Alteryx uses predictive analytics to identify patterns in historical and transactional data to identify risks as well as opportunities. The predictive tools are built using R (programming language), but there is no need to learn R to execute these predictive tools as they have already been packaged into easy-to-use macros that only require configuration.
Time series forecasting is the use of a statistical model to predict the future values based on past results. The key to time series forecasting is to capture the genuine patterns that occur in historical data without replicating past events that will not happen again. In order to do this the data needs to be structured in a certain way.
- Date field at a constant interval e.g. day/week/month etc
- Values recorded at these intervals
- Be aware of missing interval values
Tools and Models
TS Filler – flags a missing interval within the data set. Nulls can not be accepted by a forecasting model, so these values have to be replaced with an accepted alternative value (for example the median).
TS Plot – creates a number of different univariate time series plots to help determine the forecasting model required.
ARIMA and ETS models – the two pre-packaged time series methodologies.
TS Compare – allows the user to compare errors in each model. To minimise the error the model with the lowest error values across all measures is chosen.
TS Forecast – sets a percentage value for the larger and smaller confidence interval and the number of periods into the future to forecast.
HINT: always add browse tools after time series tools to visualise patterns, unusual observations, changes over time, and relationships between variables, as you are working through your workflow.