Abstract
One more and more important issue to cope with in the analysis of large systems with heavy interaction load amongst its parts (i. e., subsystems) is to deal with unreliable and noisy data caused by unreliable, error prone, faulty, etc. sensing technology. This involves, aside the well known statistical approaches for error and/or noise reduction in data sets, the development of new means and data aggregation techniques to allow for both qualitative and quantitative compensation of missing, corrupted, unreliable, noisy, and manipulated data. In particular, the questions to be addressed within this deliverable are: 1. How can noisy (or dirty) data be defined?, 2. What are general data representation techniques suitable for noisy data, 3. How to query data from complex and dynamic environments?, 4. What are statistical approaches for dealing with noisy data? data cleaning, 5. What kind of sensor faults may appear (to SOCIONICAL data)?, 6. What are methods/techniques for detecting/correcting sensor faults, 7. Which approach is suitable for what problem (in the domain of SOCIONICAL)?, 8. How to avoid data manipulation; if happened, how to detect manipulated data?, and 9. To what extent should one trust information provided by AmI technology (technical systems)?
| Original language | English |
|---|---|
| Place of Publication | Altenberger Strasse 69, 4040 Linz |
| Publisher | Institut für Pervasive Computing |
| Number of pages | 175 |
| Publication status | Published - Jan 2011 |
Fields of science
- 102 Computer Sciences
- 102001 Artificial intelligence
- 102018 Artificial neural networks
JKU Focus areas
- Engineering and Natural Sciences (in general)