|
on Big Data |
| By: | David H. Kreitmeir (SoDa Labs, Monash University); Paul A. Raschky (Department of Economics and SoDa Laboratories, Monash University) |
| Abstract: | We analyse the effects of the ban of ChatGPT, a generative pre-trained transformer chatbot, on individual productivity. We first compile data on the hourly coding output of over 8, 000 professional GitHub users in Italy and other European countries to analyse the impact of the ban on individual productivity. Combining the high-frequency data with the sudden announcement of the ban in a difference-in-differences framework, we find that the output of Italian developers decreased by around 50\% in the first two business days after the ban and recovered after that. Applying a synthetic control approach to daily Google search and Tor usage data shows that the ban led to a significant increase in the use of censorship bypassing tools. Our findings show that users swiftly implement strategies to bypass Internet restrictions but this adaptation activity creates short-term disruptions and hampers productivity. |
| Keywords: | chatgpt, productivity, internet, censorship, italy |
| JEL: | D72 D83 L86 L88 |
| Date: | 2023–04 |
| URL: | http://d.repec.org/n?u=RePEc:ajr:sodwps:2023-01&r=big |
| By: | Cavaillé, Charlotte; Van Der Straeten, Karine; Chen, Daniel L. |
| Abstract: | Survey design often approximates a prediction problem: the goal is to select instruments that best predict the value of an unobserved construct or a future outcome. We demonstrate how advances in machine learning techniques can help choose among competing instruments. First, we randomly assign respondents to one of four survey instruments to predict a behavior defined by our validation strategy. Next, we assess the optimal instrument in two stages. A machine learning model first predicts the behavior using individual covariates and survey responses. Then, using doubly robust welfare maximization and prediction error from the first stage, we learn the optimal survey method and examine how it varies across education levels. |
| Date: | 2023–04–04 |
| URL: | http://d.repec.org/n?u=RePEc:tse:wpaper:128022&r=big |
| By: | Hong, Sung Wook (Korea Institute for Industrial Economics and Trade); Min, Seong-hwan (Korea Institute for Industrial Economics and Trade) |
| Abstract: | This paper examines the use and viability of unstructured data in forecasting the real economy in order to quickly understand the current situation and trends in the real economy in light of growing uncertainty both domestically and externally. To extract an index utilizing news data, the methodology of Thorsrud (2016) was used to develop two approaches. The first approach involves conducting topic analysis to extract topics and then employing sentiment analysis for each topic and calculating a simple index. The second approach involves creating a comprehensive score by combining the sentiment score and topic score to create a sort of composite index. To prove the utility of news data-based indices, the total population was set to match the number of topics that had the maximum number of word groups of all of the news data from the six segments of the manufacturing industry for the given period. That number was then reduced based on topics that met a certain standard, and the correlation between the extracted indices and real industry indices (such as an increase in the segment-based manufacturing index) was examined. In conclusion, the numbers calculated by analyzing the news for each segment of the manufacturing industry showed a similar trend to that of the real economy, showing that data extracted in a quantitative manner from unstructured data such as news articles prove to be significantly useful in understanding trends in the real economy. |
| Keywords: | real economy; forecasting; economic forecasting; news data; topic analysis; sentiment analysis; manufacturing; unstructured data; Korea |
| JEL: | C50 C53 C59 C82 C83 C89 E37 |
| Date: | 2021–05–21 |
| URL: | http://d.repec.org/n?u=RePEc:ris:kietrp:2021_008&r=big |
| By: | Laura Felber; Simon Beyeler |
| Abstract: | In this paper, we assess the value of high-frequency payments data for nowcasting economic activity. Focusing on Switzerland, we predict real GDP based on an unprecedented 'complete' set of transaction payments data: a combination of real-time gross settlement payment system data as well as debit and credit card data. Following a strongly data-driven machine learning approach, we find payments data to bear an accurate and timely signal about economic activity. When we assess the performance of the models by the initially published GDP numbers (pseudo real-time evaluation), we find a state-dependent value of the data: the payment models slightly outperform the benchmark models in times of crisis but are clearly inferior in 'normal' times. However, when we assess the performance of the models by revised and more final GDP numbers, we find payments data to be unconditionally valuable: the payment models outperform the benchmark models by up to 11% in times of crisis and by up to 12% in 'normal' times. We thus conclude that models based on payments data should become an integral part of policymakers' decision-making. |
| Keywords: | Nowcasting, GDP, machine learning, payments data, COVID-19 |
| JEL: | C52 C53 C55 E37 |
| Date: | 2023 |
| URL: | http://d.repec.org/n?u=RePEc:snb:snbwpa:2023-01&r=big |
| By: | Charlotte Cavaillé (University of Michigan [Ann Arbor] - University of Michigan System); Karine van Der Straeten (TSE-R - Toulouse School of Economics - UT Capitole - Université Toulouse Capitole - UT - Université de Toulouse - EHESS - École des hautes études en sciences sociales - CNRS - Centre National de la Recherche Scientifique - INRAE - Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement); Daniel L. Chen (TSE-R - Toulouse School of Economics - UT Capitole - Université Toulouse Capitole - UT - Université de Toulouse - EHESS - École des hautes études en sciences sociales - CNRS - Centre National de la Recherche Scientifique - INRAE - Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement) |
| Abstract: | Survey design often approximates a prediction problem: the goal is to select instruments that best predict the value of an unobserved construct or a future outcome. We demonstrate how advances in machine learning techniques can help choose among competing instruments. First, we randomly assign respondents to one of four survey instruments to predict a behavior defined by our validation strategy. Next, we assess the optimal instrument in two stages. A machine learning model first predicts the behavior using individual covariates and survey responses. Then, using doubly robust welfare maximization and prediction error from the first stage, we learn the optimal survey method and examine how it varies across education levels. |
| Date: | 2023–04–07 |
| URL: | http://d.repec.org/n?u=RePEc:hal:wpaper:hal-04062637&r=big |
| By: | El Amine Cherrat; Snehal Raj; Iordanis Kerenidis; Abhishek Shekhar; Ben Wood; Jon Dee; Shouvanik Chakrabarti; Richard Chen; Dylan Herman; Shaohan Hu; Pierre Minssen; Ruslan Shaydulin; Yue Sun; Romina Yalovetzky; Marco Pistoia |
| Abstract: | Quantum machine learning has the potential for a transformative impact across industry sectors and in particular in finance. In our work we look at the problem of hedging where deep reinforcement learning offers a powerful framework for real markets. We develop quantum reinforcement learning methods based on policy-search and distributional actor-critic algorithms that use quantum neural network architectures with orthogonal and compound layers for the policy and value functions. We prove that the quantum neural networks we use are trainable, and we perform extensive simulations that show that quantum models can reduce the number of trainable parameters while achieving comparable performance and that the distributional approach obtains better performance than other standard approaches, both classical and quantum. We successfully implement the proposed models on a trapped-ion quantum processor, utilizing circuits with up to $16$ qubits, and observe performance that agrees well with noiseless simulation. Our quantum techniques are general and can be applied to other reinforcement learning problems beyond hedging. |
| Date: | 2023–03 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2303.16585&r=big |
| By: | Jianfei Zhang; Mathieu Rosenbaum |
| Abstract: | News can convey bearish or bullish views on financial assets. Institutional investors need to evaluate automatically the implied news sentiment based on textual data. Given the huge amount of news articles published each day, most of which are neutral, we present a systematic news screening method to identify the ``true'' impactful ones, aiming for more effective development of news sentiment learning methods. Based on several liquidity-driven variables, including volatility, turnover, bid-ask spread, and book size, we associate each 5-min time bin to one of two specific liquidity modes. One represents the ``calm'' state at which the market stays for most of the time and the other, featured with relatively higher levels of volatility and trading volume, describes the regime driven by some exogenous events. Then we focus on the moments where the liquidity mode switches from the former to the latter and consider the news articles published nearby impactful. We apply naive Bayes on these filtered samples for news sentiment classification as an illustrative example. We show that the screened dataset leads to more effective feature capturing and thus superior performance on short-term asset return prediction compared to the original dataset. |
| Date: | 2023–04 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2304.05115&r=big |
| By: | Zhizhong Tan; Min Hu; Yixuan Wang; Lu Wei; Bin Liu |
| Abstract: | It is a challenging problem to predict trends of futures prices with traditional econometric models as one needs to consider not only futures' historical data but also correlations among different futures. Spatial-temporal graph neural networks (STGNNs) have great advantages in dealing with such kind of spatial-temporal data. However, we cannot directly apply STGNNs to high-frequency future data because future investors have to consider both the long-term and short-term characteristics when doing decision-making. To capture both the long-term and short-term features, we exploit more label information by designing four heterogeneous tasks: price regression, price moving average regression, price gap regression (within a short interval), and change-point detection, which involve both long-term and short-term scenes. To make full use of these labels, we train our model in a continual manner. Traditional continual GNNs define the gradient of prices as the parameter important to overcome catastrophic forgetting (CF). Unfortunately, the losses of the four heterogeneous tasks lie in different spaces. Hence it is improper to calculate the parameter importance with their losses. We propose to calculate parameter importance with mutual information between original observations and the extracted features. The empirical results based on 49 commodity futures demonstrate that our model has higher prediction performance on capturing long-term or short-term dynamic change. |
| Date: | 2023–03 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2303.16532&r=big |
| By: | Mohamed Hamdouche (LPSM); Pierre Henry-Labordere (LPSM); Huy\^en Pham (LPSM) |
| Abstract: | We propose a novel generative model for time series based on Schr{\"o}dinger bridge (SB) approach. This consists in the entropic interpolation via optimal transport between a reference probability measure on path space and a target measure consistent with the joint data distribution of the time series. The solution is characterized by a stochastic differential equation on finite horizon with a path-dependent drift function, hence respecting the temporal dynamics of the time series distribution. We can estimate the drift function from data samples either by kernel regression methods or with LSTM neural networks, and the simulation of the SB diffusion yields new synthetic data samples of the time series. The performance of our generative model is evaluated through a series of numerical experiments. First, we test with a toy autoregressive model, a GARCH Model, and the example of fractional Brownian motion, and measure the accuracy of our algorithm with marginal and temporal dependencies metrics. Next, we use our SB generated synthetic samples for the application to deep hedging on real-data sets. Finally, we illustrate the SB approach for generating sequence of images. |
| Date: | 2023–04 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2304.05093&r=big |
| By: | Acosta-Smith, Jonathan (Bank of England); Guin, Benjamin (Bank of England); Salgado-Moreno, Mauricio (Bank of England); Vo, Quynh-Anh (Bank of England) |
| Abstract: | Climate-related disclosures reduce information asymmetries between firms and investors and help transition to a net zero economy. However, disclosure practices might differ across firms. We explore the determinants of firm disclosures by creating a unique, firm-level panel data set on climate-related disclosures of UK financial institutions. To that end, we apply Natural Language Processing techniques with Machine Learning classifiers on unique textual data which we hand-collected from their published reports. We document differences in disclosure levels across financial institutions with different sizes and over time. We show that climate‑related policy communications in the form of regulatory guidance on future mandatory disclosures is associated with a catch-up by firms previously disclosing less. |
| Keywords: | Climate-related disclosures; market discipline; Task Force on Climate-Related Financial Disclosures (TCFD) and Natural Language Processing (NLP). |
| JEL: | C40 C80 G20 |
| Date: | 2023–03–10 |
| URL: | http://d.repec.org/n?u=RePEc:boe:boeewp:1017&r=big |
| By: | Murray Z. Frank; Jing Gao; Keer Yang |
| Abstract: | There is considerable evidence that machine learning algorithms have better predictive abilities than humans in various financial settings. But, the literature has not tested whether these algorithmic predictions are more rational than human predictions. We study the predictions of corporate earnings from several algorithms, notably linear regressions and a popular algorithm called Gradient Boosted Regression Trees (GBRT). On average, GBRT outperformed both linear regressions and human stock analysts, but it still overreacted to news and did not satisfy rational expectation as normally defined. By reducing the learning rate, the magnitude of overreaction can be minimized, but it comes with the cost of poorer out-of-sample prediction accuracy. Human stock analysts who have been trained in machine learning methods overreact less than traditionally trained analysts. Additionally, stock analyst predictions reflect information not otherwise available to machine algorithms. |
| Date: | 2023–03 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2303.16158&r=big |
| By: | Marie Obidzinski (CRED - Centre de Recherche en Economie et Droit - Université Paris-Panthéon-Assas); Yves Oytana (CRESE - Centre de REcherches sur les Stratégies Economiques (UR 3190) - UFC - Université de Franche-Comté - UBFC - Université Bourgogne Franche-Comté [COMUE]) |
| Abstract: | We study the design of optimal liability rules when the use of a prediction by a human operator (she) may generate an external harm. This setting is common when using artificial intelligence (AI) to make a decision. An AI manufacturer (he) chooses the level of quality with which the algorithm is developed and the price at which it is distributed. The AI makes a prediction about the state of the world to the human operator who buys it, who can then decide to exert a judgment effort to learn the payoffs in each possible state of the world. We show that when the human operator overestimates the algorithm's accuracy (overestimation bias), imposing a strict liability rule on her is not optimal, because the AI manufacturer will exploit the bias by under-investing in the quality of the algorithm. Conversely, imposing a strict liability rule on the AI manufacturer may not be optimal either, since it has the adverse effect of preventing the human operator from exercising her judgment effort. We characterize the liability sharing rule that achieves the highest possible quality level of the algorithm, while ensuring that the human operator exercises a judgment effort. We then show that, when it can be used, a negligence rule generally achieves the first best optimum. To conclude, we discuss the pros and cons of each type of liability rule. |
| Keywords: | Liability rules, Decision-making, Artificial intelligence, Cognitive bias, Judgment, Prediction, Algorithm |
| Date: | 2022–11–22 |
| URL: | http://d.repec.org/n?u=RePEc:hal:wpaper:hal-04034871&r=big |
| By: | Rasoul Amirzadeh; Asef Nazari; Dhananjay Thiruvady; Mong Shan Ee |
| Abstract: | The growth of market capitalisation and the number of altcoins (cryptocurrencies other than Bitcoin) provide investment opportunities and complicate the prediction of their price movements. A significant challenge in this volatile and relatively immature market is the problem of predicting cryptocurrency prices which needs to identify the factors influencing these prices. The focus of this study is to investigate the factors influencing altcoin prices, and these factors have been investigated from a causal analysis perspective using Bayesian networks. In particular, studying the nature of interactions between five leading altcoins, traditional financial assets including gold, oil, and S\&P 500, and social media is the research question. To provide an answer to the question, we create causal networks which are built from the historic price data of five traditional financial assets, social media data, and price data of altcoins. The ensuing networks are used for causal reasoning and diagnosis, and the results indicate that social media (in particular Twitter data in this study) is the most significant influencing factor of the prices of altcoins. Furthermore, it is not possible to generalise the coins' reactions against the changes in the factors. Consequently, the coins need to be studied separately for a particular price movement investigation. |
| Date: | 2023–03 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2303.16148&r=big |
| By: | Costantiello, Alberto; Leogrande, Angelo |
| Abstract: | In this article, we estimate the role of Political Stability and Absence of Violence and Terrorism-PS in the context of Environmental, Social and Governance-ESG data at world level. We analyse data from 193 countries in the period 2011-2020. We apply Panel Data with Fixed Effects, Panel Data with Random Effects and Pooled Ordinary Least Square-OLS. We found that PS is positively associated, among others, to Population Density and Government Effectiveness, and negatively associated, among others, to Research and Development Expenditure and Maximum 5-day Rainfall. Furthermore, we apply the k-Means algorithm optimized with the application of the Elbow Method and we find the presence of four clusters. Finally, we propose a confrontation among eight different machine-learning algorithms for the prediction of PS and we find that the Polynomial Regression shows the higher performance. The Polynomial Regression predicts an increase in the level of PS of 0.25% on average for the analysed countries. |
| Keywords: | Analysis of Collective Decision-Making; General; Political Processes: Rent-Seeking; Lobbying; Elections; Legislatures; and Voting Behaviour; Bureaucracy; Administrative Processes in Public Organizations; Corruption; Positive Analysis of Policy Formulation; Implementation |
| JEL: | D7 D70 D72 D73 D78 |
| Date: | 2023–04–02 |
| URL: | http://d.repec.org/n?u=RePEc:pra:mprapa:116887&r=big |
| By: | Davood Pirayesh Neghab; Mucahit Cevik; M. I. M. Wahab |
| Abstract: | The complexity and ambiguity of financial and economic systems, along with frequent changes in the economic environment, have made it difficult to make precise predictions that are supported by theory-consistent explanations. Interpreting the prediction models used for forecasting important macroeconomic indicators is highly valuable for understanding relations among different factors, increasing trust towards the prediction models, and making predictions more actionable. In this study, we develop a fundamental-based model for the Canadian-U.S. dollar exchange rate within an interpretative framework. We propose a comprehensive approach using machine learning to predict the exchange rate and employ interpretability methods to accurately analyze the relationships among macroeconomic variables. Moreover, we implement an ablation study based on the output of the interpretations to improve the predictive accuracy of the models. Our empirical results show that crude oil, as Canada's main commodity export, is the leading factor that determines the exchange rate dynamics with time-varying effects. The changes in the sign and magnitude of the contributions of crude oil to the exchange rate are consistent with significant events in the commodity and energy markets and the evolution of the crude oil trend in Canada. Gold and the TSX stock index are found to be the second and third most important variables that influence the exchange rate. Accordingly, this analysis provides trustworthy and practical insights for policymakers and economists and accurate knowledge about the predictive model's decisions, which are supported by theoretical considerations. |
| Date: | 2023–03 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2303.16149&r=big |
| By: | Lin Chen; Stephanie Houle |
| Abstract: | We generate high-frequency and up-to-date indicators to monitor news media coverage of supply (raw, intermediate and final goods) and labour shortages in Canada. We use natural language processing to construct two news-based indicators and time-varying topic narratives to track Canadian media coverage of these shortages from 2000 to 2022. This makes our indicators an insightful alternative monitoring tool for policy. Notably, our indicators track well with monthly price indexes and measures from the Bank of Canada’s Business Outlook Survey, and they are highly correlated with commonly tracked indicators of supply constraint. Moreover, the news-based indicators reflect the attention of the public on pressing issues. |
| Keywords: | Coronavirus disease (COVID-19); Econometric and statistical methods; Monetary policy and uncertainty; Recent economic and financial developments |
| JEL: | C55 C82 E37 |
| Date: | 2023–03 |
| URL: | http://d.repec.org/n?u=RePEc:bca:bocadp:23-8&r=big |
| By: | Chendi Ni; Yuying Li; Peter A. Forsyth |
| Abstract: | We study the optimal multi-period asset allocation problem with leverage constraints in a persistent, high-inflation environment. Based on filtered high-inflation regimes, we discover that a portfolio containing an equal-weighted stock index partially stochastically dominates a portfolio containing a capitalization-weighted stock index. Assuming the asset prices follow the jump diffusion model during high inflation periods, we establish a closed-form solution for the optimal strategy that outperforms a passive strategy under the cumulative quadratic tracking difference (CD) objective. The closed-form solution provides insights but requires unrealistic constraints. To obtain strategies under more practical considerations, we consider a constrained optimal control problem with bounded leverage. To solve this optimal control problem, we propose a novel leverage-feasible neural network (LFNN) model that approximates the optimal control directly. The LFNN model avoids high-dimensional evaluation of the conditional expectation (common in dynamic programming (DP) approaches). We establish mathematically that the LFNN approximation can yield a solution that is arbitrarily close to the solution of the original optimal control problem with bounded leverage. Numerical experiments show that the LFNN model achieves comparable performance to the closed-form solution on simulated data. We apply the LFNN approach to a four-asset investment scenario with bootstrap resampled asset returns. The LFNN strategy consistently outperforms the passive benchmark strategy by about 200 bps (median annualized return), with a greater than 90% probability of outperforming the benchmark at the terminal date. These results suggest that during persistent inflation regimes, investors should favor short-term bonds over long-term bonds, and the equal-weighted stock index over the cap-weighted stock index. |
| Date: | 2023–04 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2304.05297&r=big |
| By: | Qianqian Xie; Weiguang Han; Yanzhao Lai; Min Peng; Jimin Huang |
| Abstract: | Recently, large language models (LLMs) like ChatGPT have demonstrated remarkable performance across a variety of natural language processing tasks. However, their effectiveness in the financial domain, specifically in predicting stock market movements, remains to be explored. In this paper, we conduct an extensive zero-shot analysis of ChatGPT's capabilities in multimodal stock movement prediction, on three tweets and historical stock price datasets. Our findings indicate that ChatGPT is a "Wall Street Neophyte" with limited success in predicting stock movements, as it underperforms not only state-of-the-art methods but also traditional methods like linear regression using price features. Despite the potential of Chain-of-Thought prompting strategies and the inclusion of tweets, ChatGPT's performance remains subpar. Furthermore, we observe limitations in its explainability and stability, suggesting the need for more specialized training or fine-tuning. This research provides insights into ChatGPT's capabilities and serves as a foundation for future work aimed at improving financial market analysis and prediction by leveraging social media sentiment and historical stock data. |
| Date: | 2023–04 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2304.05351&r=big |
| By: | Shijie Wu; Ozan Irsoy; Steven Lu; Vadim Dabravolski; Mark Dredze; Sebastian Gehrmann; Prabhanjan Kambadur; David Rosenberg; Gideon Mann |
| Abstract: | The use of NLP in the realm of financial technology is broad and complex, with applications ranging from sentiment analysis and named entity recognition to question answering. Large Language Models (LLMs) have been shown to be effective on a variety of tasks; however, no LLM specialized for the financial domain has been reported in literature. In this work, we present BloombergGPT, a 50 billion parameter language model that is trained on a wide range of financial data. We construct a 363 billion token dataset based on Bloomberg's extensive data sources, perhaps the largest domain-specific dataset yet, augmented with 345 billion tokens from general purpose datasets. We validate BloombergGPT on standard LLM benchmarks, open financial benchmarks, and a suite of internal benchmarks that most accurately reflect our intended usage. Our mixed dataset training leads to a model that outperforms existing models on financial tasks by significant margins without sacrificing performance on general LLM benchmarks. Additionally, we explain our modeling choices, training process, and evaluation methodology. As a next step, we plan to release training logs (Chronicles) detailing our experience in training BloombergGPT. |
| Date: | 2023–03 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2303.17564&r=big |
| By: | Abhishek Arora; Xinmei Yang; Shao Yu Jheng; Melissa Dell |
| Abstract: | Many applications require grouping instances contained in diverse document datasets into classes. Most widely used methods do not employ deep learning and do not exploit the inherently multimodal nature of documents. Notably, record linkage is typically conceptualized as a string-matching problem. This study develops CLIPPINGS, (Contrastively Linking Pooled Pre-trained Embeddings), a multimodal framework for record linkage. CLIPPINGS employs end-to-end training of symmetric vision and language bi-encoders, aligned through contrastive language-image pre-training, to learn a metric space where the pooled image-text representation for a given instance is close to representations in the same class and distant from representations in different classes. At inference time, instances can be linked by retrieving their nearest neighbor from an offline exemplar embedding index or by clustering their representations. The study examines two challenging applications: constructing comprehensive supply chains for mid-20th century Japan through linking firm level financial records - with each firm name represented by its crop in the document image and the corresponding OCR - and detecting which image-caption pairs in a massive corpus of historical U.S. newspapers came from the same underlying photo wire source. CLIPPINGS outperforms widely used string matching methods by a wide margin and also outperforms unimodal methods. Moreover, a purely self-supervised model trained on only image-OCR pairs also outperforms popular string-matching methods without requiring any labels. |
| Date: | 2023–04 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2304.03464&r=big |
| By: | Maximilian Tschuchnig; Petra Tschuchnig; Cornelia Ferner; Michael Gadermayr |
| Abstract: | Inflation is a major determinant for allocation decisions and its forecast is a fundamental aim of governments and central banks. However, forecasting inflation is not a trivial task, as its prediction relies on low frequency, highly fluctuating data with unclear explanatory variables. While classical models show some possibility of predicting inflation, reliably beating the random walk benchmark remains difficult. Recently, (deep) neural networks have shown impressive results in a multitude of applications, increasingly setting the new state-of-the-art. This paper investigates the potential of the transformer deep neural network architecture to forecast different inflation rates. The results are compared to a study on classical time series and machine learning models. We show that our adapted transformer, on average, outperforms the baseline in 6 out of 16 experiments, showing best scores in two out of four investigated inflation rates. Our results demonstrate that a transformer based neural network can outperform classical regression and machine learning models in certain inflation rates and forecasting horizons. |
| Date: | 2023–03 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2303.15364&r=big |
| By: | Larosa, Francesca; Mysiak, Jaroslav; Molinari, Marco; Varelas, Panagiotis; Akay, Haluk; McDowall, Will; Spadaru, Catalina; Fuso-Nerini, Francesco; Vinuesa, Ricardo |
| Abstract: | Innovation is a key component to equip our society with tools to adapt to new climatic conditions. The development of research-action interfaces shifts useful ideas into operationalized knowledge allowing innovation to flourish. In this paper we quantify the existing gap between climate research and innovation action in Europe using a novel framework that combines artificial intelligence (AI) methods and network science. We compute the distance between key topics of research interest from peer review publications and core issues tackled by innovation projects funded by the most recent European framework programmes. Our findings reveal significant differences exist between and within the two layers. Economic incentives, agricultural and industrial processes are differently connected to adaptation and mitigation priorities. We also find a loose research-action connection in bioproducts, biotechnologies and risk assessment practices, where applications are still too few compared to the research insights. Our analysis supports policy-makers to measure and track how research funding result in innovation action, and to adjust decisions if stated priorities are not achieved. |
| Keywords: | climate innovation; natural language processing; knwoledge production |
| JEL: | H54 O32 O33 O38 |
| Date: | 2023 |
| URL: | http://d.repec.org/n?u=RePEc:pra:mprapa:116771&r=big |
| By: | Zhen Zeng; Rachneet Kaur; Suchetha Siddagangappa; Saba Rahimi; Tucker Balch; Manuela Veloso |
| Abstract: | Time series forecasting is important across various domains for decision-making. In particular, financial time series such as stock prices can be hard to predict as it is difficult to model short-term and long-term temporal dependencies between data points. Convolutional Neural Networks (CNN) are good at capturing local patterns for modeling short-term dependencies. However, CNNs cannot learn long-term dependencies due to the limited receptive field. Transformers on the other hand are capable of learning global context and long-term dependencies. In this paper, we propose to harness the power of CNNs and Transformers to model both short-term and long-term dependencies within a time series, and forecast if the price would go up, down or remain the same (flat) in the future. In our experiments, we demonstrated the success of the proposed method in comparison to commonly adopted statistical and deep learning methods on forecasting intraday stock price change of S&P 500 constituents. |
| Date: | 2023–04 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2304.04912&r=big |
| By: | Marlon Azinovic; Jan \v{Z}emli\v{c}ka |
| Abstract: | Contemporary deep learning based solution methods used to compute approximate equilibria of high-dimensional dynamic stochastic economic models are often faced with two pain points. The first problem is that the loss function typically encodes a diverse set of equilibrium conditions, such as market clearing and households' or firms' optimality conditions. Hence the training algorithm trades off errors between those -- potentially very different -- equilibrium conditions. This renders the interpretation of the remaining errors challenging. The second problem is that portfolio choice in models with multiple assets is only pinned down for low errors in the corresponding equilibrium conditions. In the beginning of training, this can lead to fluctuating policies for different assets, which hampers the training process. To alleviate these issues, we propose two complementary innovations. First, we introduce Market Clearing Layers, a neural network architecture that automatically enforces all the market clearing conditions and borrowing constraints in the economy. Encoding economic constraints into the neural network architecture reduces the number of terms in the loss function and enhances the interpretability of the remaining equilibrium errors. Furthermore, we present a homotopy algorithm for solving portfolio choice problems with multiple assets, which ameliorates numerical instabilities arising in the context of deep learning. To illustrate our method we solve an overlapping generations model with two permanent risk aversion types, three distinct assets, and aggregate shocks. |
| Date: | 2023–03 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2303.14802&r=big |