Stefan Feuerriegel

Emergency Response in Natural Disaster Management: Allocation and Scheduling of Rescue Units

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%.

Putting Big Data analytics to work: Feature selection for forecasting electricity prices using the LASSO and random forests

Successful companies are increasingly those companies that excel in the task of extracting knowledge from data. Tapping the source of ‘Big Data’ requires powerful algorithms combined with a strong understanding of the data used. One of the key challenges in predictive analytics is the identification of relevant factors that may explain the variables of interest. In this paper, we present a case study in predictive analytics in which we focus on the selection of relevant exogenous variables. More specifically, we attempt to predict the German electricity spot prices with reference to historical prices and a deep set of weather variables. In order to choose the relevant weather stations, we use the least absolute shrinkage selection operation (LASSO) and random forests to implicitly execute a variable selection. Overall, in our case study of German weather data, we manage to improve forecasting accuracy by up to 16.9% in terms of mean average error.

Enhancing Sentiment Analysis of Financial News by Detecting Negation Scopes

Sentiment analysis refers to the extraction of the polarity of source materials, such as financial news. However, measuring positive tone requires the correct classification of sentences that are negated, i.e. The negation scopes. For example, around 4.74% of all sentences in German ad hoc announcements contain negations. To predict the corresponding negation scope, related literature commonly utilizes two approaches, namely, rule-based algorithms and machine learning. Nevertheless, a thorough comparison is missing, especially for the sentiment analysis of financial news. To close this gap, this paper uses German ad hoc announcements as a common example of financial news in order to pursue a two-sided evaluation. First, we compare the predictive performance using a manually-labeled dataset. Second, we examine how detecting negation scopes can improve the accuracy of sentiment analysis. In this instance, rule-based algorithms produce superior results, resulting in an improvement of up to 9.80% in the correlation between news sentiment and stock market returns.

Analysis of How Underlying Topics in Financial News Affect Stock Prices Using Latent Dirichlet Allocation

Companies listed on the stock markets are typically obliged to publicly disclose any information that might have a significant influence on their stock prices. This transparency regulation is intended to ensure that all market participants have access to the same information. The corresponding press releases are one of the most reliable news sources concerning a companys operations. Interestingly, even though researcher have investigated the timing of releases, research has invested little effort into examining the underlying news topics. In this paper, we analyze the effects of topics found in such corporate press releases on stock market returns in the German market. We determine the topic of ad hoc announcements by using Latent Dirichlet Allocation. Effectively, we succeed in extracting 40 topics. As hypothesized, the effect of these topic groups differ greatly from each other. Some topics have no resulting effect on abnormal returns of stocks, whereas other topics, such as drug testing, exhibit a large effect.

Decision support from financial disclosures with deep neural networks and transfer learning

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 short-term 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. We apply deep learning for sentiment analysis of financial news.This yields considerable improvements in forecasting the stock price movements.Additional gains stem from applying transfer learning.Long short-term memory performs overall best.Demonstrates the potential for future applications of deep learning in finance

News-based trading strategies

The marvel of markets lies in the fact that dispersed information is instantaneously processed and used to adjust the price of goods, services and assets. Financial markets are particularly efficient when it comes to processing information; such information is typically embedded in textual news that is then interpreted by investors. Quite recently, researchers have started to automatically determine news sentiment in order to explain stock price movements. Interestingly, this so-called news sentiment works fairly well in explaining stock returns. In this paper, we design trading strategies that utilize textual news in order to obtain profits on the basis of novel information entering the market. We thus propose approaches for automated decision-making based on supervised and reinforcement learning. Altogether, we demonstrate how news-based data can be incorporated into an investment system. Financial disclosures are the main source for the decision-making in finance.Sentiment analysis of financial disclosures can provide decision support.We design and compare different strategies for news trading.These can outperform our benchmarks in terms of profits but at the cost of risk.Especially viable approaches are supervised and reinforcement learning.

Negation scope detection in sentiment analysis

Decision support for financial news using natural language processing requires robust methods that process all sentences correctly, including those that are negated. To predict the corresponding negation scope, related literature commonly utilizes rule-based algorithms and generative probabilistic models. In contrast, we propose the use of a tailored reinforcement learning method, since it can conquer learning task of arbitrary length. We then perform a thorough comparison with a two-pronged evaluation. First, we compare the predictive performance using a manually-labeled dataset. Here, reinforcement learning outperforms common approaches from the related literature, leading to a balanced classification accuracy of up to 70.17%. Second, we examine how detecting negation scopes can improve the accuracy of sentiment analysis for financial news, leading to an improvement of up to 10.63% in the correlation between news sentiment and stock market returns. This reveals negation scope detection as a crucial leverage in decision support from sentiment. Enhance sentiment analysis of financial news by detecting negation scopesImprovement of up to 10.63% in the correlation between sentiment and stock returnComparison across different sets of negation words and various methodsImplement reinforcement learning, Hidden Markov models, conditional random fields and rule-based methods

Do Investors Read Too Much into News? How News Sentiment Causes Price Formation

It is a well-known fact that financial markets react to information. Even though this relationship seems simple, finding evidence is not easy since information is embedded in textual news releases. Only recent have researchers started to look at the content of news. Interestingly, previous work avoids the inference of a causal relationship between news messages and abnormal returns. In this paper, we concentrate on the oil market from which we identify a (strong) instrument, namely, the number of terroristic attacks, and can reasonably account for the endogeneity problem associated with news releases. In addition, we study how news releases affect stock prices temporally. Thus, we find that a change in news sentiment entails a large change in oil prices.

News Processing during Speculative Bubbles: Evidence from the Oil Market

Speculative bubbles are commonly referred to situations where stock prices considerably deviate from their fundamentals until the bubbles bust. Bursting of bubbles such as the dot-com or U.S. housing bubble is very costly, so there is a need for mechanisms to detect them. In this paper, we attempt to predict when bubbles may bust using the sentiment of news announcements. Accordingly, we first try to understand how news reception evolves depending on the market phase (boom or bust). The probability of bubble bursts are calculated on the basis of a Markov-regime switching model. The approach is applied and validated using the oil market which appears to be one of the most important markets in the globalized world. Our methodology can be similarly extended to other markets such as gold or wheat.

Detecting Negation Scopes for Financial News Sentiment Using Reinforcement Learning

Applying natural language processing to the domain of financial news requires robust methods that process all sentences correctly, including those that are negated. So far, related research commonly utilizes rule-based algorithms to detect negated sentence fragments, named negation scopes. Nonetheless, these methods involve certain limitations when encountering complex language or particularities of the chosen prose. As an alternative, reinforcement learning offers an opportunity to learn suitable negation classifications through trial-and-error experience. This method tries to replicate human-like learning and thus appears well-suited for natural language processing. Its episode-based and flexible structure allows for the handling of even highly complex sentences. Our results provide evidence that reinforcement learning can outperform rule-based approaches from the related literature. The best performing implementation reveals a predictive accuracy of up to 76.37% on a manually-labeled dataset, exceeding the predictive accuracy of rule-based approaches by 2.55 %. When utilizing the already trained reinforcement learning implementation for sentiment analysis, we find a potential subjectivity bias that limits the predictive performance of forecasting stock market returns.