0.25
0.5
0.75
1.25
1.5
1.75
2
Weka, MOA and Experiment Databases: Frameworks for Machine Learning
Published on Mar 27, 20143375 Views
This talk is in three parts. The first deals with an aspect of the Weka project that has received little attention, namely the use of machine learning in agricultural applications. I will outline our
Related categories
Chapter list
Fifth Asian Conference on Machine Learning00:00
Acknowledgements01:20
Part 1: Weka Developments02:52
Early Application Development Experiences06:31
Testing Laboratory Data08:23
Near Infrared Spectroscopy11:57
Advanced DAta Mining System (ADAMS), Motivation13:41
What is ADAMS?16:35
ADAMS: Experiment - sample type checker18:23
ADAMS: Deploy sample type checker20:26
ADAMS: Deployment features21:12
Conclusion - 122:12
Part 2: Massive Online Analysis (MOA)24:02
What is MOA?25:00
Why is it needed?26:39
Classification Setting27:29
Classifiers and Prediction Strategies29:43
Example: Real dataset31:38
Magic Classifier32:44
Comparison from the literature33:05
Magic Classifier revealed33:26
New evaluation measure35:10
Temporally Augmented Classifier35:42
SWT improves base classifiers36:06
State-of-the-art in data stream classification36:43
Methods used for comparison37:14
Experimental Results37:38
State-of-the-art: Recommendations38:41
Conclusion - 239:43
Part 3: Machine learning experimentation in practice40:17
Benefits of sharing for machine learning41:35
Collaborative Experimentation42:21
ExpML: A markup language for DM experiments43:02
Experiment database43:06
Learn: Gaining insights from querying the database43:26
Learning from the past (model-level) - 143:55
Learning from the past45:52
Learning from the past (model-level) - 247:04
Learning from the past (model-level) - 348:00
Learning from the past (data-level) - 149:09
Learning from the past (data-level) - 249:44
Learning from the past (data-level) - 351:01
Learning from the past (method-level)52:14
Conclusion - 353:47
References55:00