What is the best AI algorithm for time based binary prediction based on a database of previous results? With all the algorithms out there, i'm looking for advice on the best algorithm to train itself to predict the next binary number of an incomplete "real time" data set based off a database "data set."
Example data sets that were previously recorded and stored
Binary Data Set 1 100**110111**0010111000111010101
Binary Data Set 2 0110010001101110110**110111**000
Binary Data Set 3 0010110010**110111**1001011011011000
In each set we find the pattern 110111 and the next event
Set 1 - 110111 Next event is 0
Set 2 - 110111 Next event is 0
Set 3 - 110111 Next event is 1
In 2 data sets, the next event was a 0, in 1 data set the next event was a 1. So in our very tiny example...if this pattern shows up, we would make a prediction of 0 because it shows up 66% of the time. But the algorithm would train itself to make this decision.
Realtime - In real time the binary events come in 1 per second. We have the received the following incomplete data and wish to predict what is next 1000010000110111 What algorithm is best suited to predict what is next?
Thank you for any assistance!!
Recurrent neural networks and especially LSTM (Long-short-term-memory) networks are used for time series prediction. In your case, the data is quite simple and can be modelled easily. Look at this The magic of LSTM networks