# install.packages("package_name")
# library(package_name)AI Learning–> Sep.2024
1-Using Artificial Intelligence(AI) as a general tools
{10.202.10.202 , 10.202.10.102 } —> Change DNS
.
.
.
2-Using Artificial Intelligence(AI) as Machine Learning (ML), (predicting weather events)
ANN, SVR, Deep Learning( for prediction, simulation, …. Exp: Horological Forecasting)
Need some software –> R or Python, …
.
.
2-1-Install R
Windows:
Go to the official R website (https://www.r-project.org/).
Click on “Download R for Windows”.
Click on “base” and download the latest version of R.
Run the installer and follow the instructions.
Linux-Ubuntu ****************************************************************
Open the terminal.
Run the following commands:
|
************************************************************************
2-1-1- Install RStudio (Optional but Recommended)
RStudio is an Integrated Development Environment (IDE) for R.
Go to the RStudio website (https://posit.co/download/rstudio-desktop/)
Download and install the version appropriate for your operating system.
2-1-2- Install R Packages
Once R (and optionally RStudio) is installed, you can install the required packages. To do this:
Open R or RStudio.
In the console, run the following command to install the desired package:
Learn more about R
2-2-Example1: Optimization problem
[Using the optimization algorithm to estimate the regression coefficients]
************E-1-1-Example in Excel************
# ---------E-1-2- Example in R ---------
library(GA)
library(openxlsx)
setwd("C:/Users/WRI/Documents/Workshop.Sep.2024")
mydata <- read.xlsx(xlsxFile ='SampleData.xlsx',sheet = "Sheet1" )
n1 <- nrow(mydata)
x1 <- mydata[,1]
y1 <- mydata[,2]
MyFunction <- function(x) {
a <- x[1]
b <- x[2]
y1.sim <- vector(mode = "numeric")
deff <- vector(mode = "numeric")
for (i in 1:n1) {
y1.sim[i] <- a*x1[i]+b
deff[i] <- (y1.sim[i]-y1[i])^2
}
objF <- sum(deff)
write.csv(y1.sim,file = "y1.sim.csv")
return(objF)
}
# minimize
optmodel <- ga(type = "real-valued", fitness = function(x) -MyFunction(x),
lower =c(-20,-50) , upper =c(20,50),
popSize = 20, maxiter = 50,pcrossover = 0.8,pmutation = 0.15,parallel =FALSE)
plot(optmodel)
a <- optmodel@solution[1]
b <- optmodel@solution[2]
a <- round(a,digits = 2)
b <- round(b,digits = 2)
#------
y1.sim <- vector(mode = "numeric")
deff <- vector(mode = "numeric")
for (i in 1:n1) {
y1.sim[i] <- a*x1[i]+b
deff[i] <- (y1.sim[i]-y1[i])^2
}
xx <- 1:n1
plot(xx, y1,
type = "l",
col = "blue",
lwd = 2,
main = paste0("Optimization Example Y=",a,"X+",b),
xlab = "X Data",
ylab = "Y Data")
lines(xx, y1.sim,
col = "red",
lwd = 2,
lty = 2)
legend("topleft",
legend = c("y1", "y1.sim"),
col = c("blue", "red"),
lty = c(1, 2),
lwd = 2) 
https://dehban.shinyapps.io/workshopsep-1/
Example2: ANN problem:
rm(list = ls())
library(neuralnet)
library(openxlsx)
library(ggplot2)
setwd("C:/Users/WRI/Documents/Workshop.Sep.2024")
mydata2 <- read.xlsx(xlsxFile ='SampleDataANN.xlsx',sheet = "Sheet1" )
set.seed(123)
normalize <- function(x) {
return ((x - min(x)) / (max(x) - min(x)))
}
mydata2_norm <- as.data.frame(lapply(mydata2, normalize))
set.seed(42)
# Split data into training and testing sets
n <- nrow(mydata2_norm)
train_index <- sample(1:n, 0.8*n)
train_data <- mydata2_norm[train_index, ]
test_data <- mydata2_norm[-train_index, ]
ANN_Model <- neuralnet(target ~ ., data = train_data, hidden = c(4, 2),stepmax = 10e6)
#------ Evaluation on Test data -------
sim.test <- predict(object = ANN_Model,newdata =test_data )
real.sim.test <- sim.test * (max(mydata2$target) - min(mydata2$target)) + min(mydata2$target)
real.obs.test <- test_data$target * (max(mydata2$target) - min(mydata2$target)) + min(mydata2$target)
cor.test<- cor(real.sim.test,real.obs.test)
#------ Evaluation on Trian data -------
sim.train <- predict(object = ANN_Model,newdata =train_data )
real.sim.train <- sim.train * (max(mydata2$target) - min(mydata2$target)) + min(mydata2$target)
real.obs.train <- train_data$target * (max(mydata2$target) - min(mydata2$target)) + min(mydata2$target)
cor.train<- cor(real.sim.train,real.obs.train)
#------ Evaluation on All data -------
sim.all <- predict(object = ANN_Model,newdata =mydata2_norm )
real.sim.all <- sim.all * (max(mydata2$target) - min(mydata2$target)) + min(mydata2$target)
real.obs.all <- mydata2_norm$target * (max(mydata2$target) - min(mydata2$target)) + min(mydata2$target)
cor.all<- cor(real.sim.all,real.obs.all)
cat("Correalation_Test:", cor.test, "\n")
cat("Correalation_Test:", cor.train, "\n")
cat("Correalation_All:", cor.all, "\n")
# save result.....
#---------------- Plot Results-------
plot(ANN_Model, show.weights = FALSE)
xx <- 1:n
# png(filename = "neural_net_model.png",width = 800,height = 600)
plot(xx, real.obs.all,
type = "l",
col = "black",
lwd = 2,
main = "ANN Problem",
xlab = "",
ylab = "Target")
lines(xx, real.sim.all,
col = "red",
lwd = 2)
legend("topleft",
legend = c("obs", "sim"),
col = c("black", "red"),
lwd = 2)
# dev.off()
#----- Save model as .RData or another format -----
save(ANN_Model, file = "neural_net_model.RData")
load("neural_net_model.RData")
https://dehban.shinyapps.io/R_MLP_ANN/
.
2-4- Development of artificial intelligence models in R (Introduction of stream-flow forecasting project~ IDIO)
Deep Learning ~ LSTM Model
