Descripción de metodologías de Machine Learning (ML) para la identificación de actividades a través de reconocimiento de patrones.
Keywords:
ML, semi-supervised learning, classification, labellingAbstract
The application of Machine Learning (ML) models is becoming more and more frequent for the implementation, automation and systematization of processes. However, the models and techniques that are available in the literature and current development are designed with the aim of obtaining a better performance in a given problem, either to enhance the evaluation and classification of labelled data or to enhance the search for clusters or highly probable groups for the correct classification, where the former serves to improve the accuracy in the evaluation of the classification. since it does not care about labelling, while the second serves to improve classification, considering that the data is not labelled. Due to the advantages and disadvantages presented by the efficiency of these approaches to the use of an extensive database, hybrid models are used in order to obtain the correct classification more accurately, and in particular, the present study carried out an analysis of four ML approaches of supervised learning implemented by the application of algorithms, to closely track processes and understand each issue. The results showed a variable accuracy between 30% and 50% with respect to the zero accuracy of unsupervised models. In addition, it was concluded that the developed model was conditioned for a potential improvement with the implementation of the semi-supervised Hierarchical Extreme Learning Machine (HELM) model, which is suggested to be used as a necessary complement in classic ML models for unsupervised predictions.
References
Bannach et al., (2008). Rapid Prototyping of Activity Recognition Applications. IEEE Pervasive Computing, 1-19.
Baños et al., (2012). A benchmark dataset to evaluate sensor displacement in activity recognition. . In Proceedings of the 2012 ACM Conference on Ubiquitous Computing, 1026-1035.
Bengio. (2013). Deep learning of representations: looking forward. International Conference on Statistical Language and Speech Processing, 1-37.
Lachtermacher et al., (1995). Backpropagation in time-series forecasting. Journal of Forecasting, 14, 381-393.
Lapedes et al., (1988). How neural nets work In: Anderson (Ed), Neural Information Processing Systems. American Institute of Physics, 442-456.
Lara, O., & Labrador, M. (2012). A survey on human activity recognition using wearable sensors. IEEE communications surveys & tutorials, 15(3), 1192-1209.
Piqueras. (2020). Aplicación del aprendizaje profundo al reconocimiento de la actividad humana. Máster Universitario en Inteligencia Artificial. Valencia, España: Universidad Nacional de La Rioja (UNIR).
Srinivasan et al., (1994). A neural network short-term load forecaster . Electric Power System Research, 28, 227-234.
Wang et al., (2019). Deep learning for sensor-based activity recognition: A survey. Pattern Recognition Letters, 119, 3-11.
Yang et al., (2015). Deep convolutional neural networks on multichannel time series for human activity recognition. Twenty-Fourth International Joint Conference on Artificial Intelligence.
Yang, Q. (2009). Activity recognition: Linking low-level sensors to high-level intelligence. Twenty-First International Joint Conference on Artificial Intelligence.
Yao et al., (2018). Deep learning of semisupervised process data with hierarchical extreme learning machine and soft sensor application. IEEE Transactions on Industrial Electronics, 65(2), 1491-1498.
