Targeted projection pursuit

An example of targeted projection pursuit

Targeted projection pursuit is a type of statistical technique used for exploratory data analysis, information visualization, and feature selection. It allows the user to interactively explore very complex data (typically having tens to hundreds of attributes) to find features or patterns of potential interest.

Conventional, or 'blind', projection pursuit, finds the most "interesting" possible projections in multidimensional data, using a search algorithm that optimizes some fixed criterion of "interestingness" – such as deviation from a normal distribution. In contrast, targeted projection pursuit allows the user to explore the space of projections by manipulating data points directly in an interactive scatter plot.

Targeted projection pursuit has found applications in DNA microarray data analysis,[1] protein sequence analysis,[2] graph layout[3] and digital signal processing.[4] It is available as a package for the WEKA machine learning toolkit.

References

  1. ↑ Faith, Joseph; Robert Mintram; Maia Angelova (2006). "Targeted Projection Pursuit for Visualising Gene Expression Data Classifications" (PDF). Bioinformatics. 22 (21): 2667–267. doi:10.1093/bioinformatics/btl463. PMID 16954139.
  2. ↑ Haddow, Chris; Marcus Durrant; Justin Perry; Joe Faith (in press). "Predicting Functional Residues of Protein Sequence Alignments as a Feature Selection Task". International Journal of Data Mining in Bioinformatics. Check date values in: |year= (help)
  3. ↑ Gibson, Helen; Joe Faith (2011). "Node-Attribute Graph Layout for Small-World Networks". Proceedings of 15th International Conference on Information Visualisation.
  4. ↑ Sujan, Rajbhandari; Joe Faith (2010). "The Use of Linear Projections in the Visual Analysis of Signals in an Indoor Optical Wireless Link". IEEE, IET International Symposium on Communication Systems, Networks and Digital Signal Processing.

Further reading


This article is issued from Wikipedia. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.