pendigits_code.Rmd

---
title: "Data Mining Final"
output: github_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

## GitHub Documents

This is an R Markdown format used for publishing markdown documents to GitHub. When you click the **Knit** button all R code chunks are run and a markdown file (.md) suitable for publishing to GitHub is generated.

## Including Code

You can include R code in the document as follows:

```{r cars}
data.dir = "C://Users//Arti Patel//Desktop//data_mining//final"
digits = read.csv(file.path(data.dir, "dataset_32_pendigits.csv"))
set.seed(524)
x = digits[,1:16]
library(dplyr)
library(dendextend)
digits_p1 <- filter(digits, class < 5)
x1 = digits_p1[,1:16]
digits_p2 <- filter(digits, class > 4)
x2 = digits_p2[,1:16]

# pca.comp = prcomp(x, scale=TRUE)
# summary(pca.comp)
# pc.9 = pca.comp$x[, 1:9]  # first 9 PC = 95% variability

#initialize df for different kinds of predictions
predicts = data.frame(digits$class)

#kmeans--------------------------------
#using pca 9 dim
km_pc.9 = kmeans(pc.9, centers=10, nstart=50)
predicts$pc.9.predict = km_pc.9$cluster
table(predicts$digits.class,predicts$pc.9.predict)

#no pca
km_all = kmeans(x, centers=10, nstart=50)
predicts$allpredict = km_all$cluster
table(predicts$digits.class,predicts$allpredict)

#--------------------------------------
#hierarchical clustering

digit.dist = dist(scale(x),method="maximum")
hc = hclust(digit.dist, method="complete") # new hclust
#complete linkage dendrogram for all 10 digits
plot(hc, las=1,cex=.6)
#height vs k plot to determine k
plot(length(hc$height):1, hc$height, type='o', xlab="K", ylab="height", las=1, 
     xlim=c(1, 20),ylim=c(3, 7))
points(10, hc$height[n-9], col="red", pch=19) # K=10

# dend <- as.dendrogram(hc)
# dend <- color_labels(dend, k = 10)
# plot(dend)

#breaking into 2 sets: digits 0-4
p1.dist = dist(scale(x1))
hc1 = hclust(p1.dist, method="complete") # new hclust
#n = length(hc$height)     # get number of merges
#plot(hc1, las=1,cex=.6)

dend <- as.dendrogram(hc1)
dend <- rotate(dend, 1:150)
dend <- color_branches(dend, k=5)

# labels_colors(dend) <-
#   rainbow_hcl(5)[sort_levels_values(as.numeric(digist_p1[,16])[order.dendrogram(dend)]
#   )]
# dend <- hang.dendrogram(dend,hang_height=0.1)
# dend <- set(dend, "labels_cex", 0.5)
# plot(dend,
#      main = "Clustering digits 0-4",
#      horiz =  TRUE,  nodePar = list(cex = .007))

dend <- color_labels(dend, k = 5,labels=c(0,1,2,3,4),col = c(3, 1, 1, 4, 1))
plot(dend)

#to compare linkage type from ward.D to complete:
hclust_methods <- c("ward.D", "single", "complete", "average", "mcquitty", 
        "median", "centroid", "ward.D2")
digits_dendlist <- dendlist()
for(i in seq_along(hclust_methods)) {
   hc_digits <- hclust(p1.dist, method = hclust_methods[i])   
   digits_dendlist <- dendlist(digits_dendlist, as.dendrogram(hc_digits))
}
names(digits_dendlist) <- hclust_methods

digits_dendlist %>% dendlist(which = c(1,3)) %>% ladderize %>% 
   set("branches_k_color", k=5) %>% 
   # untangle(method = "step1side", k_seq = 3:20) %>%
   tanglegram(faster = TRUE)


plot(length(hc1$height):1, hc1$height, type='o', xlab="K", ylab="height", las=1, 
     xlim=c(1, 10), ylim=c(7, 12))
points(5, hc$height[n-4], col="red", pch=19) # K=5



p = cutree(hc, k=10)
predicts$hc.maxd.compl = p
table(predicts$digits.class,predicts$hc.maxd.compl)

```

```{r}
set.seed(524)
eq = read.csv(file.path(data.dir, "earthquake_4.5_10.csv"))

spatial <- eq[,2:3]
temporal <- eq[,1:2]
library(ggplot2)
ggplot(spatial) + geom_point(aes(latitude, longitude))


library(mclust)
mc = Mclust(spatial, 
            G = 6,    # set number of non-noise components
            initialization=list(noise=TRUE))

plot(mc, what="classification",asp=1, las=1, xlim=c(10, 50))
grid()

summary(mc, parameters = TRUE)


mc_d <- densityMclust(spatial)
plot(mc_d, what='density',data=spatial, type="persp")

#eval.pts = expand.grid(latitude = seq(0, 60, by=.5), longitude = seq(-140, -50, by=0.5))
#mod_pred <- predict(mc_d,eval.pts,what="dens")

#plot(mod, what="density",data=X, type="persp")
# library(plot3D)
# library(rgl)
# library(plotly)
# eval.pts$f <- mod_pred
# #plot3D(eval.pts$latitude,eval.pts$longitude,eval.pts$f)
# p <- plot_ly(eval.pts,x = eval.pts$latitude, y=eval.pts$longitude, z = eval.pts$f) %>% add_surface()
# chart_link = api_create(p, filename="surface-1")
# chart_link
# p
```



## Including Plots

You can also embed plots, for example:

```{r pressure, echo=FALSE}
plot(pressure)
```

Note that the `echo = FALSE` parameter was added to the code chunk to prevent printing of the R code that generated the plot.