|
on Big Data |
| By: | Robert-Paul Berben; Rajni Rasiawan; Jasper de Winter |
| Abstract: | This paper examines the performance of machine learning models in forecasting Dutch inflation over the period 2010 to 2023, leveraging a large dataset and a range of machine learning techniques. The findings indicate that certain machine learning models outperform simple benchmarks, particularly in forecasting core inflation and services inflation. However, these models face challenges in consistently outperforming the primary inflation forecast of De Nederlandsche Bank for headline inflation, though they show promise in improving the forecast for non-energy industrial goods inflation. Models employing path averages rather than direct forecasting achieve greater accuracy, while the inclusion of non-linearities, factors, or targeted predictors provides minimal or no improvement in forecasting performance. Overall, Ridge regression has the best forecasting performance in our study. |
| Keywords: | Inflation forecasting; Big data; Machine learning; Random Forest; Ridge regression |
| JEL: | C22 C53 C55 E17 E31 |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:dnb:dnbwpp:828 |
| By: | Fernando Perez-Cruz; Hyun Song Shin |
| Abstract: | When posed with a logical puzzle that demands reasoning about the knowledge of others and about counterfactuals, large language models (LLMs) display a distinctive and revealing pattern of failure. The LLM performs flawlessly when presented with the original wording of the puzzle available on the internet but performs poorly when incidental details are changed, suggestive of a lack of true understanding of the underlying logic. Our findings do not detract from the considerable progress in central bank applications of machine learning to data management, macro analysis and regulation/supervision. They do, however, suggest that caution should be exercised in deploying LLMs in contexts that demand rigorous reasoning in economic analysis. |
| Date: | 2024–01–04 |
| URL: | https://d.repec.org/n?u=RePEc:bis:bisblt:83 |
| By: | Jairo Flores (Central Reserve Bank of Peru); Bruno Gonzaga (Central Reserve Bank of Peru); Walter Ruelas-Huanca (Central Reserve Bank of Peru); Juan Tang (Central Reserve Bank of Peru) |
| Abstract: | This paper explores the application of machine learning (ML) techniques to nowcast the monthly year-over-year growth rate of both total and non-primary GDP in Peru. Using a comprehensive dataset that includes over 170 domestic and international predictors, we assess the predictive performance of 12 ML models. The study compares these ML approaches against the traditional Dynamic Factor Model (DFM), which serves as the benchmark for nowcasting in economic research. We treat specific configurations, such as the feature matrix rotations and the dimensionality reduction technique, as hyperparameters that are optimized iteratively by the Tree-Structured Parzen Estimator. Our results show that ML models outperformed DFM in nowcasting total GDP, and that they achieve similar performance to this benchmark in nowcasting non-primary GDP. Furthermore, the bottom-up approach appears to be the most effective practice for nowcasting economic activity, as aggregating sectoral predictions improves the precision of ML methods. The findings indicate that ML models offer a viable and competitive alternative to traditional nowcasting methods. |
| Keywords: | GDP; Machine Learning; nowcasting |
| JEL: | C14 C32 E32 E52 |
| Date: | 2025–02–03 |
| URL: | https://d.repec.org/n?u=RePEc:gii:giihei:heidwp01-2025 |
| By: | Konstantinos-Leonidas Bisdoulis |
| Abstract: | Fluctuations in the stock market rapidly shape the economic world and consumer markets, impacting millions of individuals. Hence, accurately forecasting it is essential for mitigating risks, including those associated with inactivity. Although research shows that hybrid models of Deep Learning (DL) and Machine Learning (ML) yield promising results, their computational requirements often exceed the capabilities of average personal computers, rendering them inaccessible to many. In order to address this challenge in this paper we optimize LightGBM (an efficient implementation of gradient-boosted decision trees (GBDT)) for maximum performance, while maintaining low computational requirements. We introduce novel feature engineering techniques including indicator-price slope ratios and differences of close and open prices divided by the corresponding 14-period Exponential Moving Average (EMA), designed to capture market dynamics and enhance predictive accuracy. Additionally, we test seven different feature and target variable transformation methods, including returns, logarithmic returns, EMA ratios and their standardized counterparts as well as EMA difference ratios, so as to identify the most effective ones weighing in both efficiency and accuracy. The results demonstrate Log Returns, Returns and EMA Difference Ratio constitute the best target variable transformation methods, with EMA ratios having a lower percentage of correct directional forecasts, and standardized versions of target variable transformations requiring significantly more training time. Moreover, the introduced features demonstrate high feature importance in predictive performance across all target variable transformation methods. This study highlights an accessible, computationally efficient approach to stock market forecasting using LightGBM, making advanced forecasting techniques more widely attainable. |
| Date: | 2024–12 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2501.07580 |
| By: | Magomedov, Said; Fantazzini, Dean |
| Abstract: | The popularity of cryptocurrency exchanges has surged in recent years, accompanied by the proliferation of new digital platforms and tokens. However, the issue of credit risk and the reliability of crypto exchanges remain critical, highlighting the need for indicators to assess the safety of investing through these platforms. This study examines a unique, hand-collected dataset of 228 cryptocurrency exchanges operating between April 2011 and May 2024. Using various machine learning algorithms, we identify the key factors contributing to exchange shutdowns, with trading volume, exchange lifespan, and cybersecurity scores emerging as the most significant predictors. Since individual machine learning models often capture distinct data characteristics and exhibit varying error patterns, we employ a forecast combination approach by aggregating multiple predictive distributions. Specifically, we evaluate several specifications of the generalized linear pool (GLP), beta-transformed linear pool (BLP), and beta-mixture combination (BMC). Our findings reveal that the beta-transformed linear pool and the beta-mixture combination achieve the best performances, improving forecast accuracy by approximately 4.1% based on a robust H-measure, which effectively addresses the challenges of misclassification in imbalanced datasets. |
| Keywords: | forecast combination; exchange; bitcoin; crypto assets; cryptocurrencies; credit risk; bankruptcy; default probability |
| JEL: | C35 C51 C53 C58 G12 G17 G32 G33 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:pra:mprapa:123416 |
| By: | Адилханова Зарина // Adilkhanova Zarina (National Bank of Kazakhstan); Ержан Ислам // Yerzhan Islam (National Bank of Kazakhstan) |
| Abstract: | В условиях нестабильной макроэкономической среды повышение точности прогнозирования инфляции является приоритетной задачей для центральных банков, особенно тех, которые придерживаются режима инфляционного таргетирования. Традиционные эконометрические модели сталкиваются с ограничениями при учёте волатильности, внешних шоков и нелинейных взаимосвязей. Данное исследование направлено на улучшение прогнозирования инфляции путём интеграции методов машинного обучения в существующую систему селективно-комбинированного прогнозирования инфляции. Включение таких алгоритмов, как Ridge Regression, Lasso Regression и Elastic Net, позволяет выявлять сложные паттерны в макроэкономических данных и повышать точность прогнозов. Сравнительный анализ прогнозов, полученных с использованием традиционных эконометрических моделей (OLS, LTAR, BVAR, RW) и алгоритмов машинного обучения, показывает, что гибридный подход значительно снижает ошибки прогнозирования и повышает надёжность прогнозов в краткосрочном периоде. Полученные результаты могут внести вклад в совершенствование инструментов макроэкономического прогнозирования и развитие более эффективной денежно-кредитной политики, поддерживая качество принятия решений центральными банками. // In an environment of macroeconomic instability, improving the accuracy of inflation forecasting is a priority for central banks, especially those operating under inflation targeting regimes. Traditional econometric models face limitations in accounting for volatility, external shocks, and nonlinear relationships. This study aims to enhance inflation forecasting by integrating machine learning methods into the existing Selective-Combined Inflation Forecasting System (SSCIF). The inclusion of algorithms such as Ridge Regression, Lasso Regression, and Elastic Net enables the identification of complex patterns in macroeconomic data, thereby improving forecast accuracy. A comparative analysis of forecasts generated using traditional econometric models (OLS, LTAR, BVAR, RW) and machine learning algorithms demonstrates that the hybrid approach significantly reduces forecasting errors and enhances the reliability of short-term forecasts. The results contribute to the advancement of macroeconomic forecasting tools and the development of more effective monetary policy, supporting better decision-making by central banks. |
| Keywords: | инфляция, прогнозирование, индекс потребительских цен, модель, машинное обучение, эконометрические модели, точность прогнозов, inflation, forecasting, consumer price index, model, machine learning, econometric models, forecast accuracy |
| JEL: | E31 E37 C52 C61 |
| Date: | 2024 |
| URL: | https://d.repec.org/n?u=RePEc:aob:wpaper:62 |
| By: | Laura Battaglia (Oxford University); Timothy Christensen (Yale University); Stephen Hansen (UCL, IFS, and CEPR); Szymon Sacher (Meta) |
| Abstract: | It has become common practice for researchers to use AI-powered information retrieval algorithms or other machine learning methods to estimate variables of economic interest, then use these estimates as covariates in a regression model. We show both theoretically and empirically that naively treating AI- and ML-generated variables as ÒdataÓ leads to biased estimates and invalid inference. We propose two methods to correct bias and perform valid inference: (i) an explicit bias correction with bias-corrected confidence intervals, and (ii) joint maximum likelihood estimation of the regression model and the variables of interest. Through several applications, we demonstrate that the common approach generates substantial bias, while both corrections perform well. |
| Date: | 2025–01–02 |
| URL: | https://d.repec.org/n?u=RePEc:cwl:cwldpp:2421 |
| By: | Bryan T. Kelly (Yale SOM; AQR Capital Management, LLC; National Bureau of Economic Research (NBER)); Boris Kuznetsov (Swiss Finance Institute); Semyon Malamud (Ecole Polytechnique Federale de Lausanne; Centre for Economic Policy Research (CEPR); Swiss Finance Institute); Teng Andrea Xu (École Polytechnique Fédérale de Lausanne (EPFL)) |
| Abstract: | The core statistical technology in artificial intelligence is the large-scale transformer network. We propose a new asset pricing model that implants a transformer in the stochastic discount factor. This structure leverages conditional pricing information via cross-asset information sharing and nonlinearity. We also develop a linear transformer that serves as a simplified surrogate from which we derive an intuitive decomposition of the transformer's asset pricing mechanisms. We find large reductions in pricing errors from our artificial intelligence pricing model (AIPM) relative to previous machine learning models and dissect the sources of these gains. |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:chf:rpseri:rp2508 |
| By: | Ramin Mousa; Meysam Afrookhteh; Hooman Khaloo; Amir Ali Bengari; Gholamreza Heidary |
| Abstract: | Digital currencies have become popular in the last decade due to their non-dependency and decentralized nature. The price of these currencies has seen a lot of fluctuations at times, which has increased the need for prediction. As their most popular, Bitcoin(BTC) has become a research hotspot. The main challenge and trend of digital currencies, especially BTC, is price fluctuations, which require studying the basic price prediction model. This research presents a classification and regression model based on stack deep learning that uses a wavelet to remove noise to predict movements and prices of BTC at different time intervals. The proposed model based on the stacking technique uses models based on deep learning, especially neural networks and transformers, for one, seven, thirty and ninety-day forecasting. Three feature selection models, Chi2, RFE and Embedded, were also applied to the data in the pre-processing stage. The classification model achieved 63\% accuracy for predicting the next day and 64\%, 67\% and 82\% for predicting the seventh, thirty and ninety days, respectively. For daily price forecasting, the percentage error was reduced to 0.58, while the error ranged from 2.72\% to 2.85\% for seven- to ninety-day horizons. These results show that the proposed model performed better than other models in the literature. |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2501.13136 |
| By: | Jesús Villota (CEMFI, Centro de Estudios Monetarios y Financieros) |
| Abstract: | Markets do not always efficiently incorporate news, particularly when information is complex or ambiguous. Traditional text analysis methods fail to capture the economic structure of information and its firm-specific implications. We propose a novel methodology that guides LLMs to systematically identify and classify firm-specific economic shocks in news articles according to their type, magnitude, and direction. This economically-informed classification allows for a more nuanced understanding of how markets process complex information. Using a simple trading strategy, we demonstrate that our LLM-based classification significantly outperforms a benchmark based on clustering vector embeddings, generating consistent profits out-of-sample while maintaining transparent and durable trading signals. The results suggest that LLMs, when properly guided by economic frameworks, can effectively identify persistent patterns in how markets react to different types of firm-specific news. Our findings contribute to understanding market efficiency and information processing, while offering a promising new tool for analyzing financial narratives. |
| Keywords: | Large language models, business news, stock market reaction, market efficiency. |
| JEL: | G12 G14 C45 C58 C63 D83 |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:cmf:wpaper:wp2025_2501 |
| By: | Nicola Stalder (University of Bern); Michael Mayer (Schweizerische Mobiliar Versicherungsgesellschaft); Steven C. Bourassa (University of Washington); Martin Hoesli (University of Geneva - Geneva School of Economics and Management (GSEM); Swiss Finance Institute; University of Aberdeen - Business School) |
| Abstract: | This paper describes how machine learning techniques and explainable artificial intelligence can be leveraged to estimate combined location value. We analyze listed apartment rents using gradient boosted trees, which allow for flexible modelling of non-linear effects and high order interactions among covariates. We then separate location value from structure value by imposing interaction constraints. Finally, we use the additivity property of SHapley Additive exPlanations (SHAP) to extract the combined effects of location-related covariates. These effects are then compared across different geographical levels (regional and national). The empirical analysis uses a rich dataset consisting of listed rents and property characteristics for approximately 300, 000 apartments in Switzerland. We start with an unconstrained model that allows for flexible interactions between location variables and structural characteristics. We then impose interaction constraints such that structural characteristics no longer interact with location variables or each other. This step is required to extract the pure value of location independent of any interactions with structural characteristics. The constrained model improves interpretability while retaining a high degree of accuracy. What would otherwise be a cumbersome calibration of locational values is replaced by a simple extraction of the corresponding feature effects using SHAP. The results should prove useful in improving hedonic models used by property tax assessors, mortgage underwriters, valuation firms, and regulatory authorities. |
| Keywords: | Hedonic models, SHAP values, location values, explainable artificial intelligence, machine learning, gradient boosting |
| JEL: | R31 G12 |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:chf:rpseri:rp2502 |
| By: | Harald Mayr; Mateus Souza |
| Abstract: | Without heat metering, households face strong free-riding incentives. Using data from Swiss households, we find that the staggered introduction of submetering reduced heating expenses by 17%, on average. Machine learning techniques reveal highly heterogeneous effects, consistent with coordination failure in larger buildings and strategic exit of free-riders. We find that households are price elastic even when they share a common heating bill. Our results suggest that most households do not exploit the free-riding incentive, especially in smaller buildings. “Schmeduling, ” inattention to the billing regime, and pro-social behavior can explain the low prevalence of free-riding. Nevertheless, submetering is welfare-improving for most buildings. |
| Keywords: | Free-riding, submetering, individual billing, heating energy, tragedy of the commons, welfare |
| JEL: | D61 Q41 Q52 |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:bon:boncrc:crctr224_2025_629 |
| By: | Xia Li; Hanghang Zheng; Xiao Chen; Hong Liu; Mao Mao |
| Abstract: | The advent of artificial intelligence has significantly enhanced credit scoring technologies. Despite the remarkable efficacy of advanced deep learning models, mainstream adoption continues to favor tree-structured models due to their robust predictive performance on tabular data. Although pretrained models have seen considerable development, their application within the financial realm predominantly revolves around question-answering tasks and the use of such models for tabular-structured credit scoring datasets remains largely unexplored. Tabular-oriented large models, such as TabPFN, has made the application of large models in credit scoring feasible, albeit can only processing with limited sample sizes. This paper provides a novel framework to combine tabular-tailored dataset distillation technique with the pretrained model, empowers the scalability for TabPFN. Furthermore, though class imbalance distribution is the common nature in financial datasets, its influence during dataset distillation has not been explored. We thus integrate the imbalance-aware techniques during dataset distillation, resulting in improved performance in financial datasets (e.g., a 2.5% enhancement in AUC). This study presents a novel framework for scaling up the application of large pretrained models on financial tabular datasets and offers a comparative analysis of the influence of class imbalance on the dataset distillation process. We believe this approach can broaden the applications and downstream tasks of large models in the financial domain. |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2501.10677 |
| By: | Pundit, Madhavi (Asian Development Bank); Ramayandi, Arief (Asian Development Bank Institute); Simba, Patrick Jaime (Asian Development Bank); Sorino, Dennis (Asian Development Bank); Tan, Sharyl Rose (Asian Development Bank) |
| Abstract: | Timely updates of business cycle fluctuations—commonly represented by movements in the output gap—help policymakers make informed decisions on the appropriate course of action. Unfortunately, business cycle assessments often suffer from lags in actual gross domestic product data releases. This paper proposes the use of an Economic Activity Index, developed to monitor monthly business cycle fluctuations in Asia. The index summarizes a selection of updated monthly economic indicators to gauge movements in the output gap. The paper shows that the application of machine learning models substantially improves the ability of the index to track actual fluctuations of the business cycle compared with models constructed using a traditional principal component analysis. Grouping the information used to construct the index into six categories—consumption, investment, trade, government, financial, and the external sector— makes it possible to break down and explain drivers of movements in the business cycle. |
| Keywords: | macroeconomic monitoring; tracking business cycles; economic fluctuations; nowcasting |
| JEL: | C32 C63 E32 E37 |
| Date: | 2025–01–31 |
| URL: | https://d.repec.org/n?u=RePEc:ris:adbewp:0766 |
| By: | Shijie Han; Changhai Zhou; Yiqing Shen; Tianning Sun; Yuhua Zhou; Xiaoxia Wang; Zhixiao Yang; Jingshu Zhang; Hongguang Li |
| Abstract: | Current financial Large Language Models (LLMs) struggle with two critical limitations: a lack of depth in stock analysis, which impedes their ability to generate professional-grade insights, and the absence of objective evaluation metrics to assess the quality of stock analysis reports. To address these challenges, this paper introduces FinSphere, a conversational stock analysis agent, along with three major contributions: (1) Stocksis, a dataset curated by industry experts to enhance LLMs' stock analysis capabilities, (2) AnalyScore, a systematic evaluation framework for assessing stock analysis quality, and (3) FinSphere, an AI agent that can generate high-quality stock analysis reports in response to user queries. Experiments demonstrate that FinSphere achieves superior performance compared to both general and domain-specific LLMs, as well as existing agent-based systems, even when they are enhanced with real-time data access and few-shot guidance. The integrated framework, which combines real-time data feeds, quantitative tools, and an instruction-tuned LLM, yields substantial improvements in both analytical quality and practical applicability for real-world stock analysis. |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2501.12399 |
| By: | Yuxi Hong |
| Abstract: | Accurate stock market prediction provides great opportunities for informed decision-making, yet existing methods struggle with financial data's non-linear, high-dimensional, and volatile characteristics. Advanced predictive models are needed to effectively address these complexities. This paper proposes a novel multi-layer hybrid multi-task learning (MTL) framework aimed at achieving more efficient stock market predictions. It involves a Transformer encoder to extract complex correspondences between various input features, a Bidirectional Gated Recurrent Unit (BiGRU) to capture long-term temporal relationships, and a Kolmogorov-Arnold Network (KAN) to enhance the learning process. Experimental evaluations indicate that the proposed learning structure achieves great performance, with an MAE as low as 1.078, a MAPE as low as 0.012, and an R^2 as high as 0.98, when compared with other competitive networks. |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2501.09760 |
| By: | Vishalie Shah; Julia Hatamyar; Taufik Hidayat; Noemi Kreif |
| Abstract: | This paper uses instrumental causal forests, a novel machine learning method, to explore the treatment effect heterogeneity of Indonesia's conditional cash transfer scheme on maternal health care utilisation. Using randomised programme assignment as an instrument for enrollment in the scheme, we estimate conditional local average treatment effects for four key outcomes: good assisted delivery, delivery in a health care facility, pre-natal visits, and post-natal visits. We find significant treatment effect heterogeneity by supply-side characteristics, even though supply-side readiness was taken into account during programme development. Mothers in areas with more doctors, nurses, and delivery assistants were more likely to benefit from the programme, in terms of increased rates of good assisted delivery outcome. We also find large differences in benefits according to indicators of household poverty and survey wave, reflecting the possible impact of changes in programme design in its later years. The impact on post-natal visits in 2013 displayed the largest heterogeneity among all outcomes, with some women less likely to attend post-natal check ups after receiving the cash transfer in the long term. |
| Date: | 2025–01 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2501.12803 |
| By: | Avner Seror (Aix Marseille Univ, CNRS, AMSE, Marseille, France) |
| Abstract: | As large language models (LLMs) become integrated to decision-making across various sectors, a key question arises: do they exhibit an emergent "moral mind" - a consistent set of moral principles guiding their ethical judgments - and is this reasoning uniform or diverse across models? To investigate this, we presented about forty different models from the main providers with a large array of structured ethical scenarios, creating one of the largest datasets of its kind. Our rationality tests revealed that at least one model from each provider demonstrated behavior consistent with stable moral principles, effectively acting as approximately optimizing a utility function encoding ethical reasoning. We identified these utility functions and observed a notable clustering of models around neutral ethical stances. To investigate variability, we introduced a novel non-parametric permutation approach, revealing that the most rational models shared 59% to 76% of their ethical reasoning patterns. Despite this shared foundation, differences emerged: roughly half displayed greater moral adaptability, bridging diverse perspectives, while the remainder adhered to more rigid ethical structures. |
| Keywords: | Decision Theory, revealed preference, Rationality, artificial intelligence, LLM, PSM. |
| JEL: | D9 C9 C44 |
| Date: | 2024–11 |
| URL: | https://d.repec.org/n?u=RePEc:aim:wpaimx:2433 |