Refine
Has Fulltext
- yes (26)
Year of publication
- 2019 (26) (remove)
Document Type
- Master's Thesis (26)
Language
- English (26)
Is part of the Bibliography
- yes (26)
Keywords
- Deep Learning (5)
- LSTM (5)
- Transfer Learning (4)
- CNN (2)
- GloVe (2)
- Neural Networks (2)
- Time-series Forecasting (2)
- ULMFiT (2)
- A/B experiment (1)
- ARIMA modelling (1)
Institute
- Business Intelligence and Process Management M.Sc. (26) (remove)
This thesis analyzes whether financial news articles can predict the stock prices of companies and the S&P 500 Index. The selected companies for the analysis are Facebook Inc., Apple Inc., Microsoft Corp., Google (Alphabet Inc.), and Amazon.com Inc. This thesis evaluates the predictive power of financial news by comparing the sentiment of financial news articles published on a business day to
the corresponding closing value of stock prices and the S&P 500 Index. For the analysis, financial news from various resources is used for a time period between January 2018 and May 2018. The sentiment of the news articles is analyzed using a lexicon-based approach. Then, regression models are used to predict the stock market. The models used for forecasting the stock prices and the S&P 500 Index
are Auto-Regressive Integrated Moving Average (ARIMA), Support Vector
Regression (SVR), and Linear Regression (LR). The results of the research in overall indicate that ARIMA performs better for predicting the stock prices and the S&P 500 Index.
In this research paper, the main focus was to concerns the role of Business Intelligence (BI) in decision-making processes in organizations. The biggest goal of this thesis is to explore how decision-makers use and deploy this BI output to structure collective vision and achieve organizational choices. Business intelligence
can be described not as a process or a product or even a framework, but as a new strategy in organizational architecture based on velocity in data analysis in order to make correct strategic choices in company with maximum performance in a minimum quantity of time. After a study evaluating a conceptual model of the impact of business intelligence on strategic choices was constructed together with interviews among decision-makers. In order to evaluate the model, few questions
were raised and later solved during the studies. The research concentrated on knowing how the information supplied by the business intelligence system were used by the staff in the organization. The project focused on conducting in-depth interviews in a diverse organization with managers with six different backgrounds.
Each of the interviewers make decisions based on the output of BI.
In recent years, music genre classification has been studied widely within the musical information retrieval community to detect music genre (e.g., pop, rap) automatically. The existing methods, reported in the literature, usually extract features from the melodic content or lyrics of the song and address this classification as a multi-class problem. This thesis presents a comprehensive investigation of the prediction of the music genre solely from an examination of the lyrical content of the songs. The lyrics were thoroughly analyzed to obtain the features as inputs to the various machine learning algorithms and the features were represented using tf-idf values. In order to perform the algorithms, a dataset with a total of 12,000 songs for 12 genres was created by crawling the music websites.
Furthermore, this study considers the genre classification as a multi-label task in which a song can belong to more than one genre as it is encountered. Therefore, multi-label approaches were examined in-depth and applied along with the popular classifiers. The experiments in this thesis show that binary relevance method conducted with logistic regression model outperforms among all for the lyric-based multi-label genre classification.
Search Engine Opt imisation, also known as SEO, is one of the online marketing channels t hat, when it is s et u p s uitably, it could continue to pay dividends over time without investment. Recently, SEO teams of some companies keep investigating historical data to predict the future trends of revenue associated with the number of clicks, number of imp ressions, and number of sear ches which is prop osed to help with companies’ quality planning and campaign investment. Some of the challenges experienced by SEO analysts when attempting to forecast the revenue is that there is currently no way to standardize or forecast customers’ behaviour, which means the trends could be different every day, month, and year. In this research, SEO traffic data from one of the online travel a gencies are collected for the purpose of data exploration, analysis and forecasting which are expected to bring business values and give some be neficialinsights. Moreover, different time-series forecasting models are selected to conduct experiments seeking the best fit model for SEO data; Autoregressive Integrated Moving Average (ARIMA) model is initially performed, followed by Long-Short Term Memory (LSTM) of Recurrent Neural Networks (RNNs). As a result, it is proved that ARIMA is yet a classical statistics model but powerful enough for such small-size data, albeit the data is non-stationary and has too much white noise. Me anwhile, the LSTM is a deep learning tool which could deal with different types of data, but still need to be applied with a larger size of data to prove its competence.
Process models are often used as a Knowledge Management method, because these are able to store, visualize, and distribute knowledge within an organization. These knowledge-intensive processes can have a flexible, unstructured form, which is often hard to represent within a process model. The already available approaches of process-oriented Knowledge Management methods, which are considered for Business Process Management, are conducted manually and can therefore become time-consuming and labor-intensive. In addition, in times of Big Data, it is even more challenging to consider all possible cases. Due to the possibility of automatically generating process models through Process Mining, there is a huge potential for creating these processes with real information from event data. This study identifies the possibilities of generating knowledge-intensive process models through Process Mining. At the same time, it addresses the issue of the lack of representation of discovered flexible process models. Although few articles already appeal to the challenge of generating knowledge-intensive processes through Process Mining, the current main focus lies within the operational support in the information systems. The research question of this study is whether actual Process Mining approaches are able to generate process models that are as informative as the models created by processoriented Knowledge Management methods in their current state. To answer the research question, a comprehensive literature review as well as expert interviews have been conducted. Both approaches are part of the design science research methodology, which has been followed during the whole study. The theoretical results have been compared to the practical insights gained from the interviews. As a result, an informal, textual, best practice specification for mining knowledge-intensive process models within a Process Mining project has been developed. This should be considered for implementing an Enterprise Knowledge Medium into the already available IT infrastructure, so that process knowledge can be effectively saved and leveraged through Process Mining.
This thesis contributes to the growing interest of mobile and gait-based authentication. A realworld authentication system needs to ensure performance stability, even if the walking situation of phones holder changes. In this paper we analyze different circumstances of an environmental nature, such as phone placement (pocket position and orientation), clothing (trousers’ type, trouser width, trouser pocket’s distance to the user’s hipbone, shoe and bag), surrounding(location, surface) and walking style (walking speed, direction, group walk). All these labels were recorded by the theme ”A Walk Through Berlin”. 24 participants were equipped with a Samsung Galaxy S7 smartphone and collected their walking behavior at up to three locations in Berlin, Germany in at least two day-independent workshop sessions. Besides data collection, this research addresses limitations in regard to a one-class classification problem. An application should process data directly on the device itself and should not share these sensitive data streams with external parties or servers. We execute feature preprocessing and extraction in sliding
windows and use a one-class Support Vector Machine for user classification. After best feature evaluation, we gain an Equal Error Rate of 21% for model’s total performance, where train and validation set are enrolled on different days.
As an innovative tool, gamification has gained considerable importance in recent years to navigate and find many creative solutions to the problems and challenges faced by existing companies.
In this study, a game toolkit designed by using gamification elements that can potentially serve as a guide to creative ideation is introduced and applied in a data-driven industry. The aim is to
analyze the effects of the specifically designed game toolkit and the applied game elements on creative ideation and the attributes of users in general. The toolkit is explored using insights obtained from the analyses and observation of the behaviors of 16 players during the actual experiments. Furthermore, interviews and online surveys are used to conduct the quantitative and qualitative analyses in order to enhance the game within the scope of Action Design Research.
The toolkit proposed by the researcher is assumed to contribute to the existing literature. It can provide guidance and give valuable suggestions to possible future studies about the creation and
execution of a game toolkit in a field focused on data-driven innovations. Moreover, the results specifically indicate that the game toolkit directly drives creative idea generation while promoting engagement, motivation, and enjoyment as well as altering mindsets and thinking patterns.
Companies, public services and other institutions are increasingly turning to web-based applications, but attacks are increasing in both number and variance.
Previous approaches to avoid attacks by using web application firewalls rely primarily on pattern-based detection. This document evaluates if and which machine learning methods can be used to reliably detect web-based attacks. Classifiers such as Support Vector Machines, Neural Networks, Naïve Bayes, Decision Trees and Logistic Regression are used.
Furthermore possible use cases and visualizations of the decisions are suggested.
This thesis aims to investigate the relationship between opinions of politicians expressed in speeches and politicians’ paid side-activities of politicians, a potential channel of influence for lobbyists. With a focus on methodology development to automatically extract opinions and opinion changes for a specific item on the political agenda, the thesis reveals insights hidden in political speeches. Advancements in text mining present new opportunities to design advanced algorithms to achieve the stated objective. Along the journey of developing the methodology, the richness of used and possible algorithm components is discussed.
Poker is an interesting environment for artificial intelligence research. It is a game of imperfect information with multiple competing agents facing each other in a world of limited information, risks and deception. The observed average playrate is a commonly used indicator for holecard selection patterns of agents in Texas Hold´em poker. It serves as a fundamental feature in any modeling approach. In this paper we discuss why the average playrate is a biased metric and derive an unbiased new metric, the “maximum willingness differential” from empirical data. Finally the metric is formalized mathematically in a way that one can derive it from the observed playrate. Further implementation of this new metric could
enhance agent-modeling techniques and ahead/behind-research.
In recent years, terrorism has taken a whole new dimension and becoming a global issue because of widespread attacks and comparatively high number of fatalities. Understanding the attack
characteristics of most active groups and subsequent sta- tistical analysis is, therefore, an important aspect toward counterterrorism support in the present situation. In this thesis, we use a variety of data mining techniques and descriptive analysis to determine, examine and characterize threat level from top ten most active and violent terrorist groups and then use machine learning algorithms to avail intelligence toward counterterrorism support. We use historical data of terrorist attacks that took place around the world between 1970 to 2016 from the open-source Global Terrorism Database and the primary objective is to translate terror incident related information
into actionable intelligence. In other words, we chase the trajec- tory of terrorism in the present context with statistical methods and derive insights that can be useful.
A major part of this thesis is based on supervised and unsupervised machine learning techniques. We use Apriori algorithm to discover patterns in various groups. From the discovered patterns, one of
the interesting patterns we find is that ISIL is more likely to attack other terrorists (non-state militia) with bombing/explosion while hav- ing resulting fatalities between 6 to 10 whereas Boko Haram is more likely to target civilians with explosives, without suicide attack and resulting fatalities more than 50. Within the supervised machine learning context, we extend the previous research in time-series forecasting and make use of TBATS, ETS, Auto Arima and Neural Net- work model. We predict the future number of attacks in Afghanistan and SAHEL region, and the number of fatalities in Iraq at a monthly frequency. From time- series forecasting, we prove two things; the model that works best in one time-series data may not be the best in another time-series data, and that the use of ensemble significantly improves forecasting accuracy from base models. Similarly, in the classi- fication modeling part, previous research lacks the use of
algorithms that are recently developed. We also extend the previous research in binary classification problem and make use of a cutting-edge LightGBM algorithm to predict the probability of suicide attack. Our model achieves 96% accuracy in terms of AUC and correctly classifies “Yes” instances of suicide attacks with 86.5% accuracy.
In this thesis, a conceptual framework for content marketing is developed to broaden the understanding of content marketing and its role in enhancing customer value and, as a result, improve the return on investment (ROI) in marketing. The author explores definitional aspects of content marketing and identifies its potential for process automation. The author emphasizes the need for a cross-functional, process-oriented approach that positions content marketing at a strategic level.
Today, many organizations maintain a variety of systems and databases in a complex architecture that does not seem to fulfill the needs for a cross-functional, process-oriented approach. To address this challenge the paper aims to identify the key generic processes relevant to content marketing automation. Literature was examined to identify an appropriate categorization of required processes. Three cross-functional processes were identified to develop them into a conceptual framework for strategy development and implementation: Lead Management, Asset Management and Campaign Management. A new conceptual framework is developed based on these processes and the role and function of each module is explored. Furthermore, the interacting software systems and their functions for the process architecture are identified.
The literature review found that few reference process models exist. Those that did were not based on a process-oriented cross-functional conceptualization of content marketing. This gap in literature suggests that there is a need for a new systematic process-based content marketing strategy framework. Synthesis of the diverse concepts in the literature on content marketing and marketing automation into a single, process-based framework should provide practical insights to help companies achieve greater ROI with their content marketing automation strategy development and implementation.
The developed reference processes in this paper provide a starting point for the conceptualization, but the author recommends further exploration and modification of the reference model regarding its application.
The paper tries to provide a first data-informed taxonomy of Design Thinking. In addition to that it presents a set of prototypes of tools that could be used during data-oriented Design Thinking Workshops, or Data Thinking Workshops (DTW). New products and services increasingly follow a data-driven strategy. This creates the need for designers to create products and services with data in mind at the beginning of the
innovation circle. A common technique used is Design Thinking that has a very narrow view on data so far. Hence, the first part of the paper analyzes the core principles of design processes and derives a data-informed Design Thinking taxonomy. Additionally, the paper provides a set of tools that incorporate this data-informed Design Thinking taxonomy. Using an action design research approach, the DTW format is tested and the results are analyzed using a triangulation approach.
The suggested data-driven Design Thinking taxonomy and the proposed format of DTWs and the provided tools are of potential benefit for designers focusing on developing more digital, data-driven products and services. These help designers sharpen their perspective on data
challenges and come up with a more holistic view on data in their products and services. While the tools suggested in this paper in general work, several newly raised questions and considerations for further research can be derived by the findings. It is hoped this paper will help designers to further develop the taxonomy as well as come up with new tools that can place data at the core of the design process.
This research thesis contributes to the growing interest of wireless sensor networks (WSN)in smart factory environments. About 80% of the German machinery and plant construction industry intensively attend to the topic in order to increase factory uptime, provide individual customer solutions and improve customer service (Berger, 2017). The implementation of wireless sensor networks thereby arises new challenges, due to the high distribution of sensor units and the special requirement of low energy consumption while still providing sufficient network transmission capacities. Great research efforts are proceeded to investigate the behaviour of wireless sensor networks for specific applications and provide more comprehensive interfaces and protocols. This research study examines a specific use case of acceleration sensor nodes, which are used in cable carriers to predict likely maintenance times. Inaccurate and faulty measurements,
transmission errors and different signal transformations make accurate predictions impossible. In that regard, various signal processing steps are conducted and evaluated in this thesis to compensate the resulting errors. Furthermore, a novel approach of estimating
the remaining error through machine learning in form of a linear regression is conducted. The results show, that the margin of error can be sufficiently estimated by an appropriate reference measurement as target value for the model training. Results and findings as well as recommendation for further investigations are provided.
An increasing amount of research papers published every year makes it more and more difficult for researchers to find papers that address the same or a similar topic as their own current work. Electronic online databases provide a good starting point when trying to get an overview of the papers published regarding a certain topic. The results presented by the online libraries, however, vary greatly in terms of relevance to the intended search query. In addition, most online libraries provide thousands of search results, of which a huge amount is not relevant for the user at all. There are recent approaches to increase the quality of the search results when looking for relevant research papers online. While most current search engines focus mainly on the title of the papers as well as a few tags, these approaches try to use other information found in the papers as their input.
This master’s thesis focuses on undirected graphs and presents a first prototype for a software that translates images of graphs into a metalanguage. Since many scientific publications contain graphs, they could be a possible future input for search algorithms. For this purpose, an image recognition model consisting of a convolutional neural network and two long short-term memory networks was developed in collaboration with Badr Bih. The model itself is based on a similar software prototype called pix2code developed by Tony Beltramelli in 2017. For computational reasons, we had to limit our graphs to a maximum of four nodes and trained our model on them. The results of the model are promising, but it will have to be trained again on graphs with more vertices to make a definitive evaluation of the approach possible.
In the world of internet marketing, search engine optimization is a popular term. Getting higher rankings on the search engine and thereby getting more views to advertiser’s site is basically what it is. However, those views will not mean a lot if they do not lead to sales or conversion. Search advertisers join an online auction in order to get a slot in the search engine results pages. This means, in Pay per Click online model, advertisers have to pay to the search engine for the number of clicks on the advertisement they posted. Therefore, predicting conversion likelihood for advertisers is highly crucial for the sake of the revenue. Besides, being able to understand and track the conversion rates not only allows advertisers to measure the performance of the web pages but also to identify areas for improvement. In this study, machine learning is used for predicting the conversion rate using search term queries in google shopping ads. The purpose is to analyse if a pattern on search term queries to predict conversion rate can be observed. Search term queries with a high probability of better conversion rate can be advertised more. While the bids on low-conversion segments can be lowered to reduce the costs. For this purpose, extracting features from the text to represent it in a way that can be understood by the machine like term frequency and paragraph vector are tested on machine learning and deep learning model. The results show that different patterns of search term queries do not lead to a predictable conversion rate in the specific use case. Thus, the search term itself is no indicator for good or bad conversion rate.
Sequential Statistical Testing Procedures as an Early Stopping for Binomial Bandit Experiments
(2019)
This study tries to provide an early stopping procedure on binomial bandits which is a type of multi-arm bandit experiment. In addition to that, it presents sequential statistical testing procedures which can be used as early stopping criteria for A/B experiment. The paper searches for the applicability of these procedures for binomial bandit case because multi-arm bandit experiment and A/B experiment are similar in the sense that rewards can be simulated as independent identical Bernoulli distribution.
It is often claimed that multi-armed bandit which use Thompson sampling requires dramatically less sample size than A/B testing while still controlling the type 1 and type 2 error rates on alpha and beta due to the concept of always switching to the better arm. However, it is also claimed that Bayesian procedures are not immune to peeking (early stopping) because of the structure of sequential testing.
The sequential statistical testing procedures reduce the required number of observations and allow the experiment to stop early when the collected data is good enough to make a conclusion. In this work, Wald’s SPRT and Max SPRT-I-AA (a modified version of Max SPRT) sequential statistical testing procedures are studied for the A/B testing, and their applicability to binomial bandit case is researched.
According to simulation results, Max SPRT-I-AA sequential statistical testing procedure perform well for the A/B test scenarios. The type 1 and type 2 error rates are remains on the acceptable level and the experiment time is reduced more than 50 %. However, for the binomial bandit case results are not so satisfactory and brings more questions about the applicability of Max SPRT-I-AA to multi-arm bandits because of the performance of upper boundary calculation.
Hypothesis Extraction from Academic Papers Using Neural Networks for Ontology Theory Learning
(2019)
In this study, we investigated a new use case of deep learning. We applied deep learning to extract causes and effects from the hypotheses of the scientific papers. The research presents a variety of RNN models, including RNN models with CRF layer for labelling the sequences. We used such models as Bi-LSTM, LSTM, SimpleRNN and GRU. The experiments were conducted with GloVe vector representation and character level vector representation of words. Moreover, along with RNN models, we evaluated various hyperparameters and model setups to achieve the highest performance scores. In the end, we obtained promising results and shared our thoughts on the future prospects of the following studies.
This thesis examines the effectiveness of the latest Transfer Learning techniques for Natural Language Processing applied to the classification of research methods used in scientific journals in the domain of Information Systems. The task of automated knowledge extraction from academic articles has seen ongoing progress in recent years. However, the combination of transfer and Deep Learning in order to assign research methods to scientific papers has not been addressed in the literature yet. The main contribution of this thesis is, therefore, an artifact that applies cutting-edge Transfer Learning techniques to a Deep Learning model by conducting several experiments and comparing their effectiveness. The prototype considers various ways of fine-tuning that are crucial to retain the knowledge transferred from pretrained models and avoid catastrophic forgetting. Additionally, this work discusses the literature with regard to the task-specific Theory Ontology Learning and the method-specific state of the art in Transfer Learning for Natural Language Processing. As a result, the artifact surpassed the performance of previously developed models for research method extraction, presented in the literature, without applying any custom feature engineering and only using around a thousand of labeled observations.
This research paper explores the application of the Universal Language Model Fine-tuning (ULMFiT) technique, a novel deep transfer learning approach in the Natural Language Processing (NLP) field, to the financial statements fraud detection task. Additionally, the artifact investigated a simpler model, represented by a one-dimensional Convolutional Neural Network (CNN). Both methods have been assessed with respect to the training time and predefined evaluation metrics. Overall, ULMFiT turned out to be considerably more computationally expensive to train and achieved an accuracy of 77% with F-measure of 13%, if used with a decision boundary of 0.5, and accuracy of 59% with F-measure of 42%, if calculated with a threshold of 0.2. In contrast, CNN model trained significantly faster and obtained the following metrics: accuracy of 82% with F-measure of 42% for threshold 0.5, and accuracy of 83% with F-measure of 58% for threshold 0.2. As a result, ULMFiT has been outperformed by one-dimensional CNN on all examined metrics. The results are reported by using two different values for decision boundary due to the precision-recall trade-off, depending on the use case. In addition, this thesis investigated the impact of data preprocessing. The findings have shown that removing all numbers and special symbols with supplementary text truncation, limiting the sequence length, had a positive effect on both models, mentioned above.
In this study, we investigate the usage of generative adversarial neural networks(GAN) for a sequence labeling task. We applied sequence generative adversarial neural network (SeqGAN) to extract cause-effect and moderator-mediator relations from hypotheses from scientific papers. This research focuses on the structure of SeqGAN and the problems that come with it. However, there are two main problems with GAN’s. Firstly, a vanilla GAN (Goodfellow et. Al. 2014) is designed for generating real-valued, continuous data but we want to label discrete words into tokens. The second problem is that a GAN can only provide a loss for a complete sequence (Yu, Zhang, Wang, & Yu, 2017, p. 1). To address these problems two things are changed. First, the sequence generation process is realized as a decisionmaking process. The sequences are going to be evaluated by the discriminator model. To solve the problem with the discrete data we follow Yu et. Al. and regard the generative model as a policy gradient. To approximate the sequence value a Monte Carlo Search is employed in the generative model. We show the performance of different parameter settings and which tricks improve the results. The python and r scripts are appended in this document. They can also be found in GitHub (https://github.com/clamkewitz/GANCause).
A Deep Reinforcement Learning Agent Using Multiple Assets Financial Signals for Portfolio Management
(2019)
In this study, we investigated possible applications of reinforcement learning in the area of portfolio management. Aa specific topology of reinforcement learning is choosen to study the feasibility, Deep Deterministic Policy Gradient (DDPG), to train neural networks to perform trade in an environment simulated real world trading. The results show that the DDPG agent is able to learn price pattern and perform profitable trades. In single stock backtest, the DDPG is able to generate an annual return of 7%. While in multiple stocks backtest, DDPG agent can generate an annual return of 12%.
Deep Learning, a topic of broad and current interest, has undergone rapid development in the last decade. The performance of the algorithms is already above human level. However, the area of Natural Language Processing is still a great challenge for the researchers. This study aims at exploring the novel ULMFiT method for text classification by applying it to a dataset of scientific articles with advanced rhetorical categories. This classification task is challenging even for humans and it requires a substantial analysis of the texts when conducted by machine learning algorithms. The objective of this attempt is to achieve text classification with minimal preparation. The ULMFiT method is the first successful effort to apply transfer learning to NLP tasks. Its performance will be evaluated on a task that requires a level of understanding beyond the semantic meaning of the text.
Blockchain technology is often associated with Bitcoin, but the technology enables more possibilities, such as smart contract which offers the opportunity to conclude contracts between at least two parties without a central instance e.g. a bank or notary. However, this technology is relatively new and undiscovered. It remains to be clarified whether the technology can establish in the business world. This thesis analyses the feasibility of this approach. Social, legal and technical aspects and their challenges are considered. The aim is to discuss the use of smart contracts, to examine the feasibility of establishing and to give recommendations for further measures. For this, several literatures were read to understand the different aspects and to make an overall assessment. In general, it can be said that time is needed to understand the technology and the resulting new opportunities and thereby to build trust. Finally, it can be argued that in all three aspects changes are needed to discover the full potential of the technology. Data protection should become less stringent to not hinder further progress, technology must continue to research and set standards, and society should be educated to overcome the fear of innovation.
The goal of this research is to investigate the use of deep convolutional neural networks for racing bib number recognition in sport images. Several deep neural network architectures are studied. Three final architectures are trained on three different sets of data: 1) Street View House Numbers (SVHN) Dataset, 2) A private dataset from Flashframe.io from different running events, and 3) A combination of dataset 1 and 2. This thesis investigates the performance that can be obtained on racing bib numbers from a neural network that has been trained on solely images from street house numbers, on a mixture of SVHN and RBN images as well as only on RBN images. The motivation behind this is to see how well this problem can be solved by transfer learning, as labelled images of racing bib numbers are scarce.
The models are tested on the RBNR Dataset (Ben-Ami et al., 2012) and a subset of the private dataset from Flashframe.io. The study shows that the best recognition results were obtained by a model trained on the hybrid dataset of all SVHN images plus an additional 50.000 images. This model outperformed the models that had been trained solely on the SVHN Dataset or the private racing bib number dataset.
The best model resulted in Recall of 0,92, Precision of 0,93 and F-measure of 0,93 on the RBNR Dataset (using the same formulas as previously reported on the RBNR dataset), and 0,97, 0,97 and 0,97 on the private dataset, respectively. The reported recognition results on the RBNR dataset are much higher than previously used methods and proves that neural networks can effectively be used for racing bib number recognition.