Optimized computing of parameters for functional regression in data mining

The present article deals with the issue of training models in data mining for numeric attributes. It is focused mainly on optimization techniques for training, used in cluster infrastructure. The approach taken in the paper use gradient optimization techniques from numerical mathematics, to train t...

Full description

Saved in:
Bibliographic Details
Main Authors: Krammer, P., Hluchy, L.
Format: Conference Proceeding
Language:English
Subjects:
Computational modeling
Data mining
Data models
Functional regression
Mathematical model
Numerical models
Numerical prediction
Optimization
Predictive models
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The present article deals with the issue of training models in data mining for numeric attributes. It is focused mainly on optimization techniques for training, used in cluster infrastructure. The approach taken in the paper use gradient optimization techniques from numerical mathematics, to train the model, whose structure can be defined generally. This approach is advantageous because it is not limited by particular structure of the model. Applying this approach with different structures of models to the same specific data, we can observe significant changes of model quality, expressed by several numerical characteristics. All the numerical characteristics of quality are strictly defined in the first part of the article, which also gives a brief overview of predictions in data mining. The article also presents several parallelization techniques of this approach. Numerical prediction can be applied in various sectors - meteorology, hydrology, ecology, chemical and physical processes, industry, and many other areas.
DOI:10.1109/FSKD.2012.6234314