nep-cmp New Economics Papers
on Computational Economics
Issue of 2021‒05‒17
eleven papers chosen by
Stan Miles
Thompson Rivers University

  1. Applied Algorithmic Machine Learning for Intelligent Project Prediction: Towards an AI Framework of Project Success By Hsu, Ming-Wei; Dacre, Nicholas; Senyo, PK
  2. MCTG:Multi-frequency continuous-share trading algorithm with GARCH based on deep reinforcement learning By Zhishun Wang; Wei Lu; Kaixin Zhang; Tianhao Li; Zixi Zhao
  3. Bankruptcy Prediction Model Based on Business Risk Reports : Use of Natural Language Processing Techniques By Rasolomanana, Onjaniaina Mianin'Harizo
  4. Multilevel Monte Carlo simulation for VIX options in the rough Bergomi model By Florian Bourgey; Stefano De Marco
  5. Answering the Queen: Machine Learning and Financial Crises By FOULIARD, Jeremy; Howell, Michael J.; Rey, Hélène
  6. Predictor-corrector interior-point algorithm based on a new search direction working in a wide neighbourhood of the central path By Illés, Tibor; Rigó, Petra Renáta; Török, Roland
  7. Endogenous Prediction of Bankruptcy using a Support Vector Machine By Zazueta, Jorge; Heredia, Andrea Chavez; Zazueta-Hernández, Jorge
  8. Reinforcement Learning with Expert Trajectory For Quantitative Trading By Sihang Chen; Weiqi Luo; Chao Yu
  9. Algorithmic collusion with imperfect monitoring By Calvano, Emilio; Calzolari, Giacomo; Denicolò, Vincenzo; Pastorello, Sergio
  10. Preaching to Social Media: Turkey’s Friday Khutbas and Their Effects on Twitter By Aksoy, Ozan
  11. Unmasking Partisanship: Polarization Undermines Public Response to Collective Risk By Milosh, Maria; Painter, Marcus; Sonin, Konstantin; Van Dijcke, David; Wright, Austin L.

  1. By: Hsu, Ming-Wei; Dacre, Nicholas; Senyo, PK
    Abstract: A growing number of emerging studies have been undertaken to examine the mediating dynamics between intelligent agents, activities, and cost within allocated budgets, in order to predict the outcomes of complex projects in dint of their significant uncertain nature in achieving a successful outcome. For example, prior studies have used machine learning models to calculate and perform predictions. Artificial neural networks are the most frequently used machine learning model with support vector machine, and genetic algorithm and decision trees are sometimes used in several related studies. Furthermore, most machine learning algorithms used in prior studies generally assume that inputs and outputs are independent of each other, which suggests that a project's success is expected to be independent of other projects. As the datasets used to train in prior studies often contain projects from different clients across industries, this theoretical assumption remains tenable. However, in practice projects are often interrelated across several different dimensions, for example through distributed overlapping teams. An ongoing ethnographic study at a leading project management artificial intelligence consultancy, referred to in this research as Company Alpha, suggests that projects within the same portfolio frequently share overlapping characteristics. To capture the emergent inter-project relationships, this study aims to compare two specific types of artificial neural network prediction performances; (i) multilayer perceptron and; (ii) recurrent neural networks. The multilayer perceptron has been found to be one of the most widely used artificial neural networks in the project management literature, and recurrent networks are distinguished by the memory they take from prior inputs to influence input and output. Through this comparison, this research will examine whether recurrent neural networks can capture the potential inter-project relationship towards achieving improved performance in contrast to multilayer perceptron. Our empirical investigation using ethnographic practice-based exploration at Company Alpha will contribute to project management knowledge and support developing an intelligent project prediction AI framework with future applications for project practice.
    Date: 2021–03–15
  2. By: Zhishun Wang; Wei Lu; Kaixin Zhang; Tianhao Li; Zixi Zhao
    Abstract: Making profits in stock market is a challenging task for both professional institutional investors and individual traders. With the development combination of quantitative trading and reinforcement learning, more trading algorithms have achieved significant gains beyond the benchmark model Buy&Hold (B&H). There is a certain gap between these algorithms and the real trading decision making scenarios. On the one hand, they only consider trading signals while ignoring the number of transactions. On the other hand, the information level considered by these algorithms is not rich enough, which limits the performance of these algorithms. Thus, we propose an algorithm called the Multi-frequency Continuous-share Trading algorithm with GARCH (MCTG) to solve the problems above, which consists of parallel network layers and deep reinforcement learning. The former is composed of three parallel network layers, respectively dealing with different frequencies (five minute, one day, one week) data, and day level considers the volatilities of stocks. The latter with a continuous action space of the reinforcement learning algorithm is used to solve the problem of trading stock shares. Experiments in different industries of Chinese stock market show our method achieves more extra profit comparing with basic DRL methods and bench model.
    Date: 2021–05
  3. By: Rasolomanana, Onjaniaina Mianin'Harizo
    Abstract: The purpose of this study is to assess how useful risk information is in bankruptcy prediction, by performing a sentiment analysis of the texts. The proposed method involves the use of Natural Language Processing (NLP) and machine learning techniques. The results show that neural networks performed better than other classifiers, with a classification accuracy of 96.15% for this particular text classification problem. This work demonstrates that business risks reports carry information that helps predict the likelihood of bankruptcy.
    Keywords: Bankruptcy prediction, Business risk, Natural language processing, NLP, Sentiment analysis, Neural Networks,
    Date: 2021–04
  4. By: Florian Bourgey; Stefano De Marco
    Abstract: We consider the pricing of VIX options in the rough Bergomi model [Bayer, Friz, and Gatheral, Pricing under rough volatility, Quantitative Finance 16(6), 887-904, 2016]. In this setting, the VIX random variable is defined by the one-dimensional integral of the exponential of a Gaussian process with correlated increments, hence approximate samples of the VIX can be constructed via discretization of the integral and simulation of a correlated Gaussian vector. A Monte-Carlo estimator of VIX options based on a rectangle discretization scheme and exact Gaussian sampling via the Cholesky method has a computational complexity of order $\mathcal O(\varepsilon^{-4})$ when the mean-squared error is set to $\varepsilon^2$. We demonstrate that this cost can be reduced to $\mathcal O(\varepsilon^{-2} \log^2(\varepsilon))$ combining the scheme above with the multilevel method [Giles, Multilevel Monte Carlo path simulation, Oper. Res. 56(3), 607-617, 2008], and further reduced to the asymptotically optimal cost $\mathcal O(\varepsilon^{-2})$ when using a trapezoidal discretization. We provide numerical experiments highlighting the efficiency of the multilevel approach in the pricing of VIX options in such a rough forward variance setting.
    Date: 2021–05
  5. By: FOULIARD, Jeremy; Howell, Michael J.; Rey, Hélène
    Abstract: Financial crises cause economic, social and political havoc. Macroprudential policies are gaining traction but are still severely under-researched compared to monetary policy and fiscal policy. We use the general framework of sequential predictions also called online machine learning to forecast crises out-of-sample. Our methodology is based on model averaging and is meta-statistic since we can incorporate any predictive model of crises in our set of experts and test its ability to add information. We are able to predict systemic financial crises twelve quarters ahead out-of-sample with high signal-to-noise ratio in most cases. We analyse which experts provide the most information for our predictions at each point in time and for each country, allowing us to gain some insights into economic mechanisms underlying the building of risk in economies.
    Date: 2020–12
  6. By: Illés, Tibor; Rigó, Petra Renáta; Török, Roland
    Abstract: We introduce a new predictor-corrector interior-point algorithm for solving P_*(κ)-linear complementarity problems which works in a wide neighbourhood of the central path. We use the technique of algebraic equivalent transformation of the centering equations of the central path system. In this technique, we apply the function φ(t)=√t in order to obtain the new search directions. We define the new wide neighbourhood D_φ. In this way, we obtain the first interior-point algorithm, where not only the central path system is transformed, but the definition of the neighbourhood is also modified taking into consideration the algebraic equivalent transformation technique. This gives a new direction in the research of interior-point methods. We prove that the IPA has O((1+κ)n log⁡((〖〖(x〗^0)〗^T s^0)/ϵ) ) iteration complexity. Furtermore, we show the efficiency of the proposed predictor-corrector interior-point method by providing numerical results. Up to our best knowledge, this is the first predictor-corrector interior-point algorithm which works in the D_φ neighbourhood using φ(t)=√t.
    Keywords: predictor-corrector interior-point algorithm; P_*(κ)-linear complementarity problems; wide neighbourhood; algebraic equivalent transformation technique
    JEL: C61
    Date: 2021–05–02
  7. By: Zazueta, Jorge; Heredia, Andrea Chavez; Zazueta-Hernández, Jorge
    Abstract: We build a global bankruptcy prediction model using a support vector machine trained only on firms' endogenous information in the form of financial ratios. The model is tested not only on entirely random unseen data but on samples taken from specific global regions and industries to test for prediction bias, achieving satisfactory prediction performance in all cases. While support vector machines are not easily interpretable, we explore variable importance and find it consistent with economic intuition.
    Date: 2021–05–06
  8. By: Sihang Chen; Weiqi Luo; Chao Yu
    Abstract: In recent years, quantitative investment methods combined with artificial intelligence have attracted more and more attention from investors and researchers. Existing related methods based on the supervised learning are not very suitable for learning problems with long-term goals and delayed rewards in real futures trading. In this paper, therefore, we model the price prediction problem as a Markov decision process (MDP), and optimize it by reinforcement learning with expert trajectory. In the proposed method, we employ more than 100 short-term alpha factors instead of price, volume and several technical factors in used existing methods to describe the states of MDP. Furthermore, unlike DQN (deep Q-learning) and BC (behavior cloning) in related methods, we introduce expert experience in training stage, and consider both the expert-environment interaction and the agent-environment interaction to design the temporal difference error so that the agents are more adaptable for inevitable noise in financial data. Experimental results evaluated on share price index futures in China, including IF (CSI 300) and IC (CSI 500), show that the advantages of the proposed method compared with three typical technical analysis and two deep leaning based methods.
    Date: 2021–05
  9. By: Calvano, Emilio; Calzolari, Giacomo; Denicolò, Vincenzo; Pastorello, Sergio
    Abstract: We show that if they are allowed enough time to complete the learning, Q-learning algorithms can learn to collude in an environment with imperfect monitoring adapted from Green and Porter (1984), without having been instructed to do so, and without communicating with one another. Collusion is sustained by punishments that take the form of "price wars" triggered by the observation of low prices. The punishments have a finite duration, being harsher initially and then gradually fading away. Such punishments are triggered both by deviations and by adverse demand shocks.
    Keywords: artificial intelligence; Collusion; Imperfect Monitoring; Q-Learning
    JEL: D43 D83 L13 L41
    Date: 2021–01
  10. By: Aksoy, Ozan
    Abstract: In this study I analyse through machine learning the content of all Friday khutbas (sermons) read to millions of citizens in thousands of Mosques of Turkey since 2015. I focus on six non-religious and recurrent topics that feature in the sermons, namely business, family, nationalism, health, trust, and patience. I demonstrate that the content of the sermons respond strongly to events of national importance. I then link the Friday sermons with ~4.8 million tweets on these topics to study whether and how the content of sermons affects social media behaviour. I find generally large effects of the sermons on tweets, but there is also heterogeneity by topic. It is strongest for nationalism, patience, and health and weakest for business. Overall, these results show that religious institutions in Turkey are influential in shaping the public’s social media content and that this influence is mainly prevalent on salient issues. More generally, these results show that mass offline religious activity can have strong effects on social media behavior.
    Date: 2021–05–12
  11. By: Milosh, Maria; Painter, Marcus; Sonin, Konstantin; Van Dijcke, David; Wright, Austin L.
    Abstract: Political polarization may undermine public policy response to collective risk, especially in periods of crisis, when political actors have incentives to manipulate public perceptions. We study these dynamics in the United States, focusing on how partisanship has influenced the use of face masks to stem the spread of COVID-19. Using a wealth of micro-level data, machine learning approaches, and a novel quasi-experimental design, we establish the following: (1) mask use is robustly correlated with partisanship; (2) the impact of partisanship on mask use is not offset by local policy interventions; (3) partisanship is the single most important predictor of local mask use, not COVID-19 severity or local policies; (4) president Trump's unexpected mask use at Walter Reed on July 11, 2020 and endorsement of masks on July 20, 2020 significantly increased social media engagement with and positive sentiment towards mask-related topics. These results unmask how partisanship undermines effective public responses to collective risk and how messaging by political agents can increase public engagement with policy measures.
    Keywords: COVID-19; partisanship; Polarization
    JEL: H12 I18
    Date: 2020–11

This nep-cmp issue is ©2021 by Stan Miles. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.
General information on the NEP project can be found at For comments please write to the director of NEP, Marco Novarese at <>. Put “NEP” in the subject, otherwise your mail may be rejected.
NEP’s infrastructure is sponsored by the School of Economics and Finance of Massey University in New Zealand.