Classifiers Optimization Using Swarm Algorithms Faculty of Computers and Information, Minia University and SRGE member Moataz Kilany http://www.egyptscience.net Bio-inspiring and evolutionary computation: Trends, applications and open issues workshop, 7 Nov. 2015 Faculty of Computers and Information, Cairo University
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Classifiers Optimization Using Swarm Algorithms
Faculty of Computers and Information, Minia University and SRGE member
Moataz Kilany
http://www.egyptscience.net
Bio-inspiring and evolutionary computation: Trends, applications and open issues workshop, 7 Nov. 2015 Faculty of Computers and Information, Cairo University
SRGE Workshop, Cairo University (07-November-2015)
Optimizing accelerometer-based data classification using WWO
13
Human Activity Data Set.Basic 7 Human Activities.
1) Working at Computer2) Standing Up, Walking and Going up\down stairs3) Standing4) Walking5) Going Up\Down Stairs6) Walking and Talking with Someone7) Talking while Standing
SRGE Workshop, Cairo University (07-November-2015)
Optimizing accelerometer-based data classification using WWO
14
Optimization Setting.Tuning penalty (c) and gamma
parameters of SVM Kernel function.Resampling strategy
Averaging windows of accelerometer readings of length (4 / 8 / 16 / 32 seconds).
SRGE Workshop, Cairo University (07-November-2015)
Optimizing accelerometer-based data classification using WWO
15
Primary results of classification process. Running SVM with gamma parameter
randomly initialized to small numbers resulted in accuracies in 70 – 80% for 15 folds of validation.
Windows of 4-sec average samples.
SRGE Workshop, Cairo University (07-November-2015)
Optimizing accelerometer-based data classification using WWO
16
Primary results of classification process. Sampling window of 4-seconds.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 150%
20%
40%
60%
80% 72.41%
Fold ID
Accu
racy
SRGE Workshop, Cairo University (07-November-2015)
Optimizing accelerometer-based data classification using WWO
17
1 2 3 4 5 6 7 8 9 10 11 12 13 14 150%20%40%60%80%
100%120%
4 -sec4-sec Trend8-sec12-sec32-sec32-sec Trend
Fold ID
Accu
racy
Primary results of classification process. Different Sampling Windows.
SRGE Workshop, Cairo University (07-November-2015)
Optimizing accelerometer-based data classification using WWO
18
Results upon optimizing Gamma and Penalty (C) Parameters.Gamma optimized in range [1,100].Penalty (C) optimized in range
Resulted in accuracies in 99 – 100% for 15 folds of validation.
SRGE Workshop, Cairo University (07-November-2015)
Optimizing accelerometer-based data classification using WWO