|
on Computational Economics |
|
Issue of 2021‒11‒29
fifteen papers chosen by |
| By: | Kandula, Shanthan; Krishnamoorthy, Srikumar; Roy, Debjit |
| Abstract: | E-commerce packages are notorious for their inefficient usage of space. More than one-quarter volume of a typical e-commerce package comprises air and filler material. The inefficient usage of space significantly reduces the transportation and distribution capacity increasing the operational costs. Therefore, designing an optimal set of packaging box sizes is imperative for improving efficiency. Though prior approaches for determining the optimal box sizes exist, they cannot be applied due to the wide range of SKUs hosted by the e-commerce warehouses. Besides, designing a few tens of boxes for covering hundreds of thousands of SKUs that span a wide range of sizes is impractical with the integer programming formulations used by the conventional approaches. This article proposes a scalable three-stage optimization framework that combines unsupervised learning, reinforcement learning, and tree search to design optimal box sizes. More specifically, the package optimization problem is formulated into a sequential decision-making task called the box-sizing game. A neural network agent is then designed to play the game and learn control policies to solve the problem. In addition, a tree-search operator is developed to improve the performance of the learned policies. The proposed framework is evaluated on real-world and synthetic datasets against standard metaheuristics and industry benchmarks. Results indicate the robustness and superiority of the approach in generating industry-strength solutions. Specifically, the packaging box assortments generated by the framework are 5% to 7.5% better than the industry baselines. |
| Date: | 2021–11–17 |
| URL: | http://d.repec.org/n?u=RePEc:iim:iimawp:14665&r= |
| By: | Ananda Chatterjee; Hrisav Bhowmick; Jaydip Sen |
| Abstract: | For a long-time, researchers have been developing a reliable and accurate predictive model for stock price prediction. According to the literature, if predictive models are correctly designed and refined, they can painstakingly and faithfully estimate future stock values. This paper demonstrates a set of time series, econometric, and various learning-based models for stock price prediction. The data of Infosys, ICICI, and SUN PHARMA from the period of January 2004 to December 2019 was used here for training and testing the models to know which model performs best in which sector. One time series model (Holt-Winters Exponential Smoothing), one econometric model (ARIMA), two machine Learning models (Random Forest and MARS), and two deep learning-based models (simple RNN and LSTM) have been included in this paper. MARS has been proved to be the best performing machine learning model, while LSTM has proved to be the best performing deep learning model. But overall, for all three sectors - IT (on Infosys data), Banking (on ICICI data), and Health (on SUN PHARMA data), MARS has proved to be the best performing model in sales forecasting. |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2111.01137&r= |
| By: | Narayanan, Sridhar (Stanford U); Kalyanam, Kirthi (Santa Clara U) |
| Abstract: | The availability of behavioral and other data on customers and advances in machine learning methods have enabled targeting of customers in a variety of domains, including pricing, advertising, recommendation systems and personal selling contexts. Typically, such targeting involves first training a machine learning algorithm on a training dataset, and then using that algorithm to score current or potential customers. When the score crosses a threshold, a treatment (such as an offer, an advertisement or a recommendation) is assigned. In this paper, we demonstrate that this has given rise to opportunities for causal measurement of the effects of such targeted treatments using regression discontinuity designs (RDD). Investigating machine learning in a regression discontinuity framework leads to several insights. First, we characterize conditions under which regression discontinuity designs can be used to measure not just local average treatment effects (LATE), but also average treatment effects (ATE). In some situations, we show that RD can be used to find bounds on the ATE even if we are unable to find point estimates. We then apply this to the machine learning based targeting contexts by studying two different ways in which the score required for targeting is generated, and explore the utility of RDD to these contexts. Finally, we apply our approach in the empirical context of the targeting of retargeted display advertising. Using a dataset from a context where a machine learning based targeting policy was employed in parallel with a randomized controlled trial, we examine the performance of the RDD estimate in estimating the treatment effect, validate it using a placebo test and demonstrate its practical utility. |
| Date: | 2020–12 |
| URL: | http://d.repec.org/n?u=RePEc:ecl:stabus:3925&r= |
| By: | Hu, Junjie; López Cabrera, Brenda; Melzer, Awdesch |
| Abstract: | Short Term Load Forecast (STLF) is necessary for effective scheduling, operation optimization trading, and decision-making for electricity consumers. Modern and efficient machine learning methods are recalled nowadays to manage complicated structural big datasets, which are characterized by having a nonlinear temporal dependence structure. We propose different statistical nonlinear models to manage these challenges of hard type datasets and forecast 15-min frequency electricity load up to 2-days ahead. We show that the Long-short Term Memory (LSTM) and the Gated Recurrent Unit (GRU) models applied to the production line of a chemical production facility outperform several other predictive models in terms of out-of-sample forecasting accuracy by the Diebold-Mariano (DM) test with several metrics. The predictive information is fundamental for the risk and production management of electricity consumers. |
| Keywords: | Short Term Load Forecast,Deep Neural Network,Hard Structure Load Process |
| JEL: | C51 C52 C53 Q31 Q41 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:zbw:irtgdp:2021020&r= |
| By: | Vito Polito; Yunyi Zhang |
| Abstract: | We develop a regime switching vector autoregression where artificial neural networks drive time variation in the coefficients of the conditional mean of the endogenous variables and the variance covariance matrix of the disturbances. The model is equipped with a stability constraint to ensure non-explosive dynamics. As such, it is employable to account for nonlinearity in macroeconomic dynamics not only during typical business cycles but also in a wide range of extreme events, like deep recessions and strong expansions. The methodology is put to the test using aggregate data for the United States that include the abnormal realizations during the recent Covid-19 pandemic. The model delivers plausible and stable structural inference, and accurate out-of-sample forecasts. This performance compares favourably against a number of alternative methodologies recently proposed to deal with large outliers in macroeconomic data caused by the pandemic. |
| Keywords: | nonlinear time series, regime switching models, extreme events, Covid-19, macroeconomic forecasting |
| JEL: | C45 C50 E37 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:ces:ceswps:_9395&r= |
| By: | Isaac K. Ofori (University of Insubria, Varese, Italy); Christopher Quaidoo (University of Insubria, Varese, Italy); Pamela E. Ofori (University of Insubria, Varese, Italy) |
| Abstract: | This study uses machine learning techniques to identify the key drivers of financial development in Africa. To this end, four regularization techniques— the Standard lasso, Adaptive lasso, the minimum Schwarz Bayesian information criterion lasso, and the Elasticnet are trained based on a dataset containing 86 covariates of financial development for the period 1990 – 2019. The results show that variables such as cell phones, economic globalisation, institutional effectiveness, and literacy are crucial for financial sector development in Africa. Evidence from the Partialing-out lasso instrumental variable regression reveals that while inflation and agricultural sector employment suppress financial sector development, cell phones and institutional effectiveness are remarkable in spurring financial sector development in Africa. Policy recommendations are provided in line with the rise in globalisation, and technological progress in Africa. |
| Keywords: | Africa, Elasticnet, Financial Development, Financial Inclusion, Lasso, Regularization, Variable Selection |
| JEL: | C01 C14 C52 C53 C55 E5 O55 |
| Date: | 2021–01 |
| URL: | http://d.repec.org/n?u=RePEc:exs:wpaper:21/074&r= |
| By: | Jean-Charles Bricongne; Baptiste Meunier; Thomas Pical |
| Abstract: | The Covid-19 crisis has highlighted innovative high-frequency dataset allowing to measure in real-time the economic impact. In this vein, we explore how satellite data measuring the concentration of nitrogen dioxide (NO2, a pollutant emitted mainly by industrial activity) in the troposphere can help predict industrial production. We first show how such data must be adjusted for meteorological patterns which can alter data quality and pollutant emissions. We use machine learning techniques to better account for non-linearities and interactions between variables. We then find evidence that nowcasting performances for monthly industrial production are significantly improved when relying on daily NO2 data compared to benchmark models based on PMIs and auto-regressive (AR) terms. We also find evidence of heterogeneities suggesting that the contribution of daily pollution data is particularly important during “crisis” episodes and that the elasticity of NO2 pollution to industrial production for a country depends on the share of manufacturing in the value added. Available daily, free-to-use, granular and covering all countries including those with limited statistics, this paper illustrates the potential of satellite-based data for air pollution in enhancing the real-time monitoring of economic activity. |
| Keywords: | Data Science, Big Data, Satellite Data, Nowcasting, Machine Learning, Industrial Production |
| JEL: | C51 C81 E23 E37 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:bfr:banfra:847&r= |
| By: | Tobias Schultheiss; Uschi Backes-Gellner |
| Abstract: | This paper examines the role of lifelong learning in counteracting skill depreciation and obsolescence. We build on findings showing that different skill types have structurally different depreciation rates. We differentiate between hard and soft skills and measure the relative importance of these two skill types at the occupational level. As data source we draw on a large sample of job advertisements and a categorization of their skill requirements through a machine-learning algorithm. We analyze lifelong learning effects for "harder" occupations (with relatively more hard than soft skills) versus "softer" occupations. Our results reveal important patterns of skill depreciation and counteracting lifelong learning effects: In harder occupations, the role of lifelong learning is primarily as a hedge against unemployment risks caused by fast-depreciating hard skills; in softer occupations, this role instead lies mostly in acting as a boost to wage gains and upward career mobility as workers build on a value-stable skill foundation. |
| JEL: | I26 J24 |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:iso:educat:0188&r= |
| By: | David Karpa; Torben Klarl; Michael Rochlitz |
| Abstract: | The most important resource to improve technologies in the field of artificial intelligence is data. Two types of policies are crucial in this respect: privacy and data-sharing regulations, and the use of surveillance technologies for policing. Both types of policies vary substantially across countries and political regimes. In this chapter, we examine how authoritarian and democratic political institutions can influence the quality of research in artificial intelligence, and the availability of large-scale datasets to improve and train deep learning algorithms. We focus mainly on the Chinese case, and find that -- ceteris paribus -- authoritarian political institutions continue to have a negative effect on innovation. They can, however, have a positive effect on research in deep learning, via the availability of large-scale datasets that have been obtained through government surveillance. We propose a research agenda to study which of the two effects might dominate in a race for leadership in artificial intelligence between countries with different political institutions, such as the United States and China. |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2111.00992&r= |
| By: | Nhan Huynh; Mike Ludkovski |
| Abstract: | We investigate jointly modeling Age-specific rates of various causes of death in a multinational setting. We apply Multi-Output Gaussian Processes (MOGP), a spatial machine learning method, to smooth and extrapolate multiple cause-of-death mortality rates across several countries and both genders. To maintain flexibility and scalability, we investigate MOGPs with Kronecker-structured kernels and latent factors. In particular, we develop a custom multi-level MOGP that leverages the gridded structure of mortality tables to efficiently capture heterogeneity and dependence across different factor inputs. Results are illustrated with datasets from the Human Cause-of-Death Database (HCD). We discuss a case study involving cancer variations in three European nations, and a US-based study that considers eight top-level causes and includes comparison to all-cause analysis. Our models provide insights into the commonality of cause-specific mortality trends and demonstrate the opportunities for respective data fusion. |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2111.06631&r= |
| By: | Asier Guti\'errez-Fandi\~no; Miquel Noguer i Alonso; Petter Kolm; Jordi Armengol-Estap\'e |
| Abstract: | We introduce a new language representation model in finance called Financial Embedding Analysis of Sentiment (FinEAS). In financial markets, news and investor sentiment are significant drivers of security prices. Thus, leveraging the capabilities of modern NLP approaches for financial sentiment analysis is a crucial component in identifying patterns and trends that are useful for market participants and regulators. In recent years, methods that use transfer learning from large Transformer-based language models like BERT, have achieved state-of-the-art results in text classification tasks, including sentiment analysis using labelled datasets. Researchers have quickly adopted these approaches to financial texts, but best practices in this domain are not well-established. In this work, we propose a new model for financial sentiment analysis based on supervised fine-tuned sentence embeddings from a standard BERT model. We demonstrate our approach achieves significant improvements in comparison to vanilla BERT, LSTM, and FinBERT, a financial domain specific BERT. |
| Date: | 2021–10 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2111.00526&r= |
| By: | Pallavi Basu; Luella Fu; Alessio Saretto; Wenguang Sun |
| Abstract: | In sparse large-scale testing problems where the false discovery proportion (FDP) is highly variable, the false discovery exceedance (FDX) provides a valuable alternative to the widely used false discovery rate (FDR). We develop an empirical Bayes approach to controlling the FDX. We show that for independent hypotheses from a two-group model and dependent hypotheses from a Gaussian model fulfilling the exchangeability condition, an oracle decision rule based on ranking and thresholding the local false discovery rate (lfdr) is optimal in the sense that the power is maximized subject to FDX constraint. We propose a data-driven FDX procedure that emulates the oracle via carefully designed computational shortcuts. We investigate the empirical performance of the proposed method using simulations and illustrate the merits of FDX control through an application for identifying abnormal stock trading strategies. |
| Keywords: | Cautious Data Mining; False Discovery Exceedance Control; Local False Discovery Rates; Multiple Hypotheses Testing; Poisson Binomial Distribution; Trading Strategies |
| JEL: | C11 C12 C15 |
| Date: | 2021–11–18 |
| URL: | http://d.repec.org/n?u=RePEc:fip:feddwp:93384&r= |
| By: | Francisco Estrada; Oscar Calder\'on-Bustamante; Wouter Botzen; Juli\'an A. Velasco; Richard S. J. Tol |
| Abstract: | Complex physical models are the most advanced tools available for producing realistic simulations of the climate system. However, such levels of realism imply high computational cost and restrictions on their use for policymaking and risk assessment. Two central characteristics of climate change are uncertainty and that it is a dynamic problem in which international actions can significantly alter climate projections and information needs, including partial and full compliance of global climate goals. Here we present AIRCC-Clim, a simple climate model emulator that produces regional probabilistic climate change projections of monthly and annual temperature and precipitation, as well as risk measures, based both on standard and user-defined emissions scenarios for six greenhouse gases. AIRCC-Clim emulates 37 atmosphere-ocean coupled general circulation models with low computational and technical requirements for the user. This standalone, user-friendly software is designed for a variety of applications including impact assessments, climate policy evaluation and integrated assessment modelling. |
| Date: | 2021–10 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2111.01762&r= |
| By: | Angela E. Kilby; Charlie Denhart |
| Abstract: | The Covid-19 pandemic has intersected with the opioid epidemic to create a unique public health crisis, with the health and economic consequences of the virus and associated lockdowns compounding pre-existing social and economic stressors associated with rising opioid and heroin use and abuse. In order to better understand these interlocking crises, we use social media data to extract qualitative and quantitative insights on the experiences of opioid users during the Covid-19 pandemic. In particular, we use an unsupervised learning approach to create a rich geolocated data source for public health surveillance and analysis. To do this we first infer the location of 26,000 Reddit users that participate in opiate-related sub-communities (subreddits) by combining named entity recognition, geocoding, density-based clustering, and heuristic methods. Our strategy achieves 63 percent accuracy at state-level location inference on a manually-annotated reference dataset. We then leverage the geospatial nature of our user cohort to answer policy-relevant questions about the impact of varying state-level policy approaches that balance economic versus health concerns during Covid-19. We find that state government strategies that prioritized economic reopening over curtailing the spread of the virus created a markedly different environment and outcomes for opioid users. Our results demonstrate that geospatial social media data can be used for agile monitoring of complex public health crises. |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2111.01778&r= |
| By: | Luo, Danqi (Stanford U); Bayati, Mohsen (Stanford U); Plambeck, Erica L. (Stanford U); Aratow, Michael (San Mateo Medical Center) |
| Abstract: | This paper provides evidence that the arrival of an additional low-acuity patient substantially increases the wait time to start of treatment for high-acuity patients, contradicting the long-standing prior conclusion in the medical literature that the effect is "negligible." Whereas the medical literature underestimates the effect by neglecting how delay propagates in a queuing system, this paper develops and validates a new estimation method based on queuing theory, machine learning and causal inference. Wait time information displayed to low-acuity patients provides a quasi-randomized instrumental variable. This paper shows that a low-acuity patient increases wait times for high-acuity patients through: pre-triage delay; delay of lab tests ordered for high-acuity patients; and transition delay when an ED interrupts treatment of a low-acuity patient in order to treat a high-acuity patient. Hence high-acuity patients' wait times could be reduced by: reducing the standard deviation or mean of those transition delays, particularly in bed-changeover; providing vertical or "fast track" treatment for more low-acuity patients, especially ESI 3 patients; standardizing providers' test-ordering for low-acuity patients; and designing wait time information systems to divert (especially when the ED is highly congested) low-acuity patients that do not need ED treatment. |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:ecl:stabus:3281&r= |