| () ## train an neural network for health risk prediction |
| () trainCNN <- function(trainset,risks) |
| () require(mxnet); |
| () # Define the input data |
| () data <- mx.symbol.Variable(‘data’); |
| () # 1st convolutional layer |
| () conv_1 <- mx.symbol.Convolution(data = data, kernel = c(5,5), num_filter = 20); |
| () act_1 <- mx.symbol.Activation(data = conv_1, act_type = “relu”) |
| () pool_1 <- mx.symbol.Pooling(data = act_1, pool_type = “max”, kernel = c(2,2), |
| stride = c(2,2)) |
| () # 2nd convolutional layer |
| () conv_2 <- mx.symbol.Convolution(data = pool_1, kernel = c(5,5), num_filter = 50) |
| () act_2 <- mx.symbol.Activation(data = conv_2, act_type = “relu”) |
| () pool_2 <- mx.symbol.Pooling(data=act_2, pool_type = “max”, c(2,2), stride = c(2,2)) |
| () # 1st fully connected layer |
| () flatten <- mx.symbol.Flatten(data = pool_2) |
| () fc_1 <- mx.symbol.FullyConnected(data = flatten, num_hidden = 100) |
| () act_3 <- mx.symbol.Activation(data = fc_1, act_type = “relu”) |
| () # 2nd fully connected layer |
| () fc_2 <- mx.symbol.FullyConnected(data = act_3, num_hidden = 40) |
| () # Set up the symbolic model |
| () NN_model <- mx.symbol.SoftmaxOutput(data = fc_2) |
| () # Set seed for reproducibility |
| () mx.set.seed() |
| () # Train the model |
| () model <- mx.model.FeedForward.create(NN_model, X=trainset, y=risks, ctx=mx.cpu(), |
| num.round = 50, array.batch.size = 40, learning.rate = 0.01, momentum = 0.9) |
| () return(model); |
| () |
| () ## use an neural network to predict health risk |
| () predictRisk <- function(nn,dataset,risks) |
| () require(neuralnet); |
| () nn.results <- compute(nn, dataset); |
| () pr = creditnet.results$net.result; |
| () colnames(pr) <- risks; |
| () return(pr); |
| () |
| () ## use ARMA to predict health risk and plot |
| () plotARMA <- function(dataset,start=c(2011,1)) |
| () require(forecast); |
| () datat<-ts(dataset,2,start=start,frequency=12); |
| () fit <- Arima(datat, order=c(2,0,1)); |
| () plot(forecast(fit)); |
| () |
