I want to model a Neural network where my input variable\parameter are 12 and output is a time series having 60 data-points . I have such 10 samples. Please suggest how to model such neural network.
Ten samples is a too small quantity for achieving good results with neural networks. I recommend you to gather more samples (you can try reproducing the process in a computer controlled simulation environment).
Usually neural networks take many inputs and few outputs. So, maybe there is a way you can concentrate the output in less variables by drawing conclusions from the 60 data points. For example, you could estimate the mean and the variance of the outputs and try making the neural network to learn these outputs instead of the 60 data points.
Good Luck!
Article Improved Shape Parameter Estimation in K Clutter with Neural...
Article A neural network approach to weibull distributed sea clutter...
Article MATE-CFAR: Ambiente de pruebas para detectores CFAR en MATLAB
First decide some specific features (n) to represent your time series output samples with 60 data points by using some time series analysis techniques. Then make your ANN learned about these features by using some supervised learning method. Then when you will test your learned ANN by giving some new input it will estimate some values about these n features. The toughest task will be to guess 60 data points from these n output features. Good Luck!!
Thanks for your responses . I have modify the problem by gather data of 108 samples. Each sample have 14 variables and the output is 77 data-points (cyclic in nature) 77 datapoint presents one cycle .
It is not feasible to considering mean and the variance of the outputs instead of the 60 data points. As after this the cyclic naute of datapoint is lost. Please suggest
Why do you need time series of 60-data points at the output? First, ANN is mostly used in problems that involve non-linear function approximation or classification of features, pattern, e.t.c. Modelling or shaking your problem into machine learning framework would help to determine whether or not ANN is the best candidate. Secondly, as suggested by other contributors, you need a significant amount of data for training, validation and testing of your neural network.