MLR Section 31 Artificial Neural Networks

df <-  read.csv("G:\\RStudio\\udemy\\ml\\Machine Learning AZ\\Part 8 - Deep Learning\\Section 39 - Artificial Neural Networks (ANN)\\Artificial_Neural_Networks\\Churn_Modelling.csv")
head(df)

Goal is to predict the ‘exited’ (dependent variable) 1 = exited, 0= stayed
The items not needed for the analysis are taken out. Things like the rownumber, customerID, surname.

Select the fields that we will be working with

# we include the dependent variable. 
df <- df[,4:14]
head(df)

Convert some of the independent variables as numeric factors

df$Geography = as.numeric(factor(df$Geography, levels = c("France","Spain","Germany"), labels = c(1,2,3)))
df$Gender = as.numeric(factor(df$Gender, levels = c("Female", "Male"), labels = c(1,2)))
df$Gender
df$Geography

Split dataset into training and test set (300 training, 100 test)

library(caTools)

package <U+393C><U+3E31>caTools<U+393C><U+3E32> was built under R version 3.3.3

set.seed(1234)
split <- sample.split(df$Exited, SplitRatio = 0.80)
training_set <- subset(df, split == TRUE)
test_set <- subset(df, split == FALSE)

For Neural Network, it is mandatory to do feature scaling (normalization)

# Feature Scaling 1 age, 2 is salary
training_set[,-11] <-  scale(training_set[,-11])
test_set[,-11] <-  scale(test_set[,-11])

Fitting ANN to the Training Set

# Create the classifier here
# install.packages("h2o")
library(h2o)
# establish the connection
h2o.init(nthreads = -1)

 Connection successful!

R is connected to the H2O cluster: 
    H2O cluster uptime:         53 minutes 21 seconds 
    H2O cluster version:        3.10.4.6 
    H2O cluster version age:    1 month and 21 days  
    H2O cluster name:           H2O_started_from_R_wilsonpc_fmq912 
    H2O cluster total nodes:    1 
    H2O cluster total memory:   0.66 GB 
    H2O cluster total cores:    8 
    H2O cluster allowed cores:  8 
    H2O cluster healthy:        TRUE 
    H2O Connection ip:          localhost 
    H2O Connection port:        54321 
    H2O Connection proxy:       NA 
    H2O Internal Security:      FALSE 
    R Version:                  R version 3.3.2 (2016-10-31) 

classifier <- h2o.deeplearning(y = "Exited" , training_frame = as.h2o(training_set),
                               activation = "Rectifier",
                               hidden = c(6,6),
                               epochs = 100,
                               train_samples_per_iteration = -2)

  |                                                                                                            
  |                                                                                                      |   0%
  |                                                                                                            
  |======================================================================================================| 100%

  |                                                                                                            
  |                                                                                                      |   0%
  |                                                                                                            
  |===================================================                                                   |  50%
  |                                                                                                            
  |======================================================================================================| 100%

summary(classifier)

Model Details:
==============

H2ORegressionModel: deeplearning
Model Key:  DeepLearning_model_R_1497675052674_3 
Status of Neuron Layers: predicting Exited, regression, gaussian distribution, Quadratic loss, 115 weights/biases, 5.6 KB, 800,000 training samples, mini-batch size 1
  layer units      type dropout       l1       l2 mean_rate rate_rms momentum mean_weight weight_rms mean_bias
1     1    10     Input  0.00 %                                                                               
2     2     6 Rectifier  0.00 % 0.000000 0.000000  0.000735 0.000360 0.000000    0.109288   0.397864  0.048861
3     3     6 Rectifier  0.00 % 0.000000 0.000000  0.001545 0.004108 0.000000   -0.117767   0.493791  0.609235
4     4     1    Linear         0.000000 0.000000  0.000328 0.000126 0.000000   -0.351699   0.801255  0.379231
  bias_rms
1         
2 0.493353
3 1.348580
4 0.000000

H2ORegressionMetrics: deeplearning
** Reported on training data. **
** Metrics reported on full training frame **

MSE:  0.1044161
RMSE:  0.3231348
MAE:  0.2002827
RMSLE:  0.2263449
Mean Residual Deviance :  0.1044161





Scoring History: 
            timestamp   duration training_speed    epochs iterations       samples training_rmse
1 2017-06-17 13:44:17  0.000 sec                  0.00000          0      0.000000              
2 2017-06-17 13:44:17  0.686 sec 141342 obs/sec  10.00000          1  80000.000000       0.32638
3 2017-06-17 13:44:20  2.986 sec 280898 obs/sec 100.00000         10 800000.000000       0.32313
  training_deviance training_mae
1                               
2           0.10653      0.21030
3           0.10442      0.20028

Predicting the test set results

y_pred <-  h2o.predict(classifier, newdata = as.h2o(test_set[-11]))

  |                                                                                                            
  |                                                                                                      |   0%
  |                                                                                                            
  |======================================================================================================| 100%

  |                                                                                                            
  |                                                                                                      |   0%
  |                                                                                                            
  |======================================================================================================| 100%

# convert into 1 or 0 depending on a threshold value
y_pred <-  y_pred>0.5
# convert h2o object back into a vector
y_pred <-  as.vector(y_pred)

Evaluate the prediction using confusion Matrix.

# Making the confusion matrix
# [3] refers to the outcome
cm <- table(test_set[,11], y_pred)
cm

   y_pred
       0    1
  0 1529   64
  1  211  196

# shutdown the h2o
h2o.shutdown()

[1] TRUE