R Notebook

#### Poker Data - Nueral Network Analysis --------------
##### Step 1: Collecting data -------------------
# This dataset is included on the UCI Machine Learning Repository website - http://archive.ics.uci.edu/ml/datasets/Poker+Hand 
# Each record of the dataset includes a poker hand with 5 playing cards.  The playing cards for each hand are   drawn from a standard deck of 52. 
## Step 2: Exploring and preparing the data ----
# read in data and examine structure
Poker <- read.csv("pokerdata.csv")
str(Poker)

'data.frame':   1025012 obs. of  11 variables:
 $ S1   : int  1 2 3 4 4 1 1 2 3 4 ...
 $ C1   : int  10 11 12 10 1 2 9 1 5 1 ...
 $ S2   : int  1 2 3 4 4 1 1 2 3 4 ...
 $ C2   : int  11 13 11 11 13 4 12 2 6 4 ...
 $ S3   : int  1 2 3 4 4 1 1 2 3 4 ...
 $ C3   : int  13 10 13 1 12 5 10 3 9 2 ...
 $ S4   : int  1 2 3 4 4 1 1 2 3 4 ...
 $ C4   : int  12 12 10 13 11 3 11 4 7 3 ...
 $ S5   : int  1 2 3 4 4 1 1 2 3 4 ...
 $ C5   : int  1 1 1 12 10 6 13 5 8 5 ...
 $ CLASS: int  9 9 9 9 9 8 8 8 8 8 ...

# custom normalization function
normalize <- function(x) { 
  return((x - min(x)) / (max(x) - min(x)))
}
# apply normalization to entire data frame
Poker_norm <- as.data.frame(lapply(Poker, normalize))
# confirm that the range is now between zero and one
summary(Poker_norm$S1)

     Min.   1st Qu.    Median      Mean   3rd Qu.      Max. 
0.0000000 0.3333333 0.6666667 0.5002309 0.6666667 1.0000000 

summary(Poker_norm$CLASS)

      Min.    1st Qu.     Median       Mean    3rd Qu.       Max. 
0.00000000 0.00000000 0.00000000 0.06855638 0.11111110 1.00000000 

# compared to the original minimum and maximum
summary(Poker$S1)

    Min.  1st Qu.   Median     Mean  3rd Qu.     Max. 
1.000000 2.000000 3.000000 2.500693 3.000000 4.000000 

summary(Poker$CLASS)

     Min.   1st Qu.    Median      Mean   3rd Qu.      Max. 
0.0000000 0.0000000 0.0000000 0.6170074 1.0000000 9.0000000 

# create training and test data
poker_train <- Poker_norm[1:7512, ]
poker_test <- Poker_norm[7513:10013, ]
## Step 3: Training a model on the data ----
# train the neuralnet model
library(neuralnet)
# simple ANN with only a single hidden neuron
## This will test how the different cards affect the outcome of the poker hand (CLASS) - Modeling the Strength of the Poker hand.
poker_model <- neuralnet(CLASS ~ S1 +
                              C1 + S2 + C2 + 
                              S3 + C3 + S4 + C4 + S5 + C5,
                            data = poker_train)
# visualize the network topology
plot(poker_model)
## Step 4: Evaluating model performance ----
# obtain model results
model_results <- compute(poker_model, poker_test[1:10])
# obtain predicted strength values
predicted_CLASS <- model_results$net.result
# examine the correlation between predicted and actual values
cor(predicted_CLASS, poker_test$CLASS)  

              [,1]
[1,] 0.04677776188

## Step 5: Improving model performance ----
# a more complex neural network topology with 4 hidden neurons
# to guarantee repeatable results
#hidden = 4
#set.seed(12345) 
#poker_model2 <- neuralnet(CLASS ~ S1 +
#                               C1 + S2 + C2 + 
#                               S3 + C3 + S4 + C4 + S5 + C5,
#                            data = poker_train, hidden = 4, stepmax=1e6)
#poker_model2
#plot(poker_model2)
#Error value too large
#hidden = 10
#set.seed(12345) 
#poker_model3 <- neuralnet(CLASS ~ S1 +
#                            C1 + S2 + C2 + 
#                            S3 + C3 + S4 + C4 + S5 + C5,
#                          data = poker_train, hidden = 10, stepmax=1e6)
#poker_model3
#Code failed to run generating an error output
#hidden = 6
set.seed(12345) 
poker_model5 <- neuralnet(CLASS ~ S1 +
                            C1 + S2 + C2 + 
                            S3 + C3 + S4 + C4 + S5 + C5,
                          data = poker_train, hidden = 6, stepmax=1e6)
plot(poker_model5)

# evaluate the results as we did before
model_results2 <- compute(poker_model5, poker_test[1:10])
predicted_class2 <- model_results2$net.result
cor(predicted_class2, poker_test$CLASS)  

             [,1]
[1,] 0.2800660736