|
on Computational Economics |
|
Issue of 2020‒12‒14
nineteen papers chosen by |
| By: | Xiao-Yang Liu; Hongyang Yang; Qian Chen; Runjia Zhang; Liuqing Yang; Bowen Xiao; Christina Dan Wang |
| Abstract: | As deep reinforcement learning (DRL) has been recognized as an effective approach in quantitative finance, getting hands-on experiences is attractive to beginners. However, to train a practical DRL trading agent that decides where to trade, at what price, and what quantity involves error-prone and arduous development and debugging. In this paper, we introduce a DRL library FinRL that facilitates beginners to expose themselves to quantitative finance and to develop their own stock trading strategies. Along with easily-reproducible tutorials, FinRL library allows users to streamline their own developments and to compare with existing schemes easily. Within FinRL, virtual environments are configured with stock market datasets, trading agents are trained with neural networks, and extensive backtesting is analyzed via trading performance. Moreover, it incorporates important trading constraints such as transaction cost, market liquidity and the investor's degree of risk-aversion. FinRL is featured with completeness, hands-on tutorial and reproducibility that favors beginners: (i) at multiple levels of time granularity, FinRL simulates trading environments across various stock markets, including NASDAQ-100, DJIA, S&P 500, HSI, SSE 50, and CSI 300; (ii) organized in a layered architecture with modular structure, FinRL provides fine-tuned state-of-the-art DRL algorithms (DQN, DDPG, PPO, SAC, A2C, TD3, etc.), commonly-used reward functions and standard evaluation baselines to alleviate the debugging workloads and promote the reproducibility, and (iii) being highly extendable, FinRL reserves a complete set of user-import interfaces. Furthermore, we incorporated three application demonstrations, namely single stock trading, multiple stock trading, and portfolio allocation. The FinRL library will be available on Github at link https://github.com/AI4Finance-LLC/FinRL- Library. |
| Date: | 2020–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2011.09607&r=all |
| By: | Sandrine Jacob Leal (Groupe de Recherche en Droit, Economie et Gestion); Mauro Napoletano (Observatoire français des conjonctures économiques) |
| Abstract: | We investigate the effects of a set of regulatory policies directed towards high-frequency trading (HFT) through an agent-based model of a limit order book able to generate flash crashes as the result of the interactions between low- and high-frequency traders. In particular, we study the impact of the imposition of minimum resting times, of circuit breakers, of cancellation fees and of transaction taxes on asset price volatility and on the occurrence and the duration of flash crashes. Monte-Carlo simulations reveal that HFT-targeted policies imply a trade-off between market stability and resilience. Indeed, we find that policies able to tackle volatility and flash crashes also hinder the market from quickly recovering after a crash. This result is mainly due to the dual role of HFT, as both a cause of flash crashes and a key player in the post-crash recovery. |
| Keywords: | High-frequency trading; Flash crashes; Regulatory policies; Agent-based models; Limit order book; Market volatility |
| JEL: | G12 G1 C63 |
| Date: | 2019–01 |
| URL: | http://d.repec.org/n?u=RePEc:spo:wpmain:info:hdl:2441/6ummnc8nko827b2luohnctekk7&r=all |
| By: | Marc Sabate-Vidales; David \v{S}i\v{s}ka; Lukasz Szpruch |
| Abstract: | Using a combination of recurrent neural networks and signature methods from the rough paths theory we design efficient algorithms for solving parametric families of path dependent partial differential equations (PPDEs) that arise in pricing and hedging of path-dependent derivatives or from use of non-Markovian model, such as rough volatility models in Jacquier and Oumgari, 2019. The solutions of PPDEs are functions of time, a continuous path (the asset price history) and model parameters. As the domain of the solution is infinite dimensional many recently developed deep learning techniques for solving PDEs do not apply. Similarly as in Vidales et al. 2018, we identify the objective function used to learn the PPDE by using martingale representation theorem. As a result we can de-bias and provide confidence intervals for then neural network-based algorithm. We validate our algorithm using classical models for pricing lookback and auto-callable options and report errors for approximating both prices and hedging strategies. |
| Date: | 2020–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2011.10630&r=all |
| By: | Aurélien Nioche (Department of Communications and Networking [Aalto] - Aalto University); Nicolas P. Rougier (Mnemosyne - Mnemonic Synergy - LaBRI - Laboratoire Bordelais de Recherche en Informatique - CNRS - Centre National de la Recherche Scientifique - École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB) - Université Sciences et Technologies - Bordeaux 1 - Université Bordeaux Segalen - Bordeaux 2 - Inria Bordeaux - Sud-Ouest - Inria - Institut National de Recherche en Informatique et en Automatique - IMN - Institut des Maladies Neurodégénératives [Bordeaux] - UB - Université de Bordeaux - CNRS - Centre National de la Recherche Scientifique); Marc Deffains (IMN - Institut des Maladies Neurodégénératives [Bordeaux] - UB - Université de Bordeaux - CNRS - Centre National de la Recherche Scientifique); Sacha Bourgeois-Gironde (LEMMA - Laboratoire d'économie mathématique et de microéconomie appliquée - UP2 - Université Panthéon-Assas - Sorbonne Université, IJN - Institut Jean-Nicod - DEC - Département d'Etudes Cognitives - ENS Paris - ENS Paris - École normale supérieure - Paris - PSL - Université Paris sciences et lettres - EHESS - École des hautes études en sciences sociales - CNRS - Centre National de la Recherche Scientifique - Département de Philosophie - ENS Paris - ENS Paris - École normale supérieure - Paris - PSL - Université Paris sciences et lettres); Sébastien Ballesta (UNISTRA - Université de Strasbourg); Thomas Boraud (IMN - Institut des Maladies Neurodégénératives [Bordeaux] - UB - Université de Bordeaux - CNRS - Centre National de la Recherche Scientifique) |
| Abstract: | In humans, the attitude toward risk is not neutral and is dissimilar between bets involving gains and bets involving losses. The existence and prevalence of these decision features in non-human primates are unclear. In addition, only a few studies have tried to simulate the evolution of agents based on their attitude toward risk. Therefore, we still ignore to which extent Prospect theory's claims are evolutionary rooted. To shed light on this issue, we collected data in 9 macaques that performed bets involving gains or losses. We confirmed that their overall behaviour is coherent with Prospect theory's claims. In parallel, we used a genetic algorithm to simulate the evolution of a population of agents across several generations. We showed that the algorithm selects progressively agents that exhibit risk-seeking and an inverted S-shape distorted perception of probability. We compared these two results and found that monkeys' attitude toward risk when facing losses only is congruent with the simulation. This result is consistent with the idea that gambling in the loss domain is analogous to deciding in a context of life-threatening challenges where a certain level of risk-seeking behaviours and probability distortions may be adaptive. |
| Keywords: | Genetic algorithm,Cognitive biases,Monkey,Autonomous Cognitive Testing,Experimental economics |
| Date: | 2021–01–21 |
| URL: | http://d.repec.org/n?u=RePEc:hal:journl:hal-03005035&r=all |
| By: | Nhung Luu; Nicolas Woloszko; Orsetta Causa; Christine Arriola; Frank van Tongeren; Åsa Johansson |
| Abstract: | Whether gains from trade are equally distributed within countries is the subject of a lively debate. This paper presents a novel framework to analyse the distributional effects of trade policy by linking the OECD’s CGE trade model, METRO, with consumption expenditure data from household budget surveys. Specifically, this paper describes a methodology to produce a concordance and transition matrix linking GTAP sectors to household survey classifications based on the Classification of Individual Consumption According to Purpose (COICOP). A mapping methodology is an important pre-requisite for investigating research questions concerning the influence of household behaviour changes on trade, as well as trade developments and policy on household welfare. The paper provides an illustration of the mapping of trade policy induced price changes onto household expenditures by conducting stylized tariff simulations with METRO and translating those into household expenditures by income decile for selected EU countries. |
| Keywords: | Household expenditure microdata, Inequality, modelling, Trade policy |
| JEL: | C68 D12 E21 F13 F14 |
| Date: | 2020–12–04 |
| URL: | http://d.repec.org/n?u=RePEc:oec:traaab:244-en&r=all |
| By: | Kitova, Olga; Dyakonova, Ludmila; Savinova, Victoria |
| Abstract: | The article describes a system of hybrid models ‘SGM Horizon’ as intellectual forecasting information system. The system of forecasting models includes a set of regression models and an expandable set of intelligent models, including artificial neural networks, decision trees, etc. Regression models include systems of regression equations that describe the behavior of forecast indicators of the development of the Russian economy in the system of national accounts. The functioning of the system of equations is determined by scenario conditions set by expert. For those indicators whose forecasts do not meet the requirements of quality and accuracy, intelligent models based on machine learning are used. Using the ‘SHM Horizon’ tools, predictive calculations were performed for a system of 30 indicators of the social sphere of the City of Moscow using hybrid models, and for8 indicators a significant increase in the quality and accuracy of the forecast was achieved with artificial neural network models. The process of models building requires considerable time, in this regard, the authors see the further development of the system in the application of the multi-criteria ranking method. |
| Keywords: | Regional economics, Forecasting, Socio-economic indicators, Hybrid models, Machine learning, Neural networks, Decision trees |
| JEL: | C40 C45 |
| Date: | 2020–07–24 |
| URL: | http://d.repec.org/n?u=RePEc:pra:mprapa:104234&r=all |
| By: | Nicolas Woloszko |
| Abstract: | This paper introduces the OECD Weekly Tracker of economic activity for 46 OECD and G20 countries using Google Trends search data. The Tracker performs well in pseudo-real time simulations including around the COVID-19 crisis. The underlying model adds to the previous Google Trends literature in two respects: (1) the data are adjusted for common long-term bias and (2) the data include variables based on both Google Search categories and topics (the latter being a collection of related keywords), thus further exploiting the potential of Google Trends. The paper highlights the predictive power of specific topics, including "bankruptcies", "economic crisis", "investment", "luggage" and "mortgage". Calibration is performed using a neural network that captures non-linear patterns, which are shown to be consistent with economic intuition using machine learning interpretability tools ("Shapley values"). The tracker sheds light on the recent downturn and the dynamics of the rebound, and provides evidence about lasting shifts in consumption patterns. |
| Keywords: | COVID-19, Google Trends, high-frequency, interpretability, machine learning, nowcasting |
| JEL: | C45 C53 C55 E37 |
| Date: | 2020–12–01 |
| URL: | http://d.repec.org/n?u=RePEc:oec:ecoaaa:1634-en&r=all |
| By: | Sriubaite, I.; Harris, A.; Jones, A.M.; Gabbe, B. |
| Abstract: | We perform a prediction analysis using methods of supervised machine learning on a set of outcomes that measure economic consequences of road traffic injuries. We employ several parametric and non-parametric algorithms including regularised regressions, decision trees and random forests to model statistically challenging empirical distributions and identify the key risk groups. In addition to a traditional outcome of interest – health care costs – we predict net monetary benefits from treatment, and productivity losses measured by the probability to return to work after the injury. Using the predictions of each selected algorithm we construct an ensemble machine learning algorithm - the Super Learner algorithm. Our findings demonstrate that the Super Learner is effective and performs best in predicting all outcomes. Further analysis of predictions by different groups of patients play an important role in the understanding of key risk factors for higher costs and poorer outcomes and offers a deeper understanding of risk in the health care sector. |
| Keywords: | Prediction and classification; super learner; machine learning; healthcare costs; patient outcomes; road traffic injuries; |
| JEL: | I11 I19 C14 C38 C53 |
| Date: | 2020–11 |
| URL: | http://d.repec.org/n?u=RePEc:yor:hectdg:20/20&r=all |
| By: | Balaji, S. J.; Babu, Suresh Chandra; Pal, Suresh |
| Abstract: | Policy-making processes in developing countries often continue to operate devoid of evidence. In this study, we explore the research-policy linkages between the agroeconomic research system (AERS) and the agricultural policy system (APS) in India. Specifically, we examine questions directed to the Ministry of Agriculture and Farmers’ Welfare in the two houses of the national parliament—the House of the People (Lok Sabha) and the Council of States (Rajya Sabha)—and filter them for key issues that confront the APS. In addition, using the list of research articles published in two major national agricultural economics journals, we examine the alignment of the AERS toward addressing pressing policy issues. We use 6,465 questions raised by elected representatives in the parliamentary houses and 377 research articles, both during the period 2014–2018. We use machine learning techniques for information retrieval because the required information is hidden as non-numerical text. Using tag clouds (lists of words by frequency), we identify key divergences between the concerns of the APS and the research focus of the AERS, and explore their linkages. To broaden our understanding, we employ latent Dirichlet allocation, a natural language processing technique that identifies crucial issues and automates their classification under appropriate clusters, to examine synergies between the research and policy systems. Results show remarkable alignment between the AERS and the APS, invalidating the two-communities hypothesis. We identify persistent issues in the policy domain that require the support of the research system, as well as potential areas for research system realignment. |
| Keywords: | INDIA; SOUTH ASIA; ASIA; agricultural economics; machine learning; agricultural research; agricultural policies; policies; farmers; research-policy linkages; latent Dirichlet allocation; policy systems; agroeconomic research |
| Date: | 2020 |
| URL: | http://d.repec.org/n?u=RePEc:fpr:ifprid:1970&r=all |
| By: | Marcelo T. LaFleur |
| Abstract: | Between the many resolutions, speeches, reports and other documents that are produced each year, the United Nations is awash in text. It is an ongoing challenge to create a coherent and useful picture of this corpus. In particular, there is an interest in measuring how the work of the United Nations system aligns with the Sustainable Development Goals (SDGs). There is a need for a scalable, objective, and consistent way to measure how similar any given publication is to each of the 17 SDGs. This paper explains a proof-of-concept process for building such a system using machine learning algorithms. By creating a model of the 17 SDGs it is possible to measure how similar the contents of individual publications are to each of the goals — their SDG Score. This paper also shows how this system can be used in practice by computing the SDG Scores for a limited selection of DESA publications and providing some analytics. |
| Keywords: | SDG; publications; classification; topic models; machine learning, LDA |
| JEL: | O0 O20 C88 |
| Date: | 2019–05 |
| URL: | http://d.repec.org/n?u=RePEc:une:wpaper:159&r=all |
| By: | Naftali Cohen; Srijan Sood; Zhen Zeng; Tucker Balch; Manuela Veloso |
| Abstract: | Time series forecasting is essential for agents to make decisions in many domains. Existing models rely on classical statistical methods to predict future values based on previously observed numerical information. Yet, practitioners often rely on visualizations such as charts and plots to reason about their predictions. Inspired by the end-users, we re-imagine the topic by creating a framework to produce visual forecasts, similar to the way humans intuitively do. In this work, we take a novel approach by leveraging advances in deep learning to extend the field of time series forecasting to a visual setting. We do this by transforming the numerical analysis problem into the computer vision domain. Using visualizations of time series data as input, we train a convolutional autoencoder to produce corresponding visual forecasts. We examine various synthetic and real datasets with diverse degrees of complexity. Our experiments show that visual forecasting is effective for cyclic data but somewhat less for irregular data such as stock price. Importantly, we find the proposed visual forecasting method to outperform numerical baselines. We attribute the success of the visual forecasting approach to the fact that we convert the continuous numerical regression problem into a discrete domain with quantization of the continuous target signal into pixel space. |
| Date: | 2020–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2011.09052&r=all |
| By: | Hamza Saad |
| Abstract: | Traditional statistical and measurements are unable to solve all industrial data in the right way and appropriate time. Open markets mean the customers are increased, and production must increase to provide all customer requirements. Nowadays, large data generated daily from different production processes and traditional statistical or limited measurements are not enough to handle all daily data. Improve production and quality need to analyze data and extract the important information about the process how to improve. Data mining applied successfully in the industrial processes and some algorithms such as mining association rules, and decision tree recorded high professional results in different industrial and production fields. The study applied seven algorithms to analyze production data and extract the best result and algorithm in the industry field. KNN, Tree, SVM, Random Forests, ANN, Na\"ive Bayes, and AdaBoost applied to classify data based on three attributes without neglect any variables whether this variable is numerical or categorical. The best results of accuracy and area under the curve (ROC) obtained from Decision tree and its ensemble algorithms (Random Forest and AdaBoost). Thus, a decision tree is an appropriate algorithm to handle manufacturing and production data especially this algorithm can handle numerical and categorical data. |
| Date: | 2020–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2011.12348&r=all |
| By: | CERQUA, AUGUSTO; LETTA, MARCO |
| Abstract: | Impact evaluations of the microeconomic effects of the COVID-19 upheavals are essential but nonetheless highly challenging. Data scarcity and identification issues due to the ubiquitous nature of the exogenous shock account for the current dearth of counterfactual studies. To fill this gap, we combine up-to-date quarterly local labor markets (LLMs) data, collected from the Business Register kept by the Italian Chamber of Commerce, with the machine learning control method for counterfactual building. This allows us to shed light on the pandemic impact on the local economic dynamics of one of the hardest-hit countries, Italy. We document that the shock has already caused a moderate drop in employment and firm exit and an abrupt decrease in firm entry at the country level. More importantly, these effects have been dramatically uneven across the Italian territory and spatially uncorrelated with the epidemiological pattern of the first wave. We then use the estimated individual treatment effects to investigate the main predictors of such unbalanced patterns, finding that the heterogeneity of impacts is primarily associated with interactions among the exposure of economic activities to high social aggregation risks and pre-existing labor market fragilities. These results call for immediate place- and sector-based policy responses. |
| Keywords: | impact evaluation; counterfactual approach; machine learning; local labor markets; firms; COVID-19; Italy |
| JEL: | C53 D22 E24 R12 |
| Date: | 2020–11–26 |
| URL: | http://d.repec.org/n?u=RePEc:pra:mprapa:104404&r=all |
| By: | Mengjin Zhao; Guangyan Jia |
| Abstract: | Seeking robustness of risk among different assets, risk-budgeting portfolio selections have become popular in the last decade. Aiming at generalizing risk budgeting method from single-period case to the continuous-time, we characterize the risk contributions and marginal risk contributions on different assets as measurable processes, when terminal variance of wealth is recognized as the risk measure. Meanwhile this specified risk contribution has a aggregation property, namely that total risk can be represented as the aggregation of risk contributions among assets and $(t,\omega)$. Subsequently, risk budgeting problem that how to obtain the policy with given risk budget in continuous-time case, is also explored which actually is a stochastic convex optimization problem parametrized by given risk budget. Moreover single-period risk budgeting policy is related to the projected risk budget in continuous-time case. Based on neural networks, numerical methods are given in order to get the policy with a specified budget process. |
| Date: | 2020–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2011.10747&r=all |
| By: | Hamza Saad |
| Abstract: | Many workers at the production department of Libyan Textile Company work with different performances. Plan of company management is paying the money according to the specific performance and quality requirements for each worker. Thus, it is important to predict the accurate evaluation of workers to extract the knowledge for management, how much money it will pay as salary and incentive. For example, if the evaluation is average, then management of the company will pay part of the salary. If the evaluation is good, then it will pay full salary, moreover, if the evaluation is excellent, then it will pay salary plus incentive percentage. Twelve variables with 121 instances for each variable collected to predict the evaluation of the process for each worker. Before starting classification, feature selection used to predict the influential variables which impact the evaluation process. Then, four algorithms of decision trees used to predict the output and extract the influential relationship between inputs and output. To make sure get the highest accuracy, ensemble algorithm (Bagging) used to deploy four algorithms of decision trees and predict the highest prediction result 99.16%. Standard errors for four algorithms were very small; this means that there is a strong relationship between inputs (7 variables) and output (Evaluation). The curve of (Receiver operating characteristics) for algorithms gave a high-level specificity and sensitivity, and Gain charts were very close to together. According to the results, management of the company should take a logic decision about the evaluation of production process and extract the important variables that impact the evaluation. |
| Date: | 2020–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2011.12343&r=all |
| By: | Dario Sansone; Anna Zhu |
| Abstract: | Using novel nation-wide social security data combined with machine learning tools, we develop predictive models of income support receipt intensities for any payment enrolee in the Australian social security system between 2014 and 2018. We show that off-the-shelf machine learning algorithms can significantly improve predictive accuracy compared to simpler heuristic models or early warning systems currently in use. Specifically, the former predicts the proportion of time individuals will be on income support in the next four years with greater accuracy, by a magnitude of at least 22% (14 percentage points increase in the R2), compared to the latter. This gain can be achieved at little extra cost to practitioners since it uses data currently available to them. Consequently, our machine learning algorithms can improve the detection of long-term income support recipients accruing a welfare cost nearly AUD 1 billion higher than individuals identified in the current system. |
| Date: | 2020–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2011.12057&r=all |
| By: | Tamer Boyaci, (ESMT European School of Management and Technology); Caner Canyakmaz, (ESMT European School of Management and Technology); Francis de Véricourt, (ESMT European School of Management and Technology) |
| Abstract: | The rapid adoption of AI technologies by many organizations has recently raised concerns that AI may eventually replace humans in certain tasks. In fact, when used in collaboration, machines can significantly enhance the complementary strengths of humans. Indeed, because of their immense computing power, machines can perform specific tasks with incredible accuracy. In contrast, human decision-makers (DM) are flexible and adaptive but constrained by their limited cognitive capacity. This paper investigates how machine-based predictions may affect the decision process and outcomes of a human DM. We study the impact of these predictions on decision accuracy, the propensity and nature of decision errors as well as the DM's cognitive efforts. To account for both flexibility and limited cognitive capacity, we model the human decision-making process in a rational inattention framework. In this setup, the machine provides the DM with accurate but sometimes incomplete information at no cognitive cost. We fully characterize the impact of machine input on the human decision process in this framework. We show that machine input always improves the overall accuracy of human decisions, but may nonetheless increase the propensity of certain types of errors (such as false positives). The machine can also induce the human to exert more cognitive efforts, even though its input is highly accurate. Interestingly, this happens when the DM is most cognitively constrained, for instance, because of time pressure or multitasking. Synthesizing these results, we pinpoint the decision environments in which human-machine collaboration is likely to be most beneficial. |
| Keywords: | Machine-learning, rational inattention, human-machine collaboration, cognitive effort |
| Date: | 2020–11–30 |
| URL: | http://d.repec.org/n?u=RePEc:esm:wpaper:esmt-20-02&r=all |
| By: | Anna Scherbakova (National Research University Higher School of Economics) |
| Abstract: | The paper focuses on the task of clustering essays produced by ESL (English as a Second Language) learners. The data was taken from a learner corpus REALEC. The division of texts by certain characteristics can be useful to speed up the analysis of a single corpus or access to the necessary sections of a large number of documents. The study discusses not only some existing approaches to clustering text data, as well as the possibility of clustering texts produced by ESL learners, but also ways to extract keywords in order to determine the topic of the essays in each group. |
| Keywords: | learner corpus, text documents clustering, document embedding, keywords extraction, metadata enrichment. |
| JEL: | Z |
| Date: | 2020 |
| URL: | http://d.repec.org/n?u=RePEc:hig:wpaper:97/lng/2020&r=all |
| By: | Yifan Yu; Shan Huang; Yuchen Liu; Yong Tan |
| Abstract: | Social media-transmitted online information, particularly content that is emotionally charged, shapes our thoughts and actions. In this study, we incorporate social network theories and analyses to investigate how emotions shape online content diffusion, using a computational approach. We rigorously quantify and characterize the structural properties of diffusion cascades, in which more than six million unique individuals transmitted 387,486 articles in a massive-scale online social network, WeChat. We detected the degree of eight discrete emotions (i.e., surprise, joy, anticipation, love, anxiety, sadness, anger, and disgust) embedded in these articles, using a newly generated domain-specific and up-to-date emotion lexicon. We found that articles with a higher degree of anxiety and love reached a larger number of individuals and diffused more deeply, broadly, and virally, whereas sadness had the opposite effect. Age and network degree of the individuals who transmitted an article and, in particular, the social ties between senders and receivers, significantly mediated how emotions affect article diffusion. These findings offer valuable insight into how emotions facilitate or hinder information spread through social networks and how people receive and transmit online content that induces various emotions. |
| Date: | 2020–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2011.09003&r=all |