Predict method for GRUAttention

# S3 method for GRUAttention
predict(model, data, top.k.attributions = NULL, explain.mode = NULL, ...)

Arguments

data

DataFrame
Data for prediction. Every row in the data should contain one piece of record data for prediction, i.e. it should be structured as follows:

  • First column: Record ID, type INTEGER.

  • Other columns: Time-series and external data values, arranged in time order.

The number of columns in data should be equal to time.dim * (M + 1) + 1, where M is the number of exogenous variables of the input data in the training phase.

top.k.attributions

integer, optional
Specifies the number of features with highest attributions to output.

  • If explain.mode is 'time-wise' then this value needs to be smaller than the length of time series data for prediction.

  • If explain.mode is 'feature-wise', then this value needs to be no greater than the number of exogenous variables.

Defaults to 0.

explain.mode

c('time-wise', 'feature-wise'), optional
Specifies the mechanism for generating the reason code for inference results.

  • 'time-wise': Use attention weights to assign time-dimension-wise contributions.

  • 'feature-wise': Use Bayesian Structural Time Series(BSTS) to assign feature-wise contributions.

Defaults to 'time-wise'.

Value

DataFrame
Predicted values, structured as follows:

  • ID: type INTEGER, timestamp.

  • VALUE: type DOUBLE, forecast values of data records.

  • REASON_CODE: type NCLOB, containing sorted SHAP values for test data at each time step/each feature component.

Details

Time-series prediction using GRU equipped with Attention Mechanism.