javascript - Using neural network to predict next outcome -

i struggling understand how use neural networks predict values.

as have understood, can train neural network time series.

so if have, instance, day-to-day quotes specific stock, can train network these values.

but if have day-to-day quotes vector [42, 12, 20, 53, 18] , want predict next value, wouldn't linear regression problem formula y = 0.7x + 31.1, yielding 26.9 next (6th) value?

is neural networks useful when output relies on many inputs? can show me working example of how predict next outcome in time series shows it's more useful use neural networks rather multivariate regression?

i have, instance, looked @ synaptic.js, don't provide simple examples predict outcomes.

i new @ neural networks, sorry if question stupid :-d

you can use recurrent neural network such task in following manner:

1. input: input take sequence of stock prices without last one.
2. output: output take sequence of stock prices without first one.
3. topology , network details: build recurrent neural network (e.g. using gru or lstm units) in way given input must predict next step. example feeding [42, 12, 20, 53] input should produce [12, 20, 53, 18] output (i'm using example sequence gave).
4. prediction step: if want predict next outcome of network - feed network sequence have , take last output prediction.