This thesis investigates the prediction of possible stock price changes immediately after news article publications, by automatic analysis of these news articles. Some background information about financial trading theory and text mining is given in addition to an overview of earlier related research in the field of automatic analyzes of news articles for predicting future stock prices. In this thesis a system is designed and implemented to predict stock price trends for the time immediately after the publication of news articles. This system consists mainly of four components. The first component gathers news articles and stock prices automatically from internet. The second component prepares the news articles by sending them to some document preprocessing steps and finding relevant features before they are sent to a document representation process. The third component categorizes the news articles into predefined categories, and finally the fourth component applies appropriate trading strategies depending on the category of the news article. This system requires a labeled data set to train the categorization component. This data set is labeled automatically on the basis of the price trends directly after the news article publication. An additional label refining step using clustering is added in an attempt to improve the labels given by the basic method of labeling by price trends.The findings indicate that a categorization of news articles provides additional information that can be used to forecast stock price trends. Experiments showed that the label refining method greatly improves the performance of the system. It was also shown that the timing of when to start the price trends used to label the data sets had a significant impact on the results. Trading simulations performed with the systems managed to gain positive returns (profits) on most of its trades. Some of the methods also managed to give better results than what trades performed with the manually labeled data set did.

This thesis presents a graph-based approach to the problem of text representation. The work is motivated by the need for better representations for use in textual Case-Based Reasoning (CBR). In CBR new problems are solved by reasoning based on similar past problem cases. When the cases are represented in free text format, measuring the similarity between a new problem and previously solved problems become a challenging task. The case documents need to be re-represented before they can be compared/matched.Textual CBR (TCBR) addresses this issue. We investigate automatic re-representation of textual cases, in particular measuring the salience of features (entities in the text) towards this end. We use the classical vector space model in Information Retrieval (IR) but investigate whether graph-representation and salience inference using graphs can improve on the Term Frequency (TF) and Term Frequency-Inverse Document Frequency (TF-IDF) measures, emph{bag of words} approaches predominant in IR.Our special focus is whether, and possibly how, the co-occurrence and the syntactic dependency relations between terms have an impact on feature weighting. We measure salience through the notion of graph centrality. We experiment with two types of application tasks, classification and case retrieval. Although classification is not a typical TCBR task, it is easier to find datasets for this application, and the centrality measures we have studied are not specific to TCBR. The experiments on this task are therefore relevant to the second application task which is our ultimate target. We test various centrality metrics described in the literature, make a distinction between local and global weighting measures and compare them for both application tasks. In general, our graph-based salience inference methods perform better than TF and TF-IDF.

Electroencephalography (EEG) equipment are becoming more available on thepublic market, which enables more diverse research in a currently narrow field.The Brain-Computer Interface (BCI) community recognize the need for systemsthat makes BCI more user-friendly, real-time, manageable and suited for peoplethat are not forced to use them, like clinical patients, and those who are disabled.Thus, this project is an effort to seek such improvements, having a newly availablemarket product to experiment with: a single channel brain wave reader. However,it is important to stress that this shift in BCI, from patients to healthy and ordinaryusers, should ultimately be beneficial for those who really need it, indeed.The main focus have been building a system which enables usage of the availableEEG device, and making a prototype that incorporates all parts of a functioningBCI system. These parts are 1) acquiring the EEG signal 2) process and classify theEEG signal and 3) use the signal classification to control a feature in a game. Thesolution method in the project uses the NeuroSky mindset for part 1, the Fouriertransform and an Artificial Neural Network for classifying brain wave patterns inpart 2, and a game of Snake uses the classification results to control the characterin part 3.This report outlines the step-by-step implementation and testing for this system,and the result is a functional prototype that can use user EEG to control the snakein the game with over 90% accuracy. Two mental tasks have been used to separatebetween turning the snake left or right, baseline (thinking nothing in particular)and mental counting. The solution differentiates from other appliances of the NeuroSkymindset that it does not require any pre-training for the user, and it is onlypartially real-time.

This thesis is the continuation of our specialization project, Lecture Quiz 2.5.This platform is a game-like system where lecturers can hold quizzes in lectures to increase student participation and interactivity.The current version is a finished lecture quiz system that can be used in lecture environments.Lecture Quiz 3.0 has moved away from earlier implementations, by centralizing and minimizing the effort to start and run quizzes.One focus was multi-platform and we developed the system to support Microsoft Windows, Mac OS X and Linux.This system can be used in lecture environments to promote more student participation, and enable variation in teaching methods.To run quiz games, the lecturer can use a PC, connect it to the projector, and run the Presentation Client.Students access the Player Client via a mobile device such as a smart phone or notebook, the address to the Player Client web page is presented on the Presentation Client.Once connected, they choose a username, and answer multiple choice questions, which are presented on the projector screen.To keep things interesting for the students, we focused on the visual expression of the Presentation Client and Player Client.This is to give the players the experience of playing a game, rather than answering a questionnaire.We developed the system with usability in mind.This is to ensure that the system feel easy to use, for both students and lecturers.One of the main goals is to make lecturers see the system as an alternative to a regular presentation, and not as extra work.A lecturer might be interested in collecting statistics about the students' overall progress in the course.This way they might be able to give a larger focus to the parts of the syllabus where the students lack performance.Another factor is that creating quizzes is time consuming, and needs to be done in advance of a lecture.We developed a separate quiz manager and statistics tool that can be used by lecturers, named Quiz Server.It is a Web based application, utilizing Java EE to enable multi-platform support.We performed an experiment in a lecture to get feedback from students on how they perceived the Lecture Quiz game.This experiment was performed by running a quiz in the lecture hall and then the students were asked to fill in an evaluation form.The students who participated thought that Lecture Quiz had a positive effect on the lecture.

This dissertation investigates the classification capabilities of artificial neural networks (ANNs). The goal is to generalize over the features of a writing system, and thus classify the writing system of a previously unseen glyph. The complexity of the problem necessitates a large network, which hampers the tuning of the weights. ANNs were created using three different hybrids of back-propagation (BP) learning and evolution, and a pure BP algorithm for comparison. The purpose was to find the method best suited for this kind of generalization and classification networks. The results suggest that ANNs are able to generalize enough to solve the classification task, but it is depending on the weight tuning algorithm. A pure BP algorithm is preferable to any of the hybrid algorithms, due to the size of the ANN. This algorithm had both the best classification results and the fastest runtime, in addition to the least complex implementation.

This report presents the work done to develop an autonomous driverfor the Simulated Car Racing Championship (SCRC), a competition incomputational intelligence based on The Open Racing Car Simulator(TORCS). Autonomous race driving based only on local sensory datais a complex problem, and previous SCRC entries' work show a widevariety of approaches taken to address it. We describe CRABCAR,a controller that combines oine learning prior to the competitionwith online learning during the competition to optimize its performance in the SCRC context. The presented approach extends and builds on track modelling and racing line optimization techniques introduced previously, addressing known problems said techniques have with noisy sensory input and non-perfect track information. CRABCAR's performance is compared to previous entries from the SCRC, with results showing CRABCAR at a performance level similar to the others. We conclude that a system for online adaption is essential when pre-learned strategies are applied to discretely segmented and non-perfect track models in the SCRC context.

In this report, we have explored swarm intelligence through a box-pushing taskwith physical robots called e-pucks. Research on social insects has been presentedtogether with dierent ways of controlling autonomous robots, where combiningthis knowledge has been essential in our quest to make a biological plausible antretrieving system.Inspired by ants and behavior-based robotics, we have created the system CRABS.It is based on Brooks' subsumption architecture to control six dierent behaviors,from a xed input-output scheme. The system is designed to easily handle addingor removal of behavior layers. Behavior modules can also be used separately andported to other software or hardware platforms.During this project we came across several hardware and software challenges in-vestigating cooperative behavior. With the use of the simulation tool Webots, wewere able to determine e-pucks' capabilities, and through this knowledge able todesign and construct an articial food source. This operated as the box-item in thebox-pushing task.Based on two types of sensors and two actuators (wheels), we had a strategy toaccomplish the box-pushing task following the biological principles of social insects.The guidelines of the ant retrieving model made CRABS a self-organized systemthat given three or more e-pucks, will always succeed in retrieving the box back tothe wall. The most remarkable view on this accomplishment is that is done throughthe use of only stigmergy and positive/negative feedback.One of the things we've experienced throughout this thesis is that hardware is a morework demanding and inconsistent platform than your usual software simulation.Everything is not given, and although Webots provided helpful shortcuts, a lot oftime and hard work was put down in order to get the system up and running. Withthat being said, we are pleased that we took the hardware rout and were able totest and validate our system on physical robots.

This thesis has evaluated the use of the computationally expensivemaximum-likelihood (ML) method coupled with an evolutionaryalgorithm (EA) for the problem of inferring evolutionaryrelationships among species (phylogenies) from molecular data. MLmethods allow using all the information from molecular data, suchas DNA sequences, and have several beneficial properties compared toother methods. Evolutionary algorithms is a class of optimizationalgorithms that often perform well in complex fitness landscapes.EAs are also proclaimed to be easy to parallelize, an aspect thatis increasingly more important.A parallel EA system has been implemented and tested on a clusterfor the task of phylogeny inference. The system shows promisingresults and is able to utilize processors of a massively parallelsystem in a transparent manner.

This thesis presents a case study of Game Mechanic based E-Learning. This is put forward as a new approach to E-Learning that tries to mimic games to harness some of their motivational properties. A prototype system was developed as a web application, using an Agile and Lean development approach.The system was evaluated with a class at the Norwegian University of Science and Technology.This was done to give an indication of the system's ability to make work with exercises more engaging and fun. To give context in this thesis, the growing trend of Gamification is unveiled and explained in detail.The major technological delivery posited by this thesis was the prototype, implementedas a web application (dynamic webpage). The major research acheivement was evaluatingrespondents perception of the system. It was discovered that the chosen Game Mechanicwas indeed considered to make work with exercises more engaging, although this effect wasmarginal. The evaluation was also used to arrive at a general definition for games.This definition can be used to distinguish Game Mechanics basedsystems from games. It also serves as a much needed guide to designing games andnon-game systems that tries to acheive similar motivational benefits as games.

In this project I have designed a model of features, human behavior and intentions. The model suggests a set of features that can be used to describe the interaction between a human being and an automated sliding door. The model also defines symbols representing value sets for the features. The symbols are then combined in order to describe different events, mapping features to intentions. This model provides a framework guiding the capturing process as well as the reasoning process.Further, I have designed a mechanism for capturing human movement and extracting the features as suggested by the model of features, human behavior and intentions. The solution components are based on research done within computer vision, where different tools and algorithms were reviewed and evaluated. Parts of the suggested solution are provided as software libraries, while others had to be implemented. The solution includes using an Xbox Kinect as a sensor device, and the OpenNI framework together with the middleware NITE for Human body tracking and skeletal joint extraction.A reasoning mechanism was designed, that utilizes the designed model in order to reach a conclusion about the intention of a human interacting with the door. Different reasoning techniques were reviewed in context of the sliding doors problem. Based on the review I suggest using rule-based reasoning. By using the events described in the model and by giving values to the different symbols I was able to form the rules for the reasoning process.The designed mechanisms were put together in an implementation in C/C++ comprising depth and RGB image capture, body tracking, user handling and feature extraction, rule-based reasoning and door control.A motorized sliding door was built, together with a door controller allowing a computer to interface with the door, giving open and close commands.Finally, the door was tested both through a live demo and a laboratory style, structured observation. The door proved a superior performance to the traditional sliding doors when it came to identifying negative intentions, thus reducing the number of false positives drastically. However, both false positives and false negatives occurred, leaving room for improved accuracy.With my solution I have managed to interpret the intention of a user interacting with an automated sliding door. I have lifted the reasoning process to a symbolic level, dealing with symbols and events easy to understand. Although the model is limited to a very specific domain, and the solution has got some limitations and weaknesses, this is a good starting point for further work.