Wavelet-based Data Reduction and Mining for Multiple Functional Data

Wavelet transformation;Data mining;Functional data;Signal processing

Jung, Uk ; Industrial and Systems Engineering

University:Georgia Institute of Technology

Department:Industrial and Systems Engineering

关键词: Wavelet transformation;    Data mining;    Functional data;    Signal processing;   

Others  :  https://smartech.gatech.edu/bitstream/1853/5084/1/jung_uk_200407_phd.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

Advance technology such as various types of automatic dataacquisitions, management, and networking systems has created atremendous capability for managers to access valuable productioninformation to improve their operation quality and efficiency.Signal processing and data mining techniques are more popular thanever in many fields including intelligent manufacturing. As datasets increase in size, their exploration, manipulation, andanalysis become more complicated and resource consuming. Timelysynthesized information such as functional data is needed forproduct design, process trouble-shooting, quality/efficiencyimprovement and resource allocation decisions. A major obstacle inthose intelligent manufacturing system is that tools forprocessing a large volume of information coming from numerousstages on manufacturing operations are not available. Thus, theunderlying theme of this thesis is to reduce the size of data in amathematical rigorous framework, and apply existing or newprocedures to the reduced-size data for various decision-makingpurposes. This thesis, first, proposes {it Wavelet-basedRandom-effect Model} which can generate multiple functional datasignals which have wide fluctuations(between-signal variations) inthe time domain. The random-effect wavelet atom position in themodel has {it locally focused impact} which can be distinguishedfrom other traditional random-effect models in biological field.For the data-size reduction, in order to deal with heterogeneouslyselected wavelet coefficients for different single curves, thisthesis introduces the newly-defined {it Wavelet Vertical Energy}metric of multiple curves and utilizes it for the efficient datareduction method. The newly proposed method in this thesis willselect important positions for the whole set of multiple curves bycomparison between every vertical energy metrics and a threshold({it Vertical Energy Threshold; VET}) which will be optimallydecided based on an objective function. The objective functionbalances the reconstruction error against a data reduction ratio.Based on class membership information of each signal obtained,this thesis proposes the {it Vertical Group-Wise Threshold}method to increase the discriminative capability of thereduced-size data so that the reduced data set retains salientdifferences between classes as much as possible. A real-lifeexample (Tonnage data) shows our proposed method is promising.