A Fresh Approach: Combining Methods for More Accurate Predictions
In the past, there were two distinct ways to produce time series predictions. The first used “classical algorithms” (ARIMA, Moving Average and Logistic Regressions). These functions needed a small amount of data to generate predictions, but their forecasts weren’t completely accurate.
Soon to Come: Even Better Predictions?
Dyntell’s researchers are currently working to create an even more accurate prediction model, built on new architecture which adds the correlation’s benefit to the ensemble process. Our servers collect time series data together with the Big Database and continuously process this data. If a Dyntell Bi user requests a prediction, the system will try to find a correlation between the actual business data and the Time Series Big Data. Though correlation doesn’t mean absolute cause and effect, the method operates under the premise that things are related, and that events affect each other in cascading sequences or are based on the same mathematical rules and have a similar behavior.