I want to make a model to predict logs(i.e time series data) that I have. Normally on Python I would try to design a RNN with each log encoded as one hot vector and then make a LSTM (RNN) model to fit the data and predict the future log data. I am new to Splunk and wanted to use the already existing MLTK (Machine Learning Toolkit) to counter this problem. I tried exploring TIme Series Prediction functionality of Splunk but it always required a numerical data to fit the curve which is not the case of log data I am injecting.
Hi,
You can add your customized version of the algorithm to https://github.com/splunk/mltk-algo-contrib
Here, the official guide to doing so: https://docs.splunk.com/Documentation/MLApp/4.2.0/API/Introduction
Hope this helps.