Can SVM, MLP etc do classification on Multivariate Time series problems?

Question

I have a multivariate time series dataset and I want to do classification with it. Basically, this is an EEG dataset which is trying to classify if the hand is moving or not.

Lets say I have 4 datasets and each dataset corresponds to a single target value. The datasets look something like this for example:

sample1.csv

423 651 214 685
125 749 515 475
325 458 153 621

sample2.csv

125 632 489 452
129 563 245 248
782 345 156 369

sample3.csv

321 456 302 312
156 325 186 456
325 356 792 625

sample4.csv

156 632 156 456
268 364 356 156
123 456 128 421

target.csv

1
2
2
1

So basically, the target.csv file holds the classification value for each sample#.csv file. We can think of it as a [3*4] matrix (eg sample1.csv) corresponds to or representing a single classification target value of '1' or '2'.

Is it possible to classify this kind of data with machine learning algorithms like SVM, MLP, Random Forest etc? If so, how can I feed the data into my model?

I mean like in a way, the machine learning algo considers the whole sample1.csv as 1 input with target value and and sample2.csv as input 2 with target value 2 and so on?

Thanks!

I have looked into problems like this but most of the time, I am finding examples done with LSTM.

Answer 1

I am bit confused about the data set. If i consider column as channels and rows as trials /samples, then you can apply discrete wavelet transform on it to get eeg bands. After that you need to find features out of it, such as time domain features and entropies. After that you can apply machine learning algorithm. If you can provide further details about your dataset, i can give you more precise answer.