Background: The novel coronavirus (SARS-CoV-2) has rapidly evolved into a global epidemic. To control its spread, countries have implemented non-pharmaceutical interventions (NPIs), such as school or border closures, while others have even enforced a complete lockdown. Here we study the effectiveness of NPIs in reducing documented cases of COVID-19. Methods: We empirically estimate the impact of NPIs on documented COVID-19 cases in a cross-country analysis. A Bayesian hierarchical model with a time-delayed effect for each NPI allows us to quantify the relative reduction in the number of new cases attributed to each NPI. Based on this model, a cross-country analysis is performed using documented cases through April 15, 2020 from n = 20 countries (i.e., the United States, Canada, Australia, the EU-15 countries, Norway, and Switzerland). Documented case numbers are selected because they are essential for decision-makers in the area of health-policy when monitoring and evaluating current control mechanisms. Findings: Based on our model, we compare the effectiveness of NPIs up to now, i.e., in the early stages of the outbreak. Venue closures are associated with a reduction in the number of new cases by 36 % (95% credible interval [CrI] 20-48 %), closely * Corresponding author. † These authors contributed equally. : medRxiv preprint followed by gathering bans (34 %; 95% CrI 21-45 %), border closures (31 %; 95% CrI 19-42 %), and work bans on non-essential business activities (31 %; 95% CrI 16-44 %). Event bans lead to a slightly less pronounced reduction (23 %; 95% CrI 8-35 %). School closures (8 %; 95% CrI 0-23 %) and lockdowns (5 %; 95% CrI 0-14 %) appear to be the least effective among the NPIs considered in this analysis.Interpretation: This cross-country analysis provides early estimates regarding the effectiveness of different NPIs for controlling the COVID-19 epidemic. These findings are relevant for evaluating current health-policies and will be refined as more data becomes available.

Natural disasters, such as earthquakes, tsunamis and hurricanes, cause tremendous harm each year. In order to reduce casualties and economic losses during the response phase, rescue units must be allocated and scheduled efficiently. As this problem is one of the key issues in emergency response and has been addressed only rarely in literature, this paper develops a corresponding decision support model that minimizes the sum of completion times of incidents weighted by their severity. The presented problem is a generalization of the parallel-machine scheduling problem with unrelated machines, non-batch sequence-dependent setup times and a weighted sum of completion times -thus, it is NP-hard. Using literature on scheduling and routing, we propose and computationally compare several heuristics, including a Monte Carlo-based heuristic, the joint application of 8 construction heuristics and 5 improvement heuristics, and GRASP metaheuristics. Our results show that problem instances (with up to 40 incidents and 40 rescue units) can be solved in less than a second, with results being at most 10.9 % up to 33.9 % higher than optimal values. Compared to current best practice solutions, the overall harm can be reduced by up to 81.8 %.

Company disclosures greatly aid in the process of financial decision-making; therefore, they are consulted by financial investors and automated traders before exercising ownership in stocks. While humans are usually able to correctly interpret the content, the same is rarely true of computerized decision support systems, which struggle with the complexity and ambiguity of natural language. A possible remedy is represented by deep learning, which overcomes several shortcomings of traditional methods of text mining. For instance, recurrent neural networks, such as long shortterm memories, employ hierarchical structures, together with a large number of hidden layers, to automatically extract features from ordered sequences of words and capture highly non-linear relationships such as context-dependent meanings. However, deep learning has only recently started to receive traction, possibly because its performance is largely untested. Hence, this paper studies the use of deep neural networks for financial decision support. We additionally experiment with transfer learning, in which we pre-train the network on a different corpus with a length of 139.1 million words. Our results reveal a higher directional accuracy as compared to traditional machine learning when predicting stock price movements in response to financial disclosures. Our work thereby helps to highlight the business value of deep learning and provides recommendations to practitioners and executives.

Emotions widely affect human decision-making. This fact is taken into account by affective computing with the goal of tailoring decision support to the emotional states of individuals. However, the accurate recognition of emotions within narrative documents presents a challenging undertaking due to the complexity and ambiguity of language. Performance improvements can be achieved through deep learning; yet, as demonstrated in this paper, the specific nature of this task requires the customization of recurrent neural networks with regard to bidirectional processing, dropout layers as a means of regularization, and weighted loss functions. In addition, we propose sent2affect, a tailored form of transfer learning for affective computing: here the network is pre-trained for a different task (i.e. sentiment analysis), while the output layer is subsequently tuned to the task of emotion recognition. The resulting performance is evaluated in a holistic setting across 6 benchmark datasets, where we find that both recurrent neural networks and transfer learning consistently outperform traditional machine learning. Altogether, the findings have considerable implications for the use of affective computing.