Commodity prices are notoriously hard to forecast, and whether the returns of commodity exchange-traded funds (ETFs) can be predicted remains an open question. We compare three deep learning models, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Transformer, for forecasting the returns of six Deutsche Bank commodity ETFs covering agriculture (DBA), base metals (DBB), broad commodities (DBC), energy (DBE), oil (DBO), and precious metals (DBP). Using daily price data from January 2007 to December 2025, we predict daytime returns (open to close) and overnight returns (previous close to open) separately, over five horizons of 1, 5, 30, 60, and 180 trading days. Each model sees a 20-day window of price-based features, returns, rolling averages and volatilities, momentum, and recent lags, built from all six ETFs. All models are trained on a strict chronological split and judged by two simple, decision-oriented measures: how often they call the direction correctly, and the risk-adjusted return (annualized Sharpe ratio) of a stylized long–short strategy that ignores transaction costs. Formal significance tests with HAC corrections for overlapping targets, bootstrap confidence intervals, and comparisons with ARIMA, random forest, and simpler benchmarks corroborate strong predictability in overnight DBP and daytime DBB at medium horizons. Predictability turns out to be highly specific to the asset, the trading session, and the horizon. Overnight returns of the precious metals ETF (DBP) are by far the most predictable: the correct direction is called 71.6% of the time at 60 days and 76.7% at 180 days, with Sharpe ratios reaching about 15. Base metals (DBB) daytime returns are predictable at 30 days and oil (DBO) daytime returns at 180 days, whereas one-day-ahead forecasts and agricultural returns (DBA) stay essentially unpredictable. The Transformer has a slight edge at longer horizons and the GRU at shorter ones. Key directional accuracy and Sharpe ratio results are confirmed by Newey–West HAC significance tests and Diebold–Mariano forecast comparison tests with the Harvey–Leybourne–Newbold small-sample correction; HAC standard errors at the 180-day horizon exceed naïve OLS errors by a factor of approximately 7.4, and we explicitly flag results that do not survive this correction. A three-fold expanding walk-forward validation scheme corroborates the main findings, with DBP overnight and DBO daytime predictability persisting across all evaluation windows. Deep learning architectures statistically and economically outperform logistic regression, ridge regression, and momentum baselines on the most predictable configurations. An anomalous failure of all models on DBA daytime returns at the 180-day horizon is diagnosed as a regime-driven artefact associated with post-2021 commodity inflation, not a general feature of agricultural return dynamics. The broader lesson is that splitting returns into daytime and overnight components exposes predictable structure that conventional close-to-close returns hide.

This paper conducts a comprehensive empirical evaluation of macroeconomic outcomes under two competing policy paradigms that have governed the United States since the end of World War II: the Keynesian-inflected New Deal Order (1946–1980), rooted in John Maynard Keynes’s theory of aggregate demand management, and the Neoliberal Order (1981–2024), shaped by the price mechanism epistemics of Friedrich A. Hayek and the monetarist counter-revolution of Milton Friedman. Drawing on the historiographical frameworks we operationalize the 1981 policy transition as a natural experiment and apply a fourteen-component analytical pipeline to an annual panel of fifteen Federal Reserve Economic Data (FRED) series spanning 1946–2024. Our methods include: Pruned Exact Linear Time (PELT) structural break detection; normality-adaptive hypothesis testing with Cohen’s d effect sizes; Principal Component Analysis (PCA); a gradient-boosted XGBoost classifier with SHAP interpretability; static and rolling Phillips curve estimation; bootstrapped fiscal multiplier analysis using the corrected Federal Surplus/Deficit ratio; Composite Macroeconomic Performance Index (CMPI) construction; Pearson correlation structure comparison; regime-stratified Okun’s Law analysis; real federal fund rate and yield curve decomposition; M2 velocity and monetary transmission analysis; k-means sub-period temporal clustering; and government debt trajectory with decade-level performance benchmarking. The Keynesian era produced significantly higher mean real GDP growth (3.7% vs. 2.7%), lower structural unemployment (5.0% vs. 6.2%), and a functioning Phillips curve trade-off that collapsed entirely under neoliberalism (β^≈0, p=0.94). The real federal fund rate averaged negative under the Keynesian regime and turned persistently positive after the Volcker shock. M2 velocity declined sharply in the neoliberal era, refuting the stable quantity theory link central to monetarist policy prescriptions. Temporal clustering recovers five distinct macroeconomic epochs that only partially align with the 1981 historiographical boundary, revealing substantial heterogeneity within each broad regime. The Reagan-era debt expansion, made visible by the corrected debt/GDP trajectory, constitutes a striking empirical contradiction to neoliberal fiscal rhetoric. Together, these results support the interpretation that the two policy regimes represent genuinely distinct macroeconomic equilibria with persistent and measurable consequences for growth, employment, price stability, monetary transmission, and fiscal sustainability. A battery of robustness checks addressing confounding structural shocks, alternative regime boundaries, policy endogeneity, model complexity, and index-construction sensitivity confirms that the core findings are not artifacts of the 1981 periodization, and a cluster-validity analysis formally supports a five-epoch, rather than strictly binary, characterization of the postwar policy landscape.

PURPOSE:To determine the feasibility of developing an artificial intelligence (AI) algorithm based on optical coherence tomography (OCT) images as an automated screening tool for diagnosing retinal thinning in children with sickle cell disease (SCD). METHODS:This retrospective consecutive series included children with SCD who had an ophthalmic examination at a Pediatric Tertiary Care Hospital, including OCT imaging between January 1998 and August 2022. Three different machine learning algorithms were evaluated: logistic regression, K-Nearest Neighbors (KNN), and random forest. RESULTS:A total of 348 OCT scans from 174 eyes of 87 patients (54% males) were included. Using the original data set, KNN algorithm outperformed both the random forest and logistic regression algorithms when using two OCT scans per patient. However, with cross-validation, this model's accuracy dropped to 77.11%. When duplicating the data set's values, the random forest algorithm performed best, demonstrating the highest accuracy after cross-validation of 96.0%, AUC, sensitivity, specificity, and a F1 score all reaching 1, when using one OCT scan per patient. CONCLUSION:AI-based analysis of OCT imaging is a promising tool in the early detection of sickle cell maculopathy in the pediatric population.

This paper presents a comprehensive spatiotemporal decomposition of equity returns for nine top-weighted constituents of the Dow Jones Industrial Average (DJIA) over a twenty-year period spanning January 2004 through December 2023, encompassing 5033 trading days and multiple market regimes, including the Global Financial Crisis (2008–2009), the COVID-19 crash and recovery (2020), and the Federal Reserve tightening cycle (2022–2023). Daily price movements are systematically partitioned into two orthogonal sessions: the open-to-close (OTC, or daytime) session, capturing within-session price discovery, and the close-to-open (CTO, or overnight) session, capturing the accumulated information arrival and liquidity dynamics between market closes and subsequent opens. Within this bipartite return framework, we construct and rigorously evaluate 24 distinct trading strategies, spanning directional (long/short), neutral (cash), momentum (inertia), and contrarian (reversal) approaches, applied independently to each session or in combinatorial cross-session configurations. Each strategy is evaluated under three transaction cost regimes (0, 1, and 2 basis points per trade) using an initial investment of $100, and assessed using annualized return, annualised volatility, Sharpe ratio, Sortino ratio, and maximum drawdown. The study universe—comprising UnitedHealth Group (UNH), Goldman Sachs (GS), Microsoft (MSFT), Home Depot (HD), Caterpillar (CAT), Amgen (AMGN), McDonald’s (MCD), Salesforce (CRM), and Honeywell (HON)—captures cross-sector heterogeneity across Healthcare, Financials, Technology, Consumer Discretionary, Industrials, Biotech, and Consumer Staples. The universe is selected from the top-weighted DJIA constituents as of early 2026; the paper is, therefore, best read as a focused, in-depth case study of index-representative large-cap names rather than a general cross-sectional statement about all U.S. equities. The principal findings are threefold. First, the overnight session consistently delivers superior risk-adjusted performance: seven of nine stocks record higher Sharpe ratios during the overnight period versus the daytime period, with the mean overnight Sharpe ratio (0.662) substantially exceeding the mean daytime Sharpe ratio (0.357), a statistically and economically significant overnight premium. Second, the hybrid Strategy #18—Long Overnight coupled with Daytime Reversal—emerges as the dominant cross-asset configuration, generating portfolio values as high as $8464 from a $100 initial investment (AMGN; Sharpe: 0.991) over the 20-year horizon. Third, Trajectory Change Analysis reveals (i) Lévy-stable tails with a mean stability index α¯=1.667 across all constituents, substantially below the Gaussian benchmark of α=2.0; (ii) Hurst exponents clustering below 0.5 (H¯=0.417), confirming dominant mean-reverting dynamics; and (iii) positive rolling CAPM alpha in 51–79% of rolling windows, indicating persistent risk-adjusted outperformance above the S&P 500 benchmark. These findings provide a rigorous empirical foundation for session-aware algorithmic trading system design and challenge the prevailing assumption of temporal homogeneity in equity return processes.

This study illuminates fundamental questions in behavioral science through advanced machine learning methodologies applied to large-scale public opinion data. Drawing on Kahneman and Tversky’s dual-process theory and Sunstein’s nudge architecture, we employ hierarchical unsupervised clustering and supervised predictive models to detect cognitive biases—loss aversion, availability heuristic, and partisan motivated reasoning—embedded within a nationally representative survey of 5022 American respondents. Our primary methodological contribution is a hierarchical two-stage clustering framework that uncovers latent opinion structures without imposing a priori partisan categories, permitting discovery of cross-cutting cleavages invisible to conventional survey analysis. Three principal findings emerge: (1) loss aversion is empirically confirmed in prospective economic perception, with pessimists outnumbering optimists at a 1.14:1 ratio even among respondents rating current conditions positively; (2) partisan motivated reasoning produces a 13.15 percentage-point perception gap among individuals with identical financial circumstances; and (3) multi-platform digital engagement is associated with reduced partisan bias, providing evidence that challenges simple echo chamber assumptions. Crime safety perception emerges as the strongest predictor of economic bias, surpassing party affiliation, and substantiating availability heuristic dominance in political cognition. These findings carry implications for democratic accountability, platform governance, and the ethics of AI-augmented behavioral analysis in an era of affective polarization.

This study evaluates the effectiveness of a median sector rotation strategy within the Nikkei 500 component sectors, building on prior research that demonstrated superior risk-adjusted returns by selecting midperforming assets. Unlike traditional momentum-based investing, which focuses on winners or losers, the median strategy systematically reallocates capital to sectors with moderate past performance, reducing volatility while maintaining steady growth. Our findings reveal that quarterly and semi-annual rebalancing optimize returns in Japan, differing from U.S.-based studies where monthly rebalancing was more effective. Unlike buy-and-hold investing, the median strategy tends to outperforms total return and drawdown reduction, making it a viable alternative for public investors. By applying structured sector rotation rather than passive indexing, investors gain exposure to Japan’s strongest industries while mitigating downside risk. The results highlight the strategy’s adaptability across markets and suggest broader applications in global equities, fixed income, and multi-asset portfolios for enhanced portfolio resilience.

Purpose:Hand surgery decision-making requires integration of complex anatomical understanding, diverse patient-specific factors, and nuanced operative techniques. While artificial intelligence (AI), large language models (LLMs), and retrieval-augmented generation (RAG) models have advanced significantly in various fields, no AI-driven clinical decision support systems currently exist for hand surgery. A novel retrieval-enhanced AI large language model specifically tailored for hand surgery was developed, capable of effectively utilizing peer-reviewed published hand surgery literature for clinical decision support in real-time at point of care. Methods:An AI clinical decision support system was developed integrating all available open-access 4510 peer-reviewed hand surgery publications from 2000 to 2024 identified through hand surgery-relevant keywords. Documents were processed using a hierarchical pipeline based on the RAPTOR methodology, which breaks down large texts into smaller segments to enhance accurate retrieval. The system was evaluated using 15 standardized clinical queries assessed using automated computational metrics for correctness and semantic similarity to source documents. Results:The AI system demonstrated consistent performance with an average G-Eval correctness score of 0.79, SEM with an average similarity score of 0.75 (range: 0.54-0.86) and average maximum similarity score of 0.80 (range: 0.56-0.91), predominantly at moderate confidence levels. Generated recommendations were contextually appropriate and reliably linked to relevant hand surgery literature, providing accurate and clinically meaningful guidance. Conclusion:The AI system, HandRAG, incorporating RAG and LLM approach offers potential benefits for evidence-based clinical decision support and education in hand surgery.

One of the most challenging problems in data analysis is visualizing patterns and extracting insights from multi-dimensional datasets that vary over time. The complexity of data and variations in the correlations between different features adds further difficulty to the analysis. In this paper, we provide a framework to analyze the temporal dynamics of such datasets. We use machine learning clustering techniques and examine the time evolution of data patterns by constructing the corresponding cluster trajectories. These trajectories allow us to visualize the patterns and the changing nature of correlations over time. The similarity and correlations of features are reflected in common cluster membership, whereas the historical dynamics are described by a trajectory in the corresponding (cluster, time) space. This allows an effective visualization of multi-dimensional data over time. We introduce several statistical metrics to measure duration, volatility, and inertia of changes in patterns. Using the Hamming distance of trajectories over multiple time periods, we propose a novel metric, the Hamming diversification index, to measure the spread between trajectories. The novel metric is easy to compute, has a simple machine learning implementation, and provides additional insights into the temporal dynamics of data. This parsimonious diversification index can be used to examine changes in pattern similarities over aggregated time periods. We demonstrate the efficacy of our approach by analyzing a complex multi-year dataset of multiple worldwide economic indicators.

Pharmaceutical manufacturing and logistics rely on accurate prediction and decision making to safeguard product quality, delivery reliability, and patient outcomes. Despite rapid advances in artificial intelligence (AI) and machine learning (ML), few studies benchmark model performance across the diverse operational demands of global pharmaceutical supply chains. Predictive setbacks contribute to financial losses, reduced supply chain efficacy, and potential adverse health consequences, yet understanding these failures offers firms opportunities to refine strategy and strengthen resilience. Drawing on 1.2 million shipments spanning 39 countries, we compare traditional statistical models (ARIMA), ensemble methods (random forests, gradient boosting), and deep neural networks (LSTM, GRU, CNN, ANN) across pricing, demand forecasting, vendor management, and shipment planning. Gradient boosting produced the strongest pricing performance, while ARIMA delivered the lowest demand-forecasting errors but with limited explanatory power; neural networks captured nonlinear demand shocks and achieved superior maintenance-risk classification. We also identified three vendor performance clusters—high-performing, cost-efficient, and mixed-reliability vendors—enabling firms to better align shipment criticality with vendor capabilities by prioritizing high performers for urgent deliveries, leveraging cost-efficient vendors for non-urgent volumes, and managing mixed performers through targeted oversight. These insights highlight the value of our evidence-based roadmap for selecting algorithms in high-stakes healthcare logistics, in rapidly evolving, technologically complex global contexts where increasing algorithmic sophistication elevates the standards for safer, smarter pharmaceutical supply chains.

BACKGROUND:Microsurgical decision-making requires integration of diverse patient-specific factors, advanced surgical techniques, and dynamic intraoperative insights. While artificial intelligence (AI), large language models (LLMs), and retrieval-augmented generation (RAG) models have advanced significantly in various fields, no AI-driven clinical decision support systems currently exist for microsurgery. We developed MicroRAG, the first AI-powered clinical decision support system specifically designed for microsurgery, capable of instantly providing evidence-based recommendations by searching and synthesizing the entire microsurgical literature. METHODS:We developed an AI clinical decision support system integrating 4876 peer-reviewed microsurgical publications (2000-2024) using advanced retrieval-augmented generation (RAG) technology. The system processes clinical queries through hierarchical document clustering and provides real-time, evidence-based recommendations with direct literature citations. We evaluated system performance using 10 standardized clinical scenarios covering common microsurgical decisions, measuring answer relevancy, faithfulness to source literature, and clinical accuracy. RESULTS:MicroRAG demonstrated exceptional performance with an average answer relevancy score of 0.953 (range: 0.857-1.000) and faithfulness score of 0.907 (range: 0.676-1.000). G-Eval correctness averaged 0.88 with Semantic Evaluation Metrics showing an average similarity score of 0.75 and confidence score of 0.80. The system successfully provided comprehensive, immediately actionable guidance for complex scenarios including free flap monitoring protocols, vascular complication management, and surgical technique selection. All responses were grounded in peer-reviewed literature with direct citations. CONCLUSION:MicroRAG represents a technological innovation in microsurgical practice, providing instant access to evidence-based recommendations that typically require hours of literature review. By delivering comprehensive, literature-grounded guidance in real-time, this system has the potential to standardize best practices, reduce decision-making uncertainty, and ultimately improve patient outcomes across all levels of surgical experience.

Income inequality has emerged as a defining challenge of our time, particularly in advanced economies, where the gap between rich and poor has reached unprecedented levels. This study analyzes income inequality trends from 2000 to 2023 across developed countries (the United States, the United Kingdom, Germany, and France) and developing nations using World Bank Gini coefficient data. We employ comprehensive visualization techniques, Pareto distribution analysis, and ARIMA time-series forecasting models to evaluate the effectiveness of the Kuznets curve as a predictor of income inequality. Our analysis reveals significant deviations from the traditional inverse U-shaped Kuznets curve across all examined countries, with persistent volatility rather than the predicted decline in inequality. Forecasts using ARIMA and neural networks indicate continued fluctuations in inequality through 2030, with the U.S. and Germany showing upward trends while France and the UK demonstrate relative stability. These findings challenge the conventional Kuznets hypothesis and demonstrate that contemporary inequality patterns are influenced by factors beyond economic development, including technological change, globalization, and policy choices. This research contributes to the literature by providing empirical evidence that the Kuznets curve has limited predictive power in modern economies, informing policymakers about the need for targeted interventions to address persistent inequality rather than relying on economic growth alone.

The periodic table, a fundamental tool in chemistry, has undergone a remarkable evolution from its early qualitative studies to the integration of modern machine learning applications. This paper delves into the historical journey of the periodic table, highlighting key events and contributions from renowned scientists such as Mendeleev, Moseley, and Bohr. Through their groundbreaking work, our understanding of the elements and their periodic trends has been significantly enhanced. The periodic table’s predictive power, rooted in the periodic law, has not only facilitated the systematic organization of elements but has also enabled the anticipation of properties of yet-to-be-discovered elements. With the advent of machine learning algorithms, researchers now have the capability to predict the properties of novel elements, optimize experimental conditions, and accelerate the discovery of new materials. This paper explores the enduring significance of the periodic table as a symbol of order and discovery in the field of chemistry, showcasing its continued relevance and utility in the context of modern scientific advancements and technological innovations.

Considering the United Nations Climate Change Accord and insights from the OECD Global Material Resources Outlook to 2060, this study explores the intricate interrelationships of population growth, economic expansion, energy consumption, and carbon emissions in key OECD and BRICS countries. With the global economy heavily reliant on fossil fuels—the primary drivers of carbon emissions—we examine historic and projected energy use trends in developed and emerging economies. Through a combination of exploratory data analysis and ARIMA-based statistical forecasting, we investigate the relationships among GDP growth, energy use, and emissions, drawing distinctions between OECD and BRICS nations. Our findings reveal that, while developed economies demonstrate declining energy use, emerging markets show an upsurge in usage tied to economic growth. This research presents a compelling case for transitioning to a low-carbon future, drawing on renewable energy sources and proposing a roadmap to achieve both economic resilience and environmental sustainability. Our work serves as a call to action for policy-driven, cleaner energy investments to curb emissions and safeguard the planet.

Marine pollution incidents pose significant threats to marine ecosystems and coastal communities across Pacific Island nations, necessitating advanced predictive capabilities for effective environmental management. This study analyzes 8133 marine pollution incidents from 2001–2014 across 25 Pacific Island nations to develop predictive models for pollution type classification, hotspot identification, and seasonal pattern forecasting. Our analysis reveals Papua New Guinea as the dominant pollution hotspot, experiencing 51.9% of all regional incidents, with plastic waste dumping comprising 78.8% of pollution events and exhibiting pronounced seasonal peaks during June (coinciding with critical fish breeding periods). Machine learning classification achieved 99.1% accuracy in predicting pollution types, with material composition emerging as the strongest predictor, followed by seasonal timing and geographic location. Temporal analysis identified distinct seasonal dependencies, with June representing peak pollution activity (755 average incidents), coinciding with vulnerable marine ecological periods. The predictive framework successfully distinguishes between persistent geographic hotspots and episodic pollution events, enabling targeted conservation interventions during high-risk periods. These findings demonstrate that pollution type and location are highly predictable from environmental and temporal variables, providing marine conservationists with tools to anticipate when and where pollution will most likely impact fish populations and ecosystem health. The study establishes the first comprehensive baseline for Pacific Island marine pollution patterns and validates machine learning approaches for proactive pollution monitoring, offering scalable solutions for protecting ocean ecosystems and supporting evidence-based policy formulation across the region.

The Massachusetts Bay Transportation Authority (MBTA) is the main public transit provider in Boston, operating multiple means of transport, including trains, subways, and buses. However, the system often faces delays and fluctuations in ridership volume, which negatively affect efficiency and passenger satisfaction. To further understand this phenomenon, this paper compares the performance of existing and unique methods to determine the best approach in predicting gated station entries in the subway system (a proxy for subway usage) and the number of delays in the overall MBTA system. To do so, this research considers factors that tend to affect public transportation, such as day of week, season, pressure, wind speed, average temperature, and precipitation. This paper evaluates the performance of 10 statistical and machine learning models on predicting next-day subway usage. On predicting delay count, the number of models is extended to 11 per day by introducing a self-exciting point process model, representing a unique application of a point-process framework for MBTA delay modeling. This research involves experimenting with the selective inclusion of features to determine feature importance, testing model accuracy via Root Mean Squared Error (RMSE). Remarkably, it is found that providing either day of week or season data has a more substantial benefit to predictive accuracy compared to weather data; in fact, providing weather data generally worsens performance, suggesting a tendency of models to overfit.

In recent years, there has been an increased interest in using the mean absolute deviation (MAD) around the mean and median (the L1 norm) as an alternative to standard deviation σ (the L2 norm). Till now, the MAD has been computed for some distributions. For other distributions, expressions for mean absolute deviations (MADs) are not available nor reported. Typically, MADs are derived using the probability density functions (PDFs). By contrast, we derive simple expressions in terms of the integrals of the cumulative distribution functions (CDFs). We show that MADs have simple geometric interpretations as areas under the appropriately folded CDF. As a result, MADs can be computed directly from CDFs by computing appropriate integrals or sums for both continuous and discrete distributions, respectively. For many distributions, these CDFs have a simpler form than PDFs. Moreover, the CDFs are often expressed in terms of special functions, and indefinite integrals and sums for these functions are well known. We compute MADs for many well-known continuous and discrete distributions. For some of these distributions, the expressions for MADs have not been reported. We hope this study will be useful for researchers and practitioners interested in MADs.

We often rely on human experts to assign true labels in medical datasets, which may not be 100% accurate. We investigate the impact of labeling errors on machine-learning classi- fiers applied to medical datasets. By introducing symmetric errors from 0% to 40% in True labels— simulating errors of true labels assignment by experts, inter-observer variability, and automated annotation - we evaluate the impact of such errors in binary classification for several well-known medical datasets using traditional machine learning models and metrics. Although all models experience degradation as errors increase, simpler, well-regularized methods such as Logistic Regression and SVM decline more gracefully. Our results underscore the necessity for improved data curation and error-aware training strategies in medical AI, ultimately guiding the selection of robust algorithms that maintain reliability under imperfect real-world conditions.

In the domain of digital advertising, a principal imperative is the precise identification and engagement of a target audience—comprising both extant consumers identified from historical data and potential prospects convertible into future patrons. A persistent and substantive challenge in this endeavor lies in constructing targeting constraints that not only capture existing behavioral patterns but also extrapolate toward highpropensity yet unobserved audience segments. This strategic expansion, commonly designated as audience extension, has conventionally been addressed through greedy cover algorithms, which prioritize audience volume to the exclusion of nuanced performance indicators. In this study, we present a methodological augmentation of the greedy framework by incorporating dual performance metrics—similarity and novelty—as evaluative criteria. The proposed algorithm introduces a multi-objective optimization framework that facilitates the judicious expansion of audience segments while preserving representational fidelity to the original cohort. We empirically substantiate the efficacy of our framework through multiple case studies, demonstrating its superiority in balancing quantitative performance with qualitative audience alignment.

In technical analysis-based algorithmic trading strategies, we use historical price patterns to predict future prices and trade accordingly. This is analogous to machine learning where we use the existing data patterns to classify or predict new patterns. This paper uses this analogy and explains trading strategies as a machine learning classification problem. We derive simple approximations that relate the performance of trading strategies to machine learning statistics. We introduce a new performance measure of the Return Efficiency Index. This index provides a link between trading strategy return statistics and classification accuracy. It has a simple geometric interpretation, similar to the ROC index in machine learning, and can be used to compare strategies in terms of their ability to capture the potential returns possible with the underlying assets. We illustrate the proposed approach by a detailed comparison of daily trading strategies designed by analogies to nearest neighbor classification widely used in machine learning and to some strategies based on deep learning.