Software Lab - Practical Lab "Data Mining Techniques in Sensor Networks"

Year: SS 2010
Lecturer: Univ.-Prof. Dr. rer. nat. T. Seidl
Dipl.-Ing. M. Hassani
Dipl.-Inform. S. Fries
Type: Software Lab
Form: Praktikum
Language: EN, DE
Content:  

Clustering is an established data mining technique for grouping data based on similarity. This technique is useful for many applications such as wireless sensor networks (WSNs). WSNs consist usually of sensor nodes which are energy, processing and storage limited devices. Sensor nodes are usually distributed over a certain area covered by the WSN to observe phenomena (light, temperature, humidity, ...) by collecting data. Physical clustering of sensor nodes aims at grouping together sensor nodes that are sensing correlated data and selecting one of them as a representative, while turning others off. This minimizes the energy consumption and thus extends the lifetime of the nodes. The aim of our software project is to use Java for implementing existing data mining techniques and an evaluation framework. After the implementation, the evaluation framework will be used to measure how fast the sensor network detects changes in the observed phenomenon (like spreading of fire events). Another requirement is calculating the energy consumption of the total sensor network when having such events. After attending this lab, the following skills are supposed to be gained:

  • knowledge about basic concepts and methods of data mining techniques for sensor network
  • knowledge of the functionality and effectiveness of some data mining algorithms
  • judging the effectiveness of data mining solutions through evaluation
  • constructing a big software project through team work, including
    requirements analysis, modeling, implementation, testing and integration

Recommended skills:

  • knowledge from "Data structures and algorithms" lecture

Slides: 01-softwarepraktikum.pdf

Dates:  
Type Date Room
Seminar Thu, 14:00 - 15:30 Seminar room