A Comparative Model of Feature Engineering With and Without Domain Knowledge
Keywords:
Machine Learning, Feature Engineering, Domain Knowledge, Human Action Recognition, Neural NetworksAbstract
One of the key aspects of building a good machine learning model is Feature engineering. Feature engineering is a process where we create new features from existing raw features. To create new features, we require domain experts who have knowledge of the subject. By using their knowledge they create new features which are helpful for a machine to learn better. The time taken by the domain experts to understand the data and then create new features is time-consuming and expensive. This problem is addressed with a neural network which will not require domain experts to engineer new features. Current paper deals with the case study pertaining to the data of Human Action Recognition. Using the data, the machine predicts the various physical actions and appearances of a person like if the person is sitting, standing, walking, walking up stairs, and walking downstairs or lying. We compare the accuracy of the model using data which was feature engineered by experts and the model which was not feature engineered by the domain experts.
References
[1] B. Fish A. Khan N. Chehade C. Chien and G. Pottie, “Feature selection-based on mutual information for human activity recognition,” IEEEInternational Conference on Acoustics, Speech and Signal Processing,vol.37, pp.1729-1732, 2012.
[2] D.Anguita A. Ghio L. Oneto X. Parra and J. Reyes-Ortiz, “Human Activity Recognition on Smartphones Using a Multiclass Hardware- Friendly Support Vector Machine.” International Conference on Ambient Assisted Living and Home Care, pp.216-223, 2012.
[3] C. Schuldt, I. Laptev, and B. Caputo, “Recognizing human actions:A local SVM approach,” in Proc. Int. Conf. Pattern Recognit., vol. 3.2004, pp. 32–36.
[4] Changki Lee, Gary Geunbae Lee, “Information gain and divergence based feature selection for machine learning based textcategorization,” Information Processing & Management, vol. 42,Issue 1, pp. 155-165, January 2006.
[5] J. Yang, J. Wang and Y. Chen, “Using acceleration measurements foractivity recognition: An effective learning algorithm for constructingneural classifiers,” in Pattern Recognition Letters, vol.29, no.16, pp. 2213-2220, 2008.
[6] M. B. Holte, C. Tran, M. M. Trivedi, and T. B. Moeslund, “Human pose estimation and activity recognitionfrom multi-view videos: Comparative explorations ofrecent developments,” IEEE Journal of Selected Topicsin Signal Processing, vol. 6, pp. 538–552, 2012.
[7] K. Charalampous and A. Gasteratos, “On-line deep learningmethod for action recognition,” Pattern Anal. Appl., pp. 1–18,Aug. 2014.
[8] Q. Li J. Stankovic M. Hanson and A. Barth, “Accurate, Fast Fall Detection Using Gyroscopes and
Accelerometer-Derived Posture Information,”Sixth International Workshop on Wearable and Implantable Body SensorNetworks, pp.138-143, 2009.
[9] N. Gkalelis, N. Nikolaidis, and I. Pitas, “View independent human movement recognition from multi-view video exploiting a circular invariant posture representation,” in 2009 IEEE International Conference on Multimedia and Expo, June 2009, pp. 394–397.
[10] O. Banos J. Galvez M. Damas H. Pomares and I. Rojas, “Windowsize impact in human activity recognition,” Sensors, vol.14, no.4, pp.6474-6499, 2014.
Downloads
Published
How to Cite
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors contributing to this journal agree to publish their articles under the Creative Commons Attribution 4.0 International License, allowing third parties to share their work (copy, distribute, transmit) and to adapt it, under the condition that the authors are given credit and that in the event of reuse or distribution, the terms of this license are made clear.
