For sequenceInputLayer you don't need to specify the sequence length as a feature. So you would just need numFeatures = 5.
For batches of sequence data in trainNetwork you need each observation in the batch to be a cell, this applies to the input and output sequences - this is to allow for cases where each sequence might have a different length. Additionally each cell should contain an array with size "NumFeatures x SequenceLength".
Here's one way you could do that with data like yours
batchSize = 5950;
sequenceLength = 14;
numFeatures = 5;
% generate random data for example
TrainX = randn(batchSize,sequenceLength,numFeatures);
% permute batch x sequence x features -> features x sequence x batch
TrainX = permute(TrainX,[3,2,1]);
% convert to cell
TrainX = num2cell(TrainX,[1,2]);
% flatten into column vector of cell-s
TrainX = TrainX(:);
To demonstrate, here's how you would use trainNetwork to train an LSTM that attempts to just memorise the input:
layers = [
sequenceInputLayer(numFeatures)
lstmLayer(numFeatures)
regressionLayer];
opts = trainingOptions("sgdm");
trainNetwork(TrainX,TrainX,layers,opts)
