ePrints.FRI - University of Ljubljana, Faculty of Computer and Information Science

Predicting positions of defender players using neural networks

Andrej Grah (2010) Predicting positions of defender players using neural networks. EngD thesis.

[img] PDF
Download (2685Kb)


    This thesis introduces the problem of predicting positions of defender players using neural networks. To predict the positions of the defender players, a model of a neural network - multilayered perceptron, was used. We analysed the spatial presentation of the pitch or the position of players for the determination of training and testing patterns. In order to simplify the definition of patterns, we developed an application that is used to record and to visualize entry patterns. A neural network with the architecture and learning phase is introduced. According to a different presentation of the space or positions of the players, the suitable formats of the patterns, are defined. We performed simulations with diffrent formats of patterns. Evaluation of the results, based on positions that are known in advance, is given by the expert. For patterns defined in a system of coordinates, which gave the best results amongst all, we performed additional testing with three layered perceptron. By using additional testing, the goal was to find optimal architecture by changing the number of neurons in a neural network.

    Item Type: Thesis (EngD thesis)
    Keywords: indoor hockey, free push, neural networks, 3-layered perceptron, learning, prediction
    Number of Pages: 47
    Language of Content: Slovenian
    Mentor / Comentors:
    Name and SurnameIDFunction
    doc. dr. Mira Trebar252Mentor
    Link to COBISS: http://www.cobiss.si/scripts/cobiss?command=search&base=50070&select=(ID=7629140)
    Institution: University of Ljubljana
    Department: Faculty of Computer and Information Science
    Item ID: 1041
    Date Deposited: 18 Mar 2010 12:58
    Last Modified: 13 Aug 2011 00:36
    URI: http://eprints.fri.uni-lj.si/id/eprint/1041

    Actions (login required)

    View Item