|
on Big Data |
| By: | Helyette Geman |
| Abstract: | In 2017, eight scientists from the Google research team published in the journal Advances in Neural Info Processing Systems the remarkable article “Attention is all you need, ” which introduced a Transformer neural network architecture. The paper has been cited over 173, 000 times and ranks among the top 100 most cited papers of the 21st century. It builds on the attention principle introduced in 2014 by Bahdanau, Cho and Turing Award winner Bengio, who proposed neural machine translation by jointly learning to align and translate. This transformer approach has become the main architecture for a wide variety of AI tasks, including large language models. In machine learning, “attention” refers to a mechanism that allows models to focus on specific parts of the input data during the learning process and to determine the relative importance of each component within a sequence. Turning to financial economics, financial news—whether in terms of volume, unusual frequency, or sentiment (positive versus negative tone)—has long attracted the attention of researchers seeking to forecast market dynamics—“buy on rumors, sell on news.” Financial bubbles, however, remain among the most challenging phenomena to model and trade. Traditional models relying solely on price dynamics often fail to detect bubbles in real time, a key objective for stock picking and portfolio selection. Advances in natural language processing (NLP) now enable researchers to quantify market attention and sentiment using financial news and social media activity. This paper builds on recent research on sentiment in financial markets and integrates these insights into quantitative bubble detection models derived from the Log-Periodic Power Law (LPPL) literature, while incorporating a Hype Index that measures disproportionate news attention at a given moment, in order to obtain a hype-adjusted view of speculative dynamics. Within this framework, sentiment and news intensity modify bubble scores derived from price dynamics. The resulting Hyped Log-Periodic Power Law (HLPPL) model improves the identification of emerging bubbles and enables the detection of negative bubbles, corresponding to temporarily overvalued assets. The approach further highlights the importance of the choice of numéraire with respect to which prices are expressed (e.g., gold versus the dollar), emphasizing that bubbles must be assessed relative to a chosen reference asset. Empirical illustrations across equities and cryptocurrencies show how media attention and narrative amplification interact with price dynamics during speculative episodes. Taken together, these results suggest that incorporating information flows and market narratives can significantly improve the early detection and interpretation of financial bubbles. |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:ocp:pbcoen:pb27_26 |
| By: | Yuqi Li; Siyuan Liu; Bingjun Liu |
| Abstract: | While deep learning has excelled in various domains, its application to sequential decision-making in finance remains challenging due to the low Signal-to-Noise Ratio (SNR) and non-stationarity of financial data. Leveraging the reasoning capabilities of Large Language Models (LLMs), we propose \textbf{PandaAI}, a closed-loop neuro-symbolic LLM agent with market regime modeling and constrained alpha generation, which bridges general LLM reasoning with financial rigor and suppresses the financial toxicity of LLM-generated outputs. To bridge the gap between general linguistic capability and financial rigor, we fine-tune a domain-specific LLM. Furthermore, we integrate this LLM into a modular architecture and form a closed-loop system. Unlike traditional models that optimize isolated prediction metrics, \textbf{PandaAI} is designed as a neuro-symbolic agent that navigates the complex, real-world financial environment with explicit risk awareness. Extensive experiments on CSI 300 stock data show that \textbf{PandaAI} achieves a $18.2\%$ higher Rank IC and $25.7\%$ lower maximum drawdown than state-of-the-art time-series models. Our constrained LLM generation and dual-channel adaptation method provide a general paradigm for LLM deployment in high-stakes sequential decision-making scenarios. |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2606.06823 |
| By: | Delgado-Téllez, Mar; Ceglar, Andrej; Spiteri, Sarah; Lebouteiller, Léonore; Vorderobermeier, Nicole |
| Abstract: | Europe is increasingly exposed to heat waves and droughts, but their short-term economic effects across sectors remain hard to predict. This study develops climate-augmented models to predict real growth in per capita value added across 1, 117 EU regions (2002–2022), by combining economic indicators with high-frequency climate data. When using machine learning (ML, Random Forest and XGBoost), climate variables improve predictions in agriculture, while gains for other sectors are limited and do not outperform economic models. Heat wave indicators consistently enhance predictive performance, whereas drought effects vary by sector. Simulations of extreme combined heat and drought scenarios suggest that agricultural annual growth could fall by 1.9 to 7.6 percentage points in most regions, whereas industry, and manufacturing in particular, is less affected, although impacts are more pronounced in Eastern Europe and the Baltic states. Overall, ML models better reflect complex climate–economic interactions, supporting their use for early warning, policy planning, and targeted adaptation. JEL Classification: C53, E37, Q54, R15 |
| Keywords: | climate extremes, machine learning, production, regional predictions |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:ecb:ecbwps:20263248 |
| By: | Albanese, Andrea (Luxembourg Institute of Socio-Economic Research (LISER)); Marguerit, David (Luxembourg Institute of Socio-Economic Research (LISER)) |
| Abstract: | Cross-border work is expanding in the EU, yet its labor-market effects on the cross-border workers themselves remain largely undocumented. Using linked Belgian administrative registers that identify cross-border spells in Luxembourg, we estimate the effects of cross-border employment on post-return labor-market outcomes through dynamic double machine learning. Returnees face a short-run employment penalty that fades with cross-border tenure and time since return. They are also more likely to receive Belgian unemployment benefits than comparable stayers, with higher daily benefit levels among recipients. |
| Keywords: | cross-border commuting, return migration, unemployment insurance, EU labor mobility, administrative data |
| JEL: | J61 J65 J64 I38 C21 C14 |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18674 |
| By: | Andreas Aigner |
| Abstract: | We present a hybrid news sentiment engine that continuously learns market sentiment from paired news headlines and concurrent asset-price snapshots without requiring any neural network training or GPU compute. The system uses a three-way ensemble combining (1) a financial-domain lexicon (FinBERT-style keyword scoring), (2) an adaptive statistical TF-IDF cluster learner that organizes headlines into semantic neighborhoods and tracks their average realized price reactions, and (3) an auto-calibrating weighting mechanism that adjusts ensemble contributions based on each signal's historical correlation with actual price movements. The engine runs on a 3-hour polling cycle from the Tradeflags NewsFeed API, which provides 22 price-snapshot fields per news item spanning equity indices (ES, NQ, SPY, DJIA, NDX, IWM), commodities (CL), and cryptocurrencies (BTC, ETH). All processing occurs at sub-second latency on a CPU-only server at effectively zero marginal cost per analytic cycle. We compare our approach against established methods -- FinBERT, GPT-based scoring, VADER, and commercial sentiment APIs -- across dimensions of cost, latency, accuracy, and adaptability. Our statistical cluster learner, which adapts to changing market regimes without retraining, represents a novel contribution not found in existing sentiment systems. |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2606.03457 |
| By: | Ferrara, Andreas (University of Pittsburgh, Department of Economics, and NBER) |
| Abstract: | Large language models (LLMs) are lowering the entry barriers to working with exciting data sources that used to require strong data science skills, such as handwritten ledgers, text, images, or sound recordings. This guide provides an introduction for researchers who are new to LLMs. It sets out a step-by-step workflow for turning a research idea into working code and data, and describes the four main ways of interacting with an LLM: the chat window, editor-integrated assis tants, agentic coding tools, and the API. It then works through the decisions a practitioner meets in sequence, beginning with whether an LLM is the right tool and whether the data are allowed to be sent to one, then how to select models, write prompts, manage context limits, and control costs, and finally how to validate, reproduce, document, and correct LLM-generated measures in regression settings. A review of recent research shows how these tools already extract, link, har monize, and classify historical data at scale. Four worked examples with replication files illustrate the use of LLMs. They classify emotions in paintings, link census records without names, measure newspaper salience and sentiment around the 1882 Chinese Exclusion Act, and score the emotional delivery of Franklin D. Roosevelt's wartime speeches. The guide also condenses the workflow, the best-practice recommendations, and the preparation of replication packages into summary tables and checklists to aid applied economists. |
| Keywords: | Large Language Models, Artificial Intelligence, Economic History, Practitioner's Guide JEL Classification: C8, N0, C55 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:cge:wacage:810 |
| By: | Bryan T. Kelly; Semyon Malamud; Johannes Schwab; Teng Andrea Xu |
| Abstract: | Large language models trained on unrestricted internet corpora inevitably embed information from the future, introducing lookahead bias that compromises the validity of backtests and causal inference in finance and the social sciences. Point-in-time language models—trained exclusively on text available up to each calendar date—eliminate this leakage by construction, but existing efforts typically produce models that lag substantially behind their unconstrained counterparts. We show that this performance gap can be narrowed through scale. Training decoder-only transformers with up to 4 billion parameters on 1 trillion chronologically filtered tokens from FineWeb, we construct a sequence of monthly model checkpoints spanning 2013–2024. Across a range of common-sense reasoning and language understanding benchmarks, our models approach the performance of leading open-weight models of comparable size (such as Gemma-3-4B and LLaMA-7B) trained on temporally unrestricted data, although a performance gap remains on several tasks. Finally, in a strict out-of-sample economic evaluation task, portfolios built from point-in-time embeddings achieve robust positive Sharpe ratios and perform close to full-sample counterparts that violate temporal validity, indicating that chronologically consistent language models can extract economically meaningful signals without relying on look-ahead bias. We release the complete pipeline—including dataset construction, training infrastructure, and evaluation code—to enable reproducible point-in-time language modeling and to support research applications that require strict temporal validity. |
| JEL: | C14 C45 G11 G14 G17 |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35247 |
| By: | Vanesa Jordá (Department of Economics, Cantabria University, Avda. de los Castros, 56. 39005 Santander, Spain); Miguel Niño-Zarazúa (Department of Economics, SOAS University of London, Thornhaugh Street, Russell Square, London WC1H 0XG, UK) |
| Abstract: | Reliable measurement of income and consumption is essential for monitoring poverty and inequality in low- and middle-income countries, yet full household surveys are costly and difficult to implement regularly. This paper examines whether reduced survey instruments can preserve key distributional information. We apply Random Forest Recursive Feature Elimination (RF-RFE) to the 2018/19 Nigeria General Household Survey-Panel to identify the income sources, consumption categories and household characteristics that best classify individuals within the welfare distribution. The analysis focuses on three outcomes: poverty status, location in the quintile distribution and position relative to the Gini-based inequality line. The survey's post-planting and post-harvest periods allow us to assess performance under different seasonal contexts. Results show that RF-RFE achieves strong classification accuracy with few predictors. For consumption, poverty status and inequality-line position are accurately predicted using a small set of expenditure categories, while quintile classification reaches about 80 percent accuracy for seasonal consumption and 60-65 percent for annual consumption predicted from a single seasonal visit. For income, poverty status reaches around 90 percent accuracy with five predictors, and inequality-line position is largely captured by labour earnings. The findings suggest that machine-learning methods can help improve survey design and reduce data requirements while retaining much of the distributional information needed to measure and monitor poverty and inequality. |
| Keywords: | consumption measurement, poverty, inequality, Random Forests, household surveys, Nigeria |
| JEL: | C38 C55 D31 I32 O55 |
| URL: | https://d.repec.org/n?u=RePEc:soa:wpaper:275 |
| By: | Andrei Bysik; Robert \'Slepaczuk |
| Abstract: | This paper investigates whether machine learning forecasts of hourly BTC-USDT returns can be converted into economically meaningful trading performance after transaction costs. Using approximately 70, 000 hourly observations from 2018-2026, XGBoost, LSTM, and iTransformer are evaluated in a 27-fold walk-forward protocol. All three models produce positive gross trading performance in selected configurations, but naive sign-based strategies fail once transaction costs of ten basis points are imposed. A cost-aware execution filter, which prevents trades only when the forecast magnitude exceeds a transaction-cost-based threshold, sharply reduces turnover and restores profitability in selected configurations. The strongest long-only XGBoost strategy produces annualised returns above 65% with a Sharpe ratio above one. Additional tests show that technical indicators improve performance in selected cases, EGARCH-derived features do not provide uniformly robust gains, and XGBoost is descriptively stronger than the neural alternatives, although bootstrap evidence does not support formal statistical dominance. Loss-function and model-selection effects are secondary and statistically fragile. The results show that the main obstacle in hourly cryptocurrency trading is not only weak predictability, but also the way forecasts are converted into trades. |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2606.00060 |