load necessary packages
library(neuralnet)
library(nnet)
library(NeuralNetTools)
step 1: collecting data
step 2: exploring and preparing the data
read in data and examine structure
abalone <- read.csv("abalone.csv")
str(abalone)
## 'data.frame': 4177 obs. of 9 variables:
## $ Sex : Factor w/ 3 levels "F","I","M": 3 3 1 3 2 2 1 1 3 1 ...
## $ Length : num 0.455 0.35 0.53 0.44 0.33 0.425 0.53 0.545 0.475 0.55 ...
## $ Diameter : num 0.365 0.265 0.42 0.365 0.255 0.3 0.415 0.425 0.37 0.44 ...
## $ Height : num 0.095 0.09 0.135 0.125 0.08 0.095 0.15 0.125 0.125 0.15 ...
## $ Whole.weight. : num 0.514 0.226 0.677 0.516 0.205 ...
## $ Shucked.weight : num 0.2245 0.0995 0.2565 0.2155 0.0895 ...
## $ Viscera.weight.: num 0.101 0.0485 0.1415 0.114 0.0395 ...
## $ Shell.weight : num 0.15 0.07 0.21 0.155 0.055 0.12 0.33 0.26 0.165 0.32 ...
## $ Rings. : int 15 7 9 10 7 8 20 16 9 19 ...
Encode as a one hot vector multilabel data
train.aba <- cbind(abalone[, 2:9], class.ind(as.factor(abalone$Sex)))
Set labels name
names(train.aba) <- c(names(abalone)[2:9],"Female","Infant","Male")
Scale data
scl <- function(x){ (x - min(x))/(max(x) - min(x)) }
train.aba[, 1:8] <- data.frame(lapply(train.aba[, 1:8], scl))
head(train.aba)
step 3: training a model on the data
n <- names(train.aba)
f <- as.formula(paste("Female + Infant + Male ~", paste(n[!n %in% c("Female","Infant","Male")], collapse = " + ")))
f
## Female + Infant + Male ~ Length + Diameter + Height + Whole.weight. +
## Shucked.weight + Viscera.weight. + Shell.weight + Rings.
nn <- neuralnet(f,
data = train.aba,
hidden = c(4,2),
act.fct = "logistic",
linear.output = FALSE,
lifesign = "minimal", threshold = 0.1)
## hidden: 4, 2 thresh: 0.1 rep: 1/1 steps: 66850 error: 1045.98411 time: 5.85 mins
summary(nn)
## Length Class Mode
## call 8 -none- call
## response 12531 -none- numeric
## covariate 33416 -none- numeric
## model.list 2 -none- list
## err.fct 1 -none- function
## act.fct 1 -none- function
## linear.output 1 -none- logical
## data 11 data.frame list
## net.result 1 -none- list
## weights 1 -none- list
## startweights 1 -none- list
## generalized.weights 1 -none- list
## result.matrix 58 -none- numeric
Appendix
#### R code ####
# load necessary packages
library(neuralnet)
library(nnet)
library(NeuralNetTools)
##### step 1: collecting data
##### step 2: exploring and preparing the data
# read in data and examine structure
abalone <- read.csv("abalone.csv")
str(abalone)
# Encode as a one hot vector multilabel data
train.aba <- cbind(abalone[, 2:9], class.ind(as.factor(abalone$Sex)))
# Set labels name
names(train.aba) <- c(names(abalone)[2:9],"Female","Infant","Male")
# Scale data
scl <- function(x){ (x - min(x))/(max(x) - min(x)) }
train.aba[, 1:8] <- data.frame(lapply(train.aba[, 1:8], scl))
head(train.aba)
##### step 3: training a model on the data
n <- names(train.aba)
f <- as.formula(paste("Female + Infant + Male ~", paste(n[!n %in% c("Female","Infant","Male")], collapse = " + ")))
f
nn <- neuralnet(f,
data = train.aba,
hidden = c(4,2),
act.fct = "logistic",
linear.output = FALSE,
lifesign = "minimal", threshold = 0.1)
summary(nn)
##### step 4: evaluating model performance
# visualize the network topology
plot(nn)
# plotnet
par(mar = numeric(4), family = 'serif')
plotnet(nn, alpha = 0.6)
# Compute predictions
predicted.nn <- compute(nn, train.aba[, 1:8])
# Extract results
result.predicted.nn <- predicted.nn$net.result
head(result.predicted.nn)
# Accuracy (training set)
original_values <- max.col(train.aba[, 9:11])
result.predicted.nn_2 <- max.col(result.predicted.nn)
mean(result.predicted.nn_2 == original_values)
##### step 5: improving model performance
# Crossvalidate
set.seed(10)
k <- 10
outs <- NULL
proportion <- 0.995
library(plyr)
pbar <- create_progress_bar('text')
pbar$init(k)
for(i in 1:k)
{
index <- sample(1:nrow(train.aba), round(proportion*nrow(train.aba)))
cross.train <- train.aba[index, ]
cross.test <- train.aba[-index, ]
cross.nn <- neuralnet(f,
data = cross.train,
hidden = c(4,2),
act.fct = "logistic",
linear.output = FALSE, threshold = 0.08)
# Compute predictions
predicted.nn <- compute(cross.nn, cross.test[, 1:8])
# Extract results
result.predicted.nn <- predicted.nn$net.result
# Accuracy (test set)
original_values <- max.col(cross.test[, 9:11])
result.predicted.nn_2 <- max.col(result.predicted.nn)
outs[i] <- mean(result.predicted.nn_2 == original_values)
pbar$step()
}
# Average accuracy
mean(outs)
#### R code ####
