI in their execution of securities trading. Due to a large volume of transactions a detection of unlawful insider trading becomes an arduous task for humans to examine and identify underlying patterns from the insider's behavior. On the other hand, innovative machine learning architectures have shown promising results for analyzing large-scale and complex data with hidden patterns. One such popular technique is eXtreme Gradient Boosting (XGBoost), the state-of-the-arts supervised classifier. We, hence, resort to and apply XGBoost to alleviate challenges of identification and detection of unlawful activities. The results demonstrate that XGBoost can identify unlawful transactions with a high accuracy of 97 percent and can provide ranking of the features that play the most important role in detecting fraudulent activities.

q-fin.CP↗

arXiv2024

A Random Forest approach to detect and identify Unlawful Insider Trading

According to The Exchange Act, 1934 unlawful insider trading is the abuse of access to privileged corporate information. While a blurred line between "routine" the "opportunistic" insider trading exists, detection of strategies that insiders mold to maneuver fair market prices to their advantage is an uphill battle for hand-engineered approaches. In the context of detailed high-dimensional financial and trade data that are structurally built by multiple covariates, in this study, we explore, implement and provide detailed comparison to the existing study (Deng et al. (2019)) and independently implement automated end-to-end state-of-art methods by integrating principal component analysis to the random forest (PCA-RF) followed by a standalone random forest (RF) with 320 and 3984 randomly selected, semi-manually labeled and normalized transactions from multiple industry. The settings successfully uncover latent structures and detect unlawful insider trading. Among the multiple scenarios, our best-performing model accurately classified 96.43 percent of transactions. Among all transactions the models find 95.47 lawful as lawful and $98.00$ unlawful as unlawful percent. Besides, the model makes very few mistakes in classifying lawful as unlawful by missing only 2.00 percent. In addition to the classification task, model generated Gini Impurity based features ranking, our analysis show ownership and governance related features based on permutation values play important roles. In summary, a simple yet powerful automated end-to-end method relieves labor-intensive activities to redirect resources to enhance rule-making and tracking the uncaptured unlawful insider trading transactions. We emphasize that developed financial and trading features are capable of uncovering fraudulent behaviors.

q-fin.ST↗