Let us split our data into training and test datasets. Out of the whole time series, we will use 80% of the data for training and the rest for testing. We need to keep the order of the time series intact while splitting it.
Split the data as follows:
train.perc = 0.8 train.indx = 1:as.integer(dim(final.data)[1] * train.perc) train.data <- final.data[train.indx,] test.data <- final.data[-train.indx ,]
train.data
contains 80% of our data and test.data
now contains the final 20%.
Let us now split the data into the dependent variable, y
, and independent variable, x
:
train.x.data <- data.matrix(train.data[,-1]) train.y.data <- train.data[,1] test.x.data <- data.matrix(test.data[,-1]) test.y.data <- test.data[,1]
We convert our x values to matrix form in both test
and train
.
Let us now build our deep learning model:
> mx.set.seed(1000) > mx.set.seed(100) > deep.model <- mx.mlp(data = train.x.data, label = train.y.data, + hidden_node...