Day #324 - Plumbing Old 'R' Code

What I like in my Approach

• R Example with 
• Data Normalization
• Cluster Data
• Build Regression on top of Clustered Data

	#http://www3.dsi.uminho.pt/pcortez/forestfires/
	#Step 1 - Load Data and check summary details
	setwd("E:/RNotes/RData/")
	forestfires = NULL
	forestfires <- read.csv(file = "forestfires.csv")
	head(forestfires)
	summary(forestfires)
	#Step 2 - Visual Correlation between variables
	#library(psych)
	#pairs.panels(forestfires)
	#Step 3 - Normalize columns where min and max values have large differences
	#forestfires$DC <- log(forestfires$DC)
	#forestfires$DMC <- log(forestfires$DMC)
	forestfires = subset(forestfires, forestfires$area>0)
	nrow(forestfires)
	#forestfires$area <- (forestfires$area-min(forestfires$area))/(max(forestfires$area)-min(forestfires$area))
	forestfires$area <- log(forestfires$area)
	forestfires$X = (forestfires$X-min(forestfires$X))/(max(forestfires$X)-min(forestfires$X))
	forestfires$Y = (forestfires$Y-min(forestfires$Y))/(max(forestfires$Y)-min(forestfires$Y))
	forestfires$XY = forestfires$X*forestfires$Y
	forestfires$XXYY = forestfires$X*forestfires$X*forestfires$Y*forestfires$Y
	forestfires$XY = (forestfires$XY-min(forestfires$XY))/(max(forestfires$XY)-min(forestfires$XY))
	forestfires$XXYY = (forestfires$XXYY-min(forestfires$XXYY))/(max(forestfires$XXYY)-min(forestfires$XXYY))
	forestfires$FFMC = (forestfires$FFMC-min(forestfires$FFMC))/(max(forestfires$FFMC)-min(forestfires$FFMC))
	#forestfires$DC <- log(forestfires$DC)
	#forestfires$DC <- forestfires$DC
	#forestfires$DMC <- forestfires$DMC
	forestfires$DMCDC <- forestfires$DMC*forestfires$DC
	forestfires$tempwind <- forestfires$temp*forestfires$wind
	forestfires$FFMCDMCDC <- forestfires$FFMC*forestfires$DMC*forestfires$DC
	#print(unique(forestfires$month))
	#print(unique(forestfires$day))
	sample(forestfires)
	#Step 4 - Replace month and data text to numeric values as below
	forestfires$month1[forestfires$month=='jan'] = 1
	forestfires$month1[forestfires$month=='feb'] = 2
	forestfires$month1[forestfires$month=='mar'] = 3
	forestfires$month1[forestfires$month=='apr'] = 4
	forestfires$month1[forestfires$month=='may'] = 5
	forestfires$month1[forestfires$month=='jun'] = 6
	forestfires$month1[forestfires$month=='jul'] = 7
	forestfires$month1[forestfires$month=='aug'] = 8
	forestfires$month1[forestfires$month=='sep'] = 9
	forestfires$month1[forestfires$month=='oct'] = 10
	forestfires$month1[forestfires$month=='nov'] = 11
	forestfires$month1[forestfires$month=='dec'] = 12
	forestfires$day1[forestfires$day=='sun'] = 0
	forestfires$day1[forestfires$day=='mon'] = 1
	forestfires$day1[forestfires$day=='tue'] = 2
	forestfires$day1[forestfires$day=='wed'] = 3
	forestfires$day1[forestfires$day=='thu'] = 4
	forestfires$day1[forestfires$day=='fri'] = 5
	forestfires$day1[forestfires$day=='sat'] = 6
	#Step 5 - Remove original columns
	forestfires$day <- NULL
	forestfires$month <- NULL
	#Step 6 - Run K means
	library(NbClust)
	set.seed(1)
	numberofclusters <- NbClust(forestfires,min.nc=2,max.nc=15,method="kmeans")
	#Step 7 - Print results
	table(numberofclusters$Best.n[1,])
	grpForest <- kmeans( forestfires, centers=3,nstart=10)
	grpForest$cluster
	grpForest$centers
	grpForest$withinss
	grpForest$size
	forestfirescluster1 = NULL
	forestfirescluster2 = NULL
	forestfirescluster3 = NULL
	clusterresults = c(grpForest$cluster)
	length(clusterresults)
	nrow(forestfires)
	for(i in 1:nrow(forestfires))
	{
	if(clusterresults[i]==1)
	{
	print('Cluster 1')
	deldata = forestfires[i,]
	forestfirescluster1 = rbind(forestfirescluster1,deldata)
	}
	if(clusterresults[i]==2)
	{
	print('Cluster 2')
	deldata = forestfires[i,]
	forestfirescluster2 = rbind(forestfirescluster2,deldata)
	}
	if(clusterresults[i]==3)
	{
	print('Cluster 3')
	deldata = forestfires[i,]
	forestfirescluster3 = rbind(forestfirescluster3,deldata)
	}
	}
	nrow(forestfirescluster1)
	nrow(forestfirescluster2)
	nrow(forestfirescluster3)
	#df2<-forestfirescluster1[complete.cases(forestfirescluster1),]
	#forestfirescluster1 = NULL
	#forestfirescluster1 = df2
	#df2<-forestfirescluster3[complete.cases(forestfirescluster3),]
	#forestfirescluster3 = NULL
	#forestfirescluster3 = df2
	nrow(forestfirescluster1)
	nrow(forestfirescluster2)
	nrow(forestfirescluster3)
	model1 = lm(forestfirescluster1$area~forestfirescluster1$DMCDC+forestfirescluster1$tempwind +forestfirescluster1$FFMCDMCDC+forestfirescluster1$DMCDC+forestfirescluster1$XY+forestfirescluster1$XXYY+forestfirescluster1$X+forestfirescluster1$Y+forestfirescluster1$FFMC+forestfirescluster1$DMC+forestfirescluster1$DC+forestfirescluster1$ISI+forestfirescluster1$temp+forestfirescluster1$RH+forestfirescluster1$wind+forestfirescluster1$rain+factor(forestfirescluster1$month1)+factor(forestfirescluster1$day1),na.action="na.exclude")
	model1
	summary(model1)
	plot(model1$fitted.values, model1$residual.values)
	hist(model1$residuals)
	library(leaps)
	model1 = regsubsets(forestfirescluster1$area~forestfirescluster1$DMCDC+forestfirescluster1$tempwind +forestfirescluster1$FFMCDMCDC+forestfirescluster1$DMCDC+forestfirescluster1$XY+forestfirescluster1$XXYY+forestfirescluster1$X+forestfirescluster1$Y+forestfirescluster1$FFMC+forestfirescluster1$DMC+forestfirescluster1$DC+forestfirescluster1$ISI+forestfirescluster1$temp+forestfirescluster1$RH+forestfirescluster1$wind+forestfirescluster1$rain+factor(forestfirescluster1$month1)+factor(forestfirescluster1$day1),na.action="na.exclude",data=forestfirescluster1,nbest = 8)
	model1
	summary(model1)
	plot(model1$fitted.values, model1$residual.values)
	hist(model1$residuals)
	model2 = lm(forestfirescluster2$area~forestfirescluster2$DMCDC+forestfirescluster2$tempwind +forestfirescluster2$FFMCDMCDC+forestfirescluster2$DMCDC+forestfirescluster2$XY+forestfirescluster2$XXYY+forestfirescluster2$X+forestfirescluster2$Y+forestfirescluster2$FFMC+forestfirescluster2$DMC+forestfirescluster2$DC+forestfirescluster2$ISI+forestfirescluster2$temp+forestfirescluster2$RH+forestfirescluster2$wind+forestfirescluster2$rain+factor(forestfirescluster2$month1)+factor(forestfirescluster2$day1),na.action="na.exclude")
	model2
	summary(model2)
	plot(model2$fitted.values, model2$residual.values)
	hist(model2$residuals)
	model3 = lm(forestfirescluster3$area~forestfirescluster3$DMCDC+forestfirescluster3$tempwind +forestfirescluster3$FFMCDMCDC+forestfirescluster3$DMCDC+forestfirescluster3$XY+forestfirescluster3$XXYY+forestfirescluster3$X+forestfirescluster3$Y+forestfirescluster3$FFMC+forestfirescluster3$DMC+forestfirescluster3$DC+forestfirescluster3$ISI+forestfirescluster3$temp+forestfirescluster3$RH+forestfirescluster3$wind+forestfirescluster3$rain+factor(forestfirescluster3$month1)+factor(forestfirescluster3$day1),na.action="na.exclude")
	model3
	summary(model3)
	plot(model3$fitted.values, model3$residual.values)
	hist(model3$residuals)

view raw ForestFires.R hosted with ❤ by GitHub

Happy Learning!!!