Regression Example with Keras in R

   Regression data can be easily fitted with a Keras Deep Learning API. In this tutorial, we'll briefly learn how to fit and predict regression data by using the Keras neural networks model in R. Here, we'll see how to create simple regression data, build the model, train it, and finally predict the input data. The tutorial covers:

• Generating sample dataset
• Building the model
• Training the model and checking the accuracy
• Predicting test data 
• Source code listing
  

We'll start by loading the Keras library for R.

library(keras)

Generating sample dataset

First, we'll create sample regression dataset for this tutorial.

set.seed(123)

N = 450

n = seq(1:N)
a = n/10+4*sin(n/10)+sample(-1:6,N,replace=T)+rnorm(N)
b = n/8+4*sin(n/10)+sample(-3:3,N,replace=T)+rnorm(N) 
c = n/6+4*sin(n/10)+sample(-5:1,N,replace=T)+rnorm(N)
y = (a+b+c)/3+rnorm(N)
 
plot(n, c, col="orange", pch=20, cex=.9)
points(n, a, col="blue", pch=20, cex=.9)
points(n, b, col="green", pch=20, cex=.9)
points(n, y, col="red", type = "l",lwd=2)
 

Red-line is y output, and the remaining dots are series of x input.
We need to convert x input data into matrix type.

x = as.matrix(data.frame(a,b,c))
y = as.matrix(y)
head(x)
            a          b          c
[1,] 1.559084 -3.4897805 -0.6143715
[2,] 5.290081 -2.7466366  1.3203803
[3,] 2.764863 -1.1433255 -3.1107472
[4,] 8.842324  0.6967253  1.1809927
[5,] 7.402110  0.6273192 -1.7591250
[6,] 3.813864  2.3840024  3.4351513

Building the model

Next, we'll create a keras sequential model.

model = keras_model_sequential() %>% layer_dense(units=64, activation="relu", input_shape=3) %>% layer_dense(units=32, activation = "relu") %>% layer_dense(units=1, activation="linear") model %>% compile( loss = "mse", optimizer = "adam", metrics = list("mean_absolute_error") ) model %>% summary() _________________________________________________________________________________ Layer (type) Output Shape Param # ================================================================================= dense_166 (Dense) (None, 64) 256 _________________________________________________________________________________ dense_167 (Dense) (None, 32) 2080 _________________________________________________________________________________ dense_168 (Dense) (None, 1) 33 ================================================================================= Total params: 2,369 Trainable params: 2,369 Non-trainable params: 0 _________________________________________________________________________________

Training the model and checking the accuracy

Next, we'll fit the model with x, y data and check the accuracy.

model %>% fit(x, y, epochs = 100,verbose = 0)
 
scores = model %>% evaluate(x, y, verbose = 0)
print(scores)
$loss
[1] 0.9134331

$mean_absolute_error
[1] 0.7495633

Next, we'll predict x data and compare it with original y value in a plot.

y_pred = model %>% predict(x)

x_axes = seq(1:length(y_pred))
plot(x_axes, y, type="l", col="red")
lines(x_axes, y_pred, col="blue")
legend("topleft", legend=c("y-original", "y-predicted"),
        col=c("red", "blue"), lty=1,cex=0.8)

Predicting test data

Next, we'll split the dataset into the train and test parts, train the model again, predict test data.

train_x = x[1:400,]
test_x = x[401:N,]
 
train_y = y[1:400]
test_y = y[401:N]
 
model %>% fit(train_x,train_y,epochs = 100,verbose = 0)
y_pred = model %>% predict(test_x)

Finally, we'll plot original test data y value and predicted one.

x_axes = seq(1:length(y_pred))
plot(x_axes, test_y, col="red", type="l")
lines(x_axes, y_pred, col="blue")
legend("topleft", legend=c("y-original", "y-predicted"),
        col=c("red", "blue"), lty=1,cex=0.8)

 

   In this tutorial, we've briefly learned how to fit regression data with keras neural networks model in R. The full source code is listed below.

Source code listing 

library(keras)

set.seed(123)

N = 450 

n = seq(1:N)
a = n/10+4*sin(n/10)+sample(-1:6,N,replace=T)+rnorm(N)
b = n/8+4*sin(n/10)+sample(-3:3,N,replace=T)+rnorm(N) 
c = n/6+4*sin(n/10)+sample(-5:1,N,replace=T)+rnorm(N)
y = (a+b+c)/3+rnorm(N)
 
plot(n, c, col="orange", pch=20, cex=.9)
points(n, a, col="blue", pch=20, cex=.9)
points(n, b, col="green", pch=20, cex=.9)
points(n, y, col="red", type = "l",lwd=2)

x = as.matrix(data.frame(a,b,c))
y = as.matrix(y)
head(x)
 

model = keras_model_sequential() %>% 
   layer_dense(units=64, activation="relu", input_shape=3) %>% 
   layer_dense(units=32, activation = "relu") %>% 
   layer_dense(units=1, activation="linear")
 
model %>% compile(
   loss = "mse",
   optimizer =  "adam", 
   metrics = list("mean_absolute_error")
 )
 
model %>% summary()

model %>% fit(x, y, epochs = 100,verbose = 0)
scores = model %>% evaluate(x, y, verbose = 0)
print(scores)

y_pred = model %>% predict(x)

x_axes = seq(1:length(y_pred))
plot(x_axes, y, type="l", col="red")
lines(x_axes, y_pred, col="blue")
legend("topleft", legend=c("y-original", "y-predicted"),
        col=c("red", "blue"), lty=1,cex=0.8)

# split into train and test parts 

train_x = x[1:400,]
test_x = x[401:N,]
 
train_y = y[1:400]
test_y = y[401:N]
 
model %>% fit(train_x,train_y,epochs = 100,verbose = 0)
y_pred = model %>% predict(test_x)

x_axes = seq(1:length(y_pred))
plot(x_axes, test_y, col="red", type="l")
lines(x_axes, y_pred, col="blue")
legend("topleft", legend=c("y-original", "y-predicted"),
        col=c("red", "blue"), lty=1,cex=0.8)

References:

1. https://keras.rstudio.com/index.html
2. https://keras.rstudio.com/reference/keras_model_sequential.html