bbn.timeseries: Time Series Prediction with Bayesian Belief Network
Description
bbn.timeseries() performs time series predictions using a Bayesian Belief Network (BBN) model based on a single prior scenario.
It generates figures illustrating how parameters change over time for all or selected nodes.
Plots for each node showing the predicted change over time.
Arguments
bbn.model
A matrix or dataframe of interactions between different model nodes.
priors1
An X by 2 array of initial changes to the system under investigation.
The first column should be a -4 to 4 (including 0) integer value for each node in the network with negative values
indicating a decrease and positive values representing an increase. 0 represents no change.
timesteps
This is the number of timesteps the model performs. Default = 5.
Note, timesteps are arbitrary and non-linear. However, something occurring in timestep 2, should occur before timestep 3.
disturbance
Default = 1.
1 creates a prolonged or press disturbance as per bbn.predict.
Essentially prior values for each manipulated node are at least maintained (if not increased through reinforcement in the model) over all timesteps.
2 shows a brief pulse disturbance, which can be useful to visualise changes as peaks and troughs in increase and decrease of nodes can propagate through the network.