multivariate time series classification using Convolutional Neural Networks

3 visualizaciones (últimos 30 días)
The research data belongs to 160 observation (50 experienced the event and 110 not experienced the event) during time period 2001-2012. Event is described by a binary variable (0=lack of event, 1=event). For example, observation A is not experienced the event during 2001-2012, but observation B is experienced the event at 2003. There are about 150 features for each observation during each year, so some of these features should be chosen and then the observation’s event condition should be predicted.
Observation Year Event Feature 1 Feature 2 … … Feature 149 Feature 150
A 2001 0
A 2002 0
A 2003 0
A 2004 0
A 2005 0
A 2006 0
A 2007 0
A 2008 0
A 2009 0
A 2010 0
A 2011 0
A 2012 0
B 2001 0
B 2002 0
B 2003 1
I should mention that the features are real numbers. I think my problem is similar to the following article (multivariate time series classification):
"Time Series Classification Using Multi-Channels Deep Convolutional Neural Networks" by Zheng et. al.,
staff.ustc.edu.cn/~cheneh/.../Yi-Zheng-WAIM2014.pdf
especially the second data set which is used in the article except that the length of time series in the article are equal or the same. Am I right? If so, I would appreciate if you could let me have access to some materials that illustrates how to to do it using Matlab software. The following article is also similar to my problem but it intends to do image classification, so I should represent my features as multi-channel 1D signal. How to do it? "Deep Convolutional Neural Networks On Multichannel Time Series For Human Activity Recognition"
Thanks beforehand
  2 comentarios
Greg Heath
Greg Heath el 13 de En. de 2016
I do not understand. What is a typical input column vector and the corresponding target column vector?
If you are just using patternnet, the target vectors for 2 classes must be either [ 1;0] or [ 0;1]
Why do you need a deep net?
arash ebrahimi
arash ebrahimi el 13 de En. de 2016
Dear friend There is multiple features. Whats more, for observation A at 2012 the output class is [0,1], but the inputs are multiple features during 2000-2011, so it is multiple time series with different length.

Iniciar sesión para comentar.

Respuestas (0)

Categorías

Más información sobre Deep Learning Toolbox en Help Center y File Exchange.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by