BoostedDecisionTreeRegression - Per hour data

April 26, 2017
Issue in predicting System Parameters Utilization
I am trying to predict CPU usage on the basis of history data. I am using BoostedDecisionTreeRegression algorithm. But there is vast difference between actual value and predicted value. Below are the parameters on which we are training the model: Year Month Day Weekday Hour Average CPU_USAGE We are using Weekday and Hour as "Categorical Columns" Please suggest for optimizing the output. Note: As per my analysis the difference might be due to large spikes in data set values