Part 2.Rmd

---
title: "Question 2"
author: "Maryam Imani, Azadeh Samadian"
date: "October 23, 2018"
output: html_document
---

```{r , echo=TRUE}
mydata <- read.table("token/networkomisegoTX.txt", header=FALSE, 
  	sep=" ",  col.names= c("fromNodeID", "toNodeID", "unixTime", "tokenAmount"), colClasses=c('factor', 'factor', 'numeric', 'numeric') )

library(anytime)
mydata$newDate <- anydate(mydata$unixTime)
df <- as.data.frame.matrix(mydata) 
head(mydata)
```

```{r , echo=TRUE}
mydata2 <- read.table("tokenPrices/OMISEGO", header=TRUE, 
  	sep="\t",  col.names= c("Date",	"Open",	"High"	,"Low"	,"Close",	"Volume",	"MarketCap"), colClasses=c('factor', 'numeric', 'numeric','numeric', 'numeric', 'numeric', 'numeric') )
mydata2$newDate <- as.Date(strptime(x = as.character(mydata2$Date), format = "%m/%d/%Y"))
df_set2 <- as.data.frame.matrix(mydata2) 
head(mydata2)

```

# Outlier

```{r , echo=TRUE}
maxTotalSupply<-140245398
subUnit<-10^18
outlier_value <- maxTotalSupply*subUnit
outlierData <- mydata [which(mydata$tokenAmount > outlier_value),]
#outlierData2 <- mydata [which(mydata$tokenAmount < 1.0e+16),]
#head(outlierData)
#library("dplyr")
#filter(mydata, tokenAmount < outlier_value)
message("Total Number of Outliers: ", length(outlierData$tokenAmount) )
```

## Number of users that included in outliers transactions: 
```{r , echo=TRUE}
users <- c(outlierData$fromNodeID, outlierData$toNodeID)
uniqueUsers <- unique(users)
message(length(uniqueUsers), " users are included in outliers transactions.")
```
## Remove outliers:
```{r , echo=TRUE}
withoutOutlierData <- mydata [which(mydata$tokenAmount < outlier_value),]
withoutOutlierData <- withoutOutlierData [which(withoutOutlierData$tokenAmount > 1.000000e+17),]

cleanDF <- as.data.frame.matrix(withoutOutlierData) 

```

## Take sample from population
```{r, echo = TRUE}

#cleanDF[sample(nrow(cleanDF),size=1000,replace=TRUE),]

```

##find max and min
```{r , echo=TRUE}
library('anytime')
minimum <- min(cleanDF$tokenAmount, na.rm=TRUE)
cleanDF$NewAmount <- cleanDF$tokenAmount / minimum # Normalizing the tokenAmount (we can ignore this part)

cleanDF <- cleanDF[order(cleanDF$NewAmount),]

minimum <- min(cleanDF$NewAmount, na.rm=TRUE)
maximum<-max(cleanDF$NewAmount, na.rm=TRUE)
message("maximum amount is: ", maximum )
numOfLayers <- 3
layerWidth <- maximum/numOfLayers

#cleanDF$layers <- cut(cleanDF$NewAmount, c(minimum:maximum , layerWidth))
head(cleanDF)

```

## finding layers and correlation 

```{r, echo = TRUE}
# In this part, we used frequncy of transaction per date as a feature. Then, we found the corrolation of this feature and the amount of close date. 

minimum <- min(cleanDF$tokenAmount, na.rm=TRUE)
maximum<-max(cleanDF$tokenAmount, na.rm=TRUE)

x2 <- minimum
layerNum <- 0
resultVec <- c()
xlabel <-c()

while ( x2*2<maximum) {
  
  x1 <- x2
  x2 <- x2 * 2
  
  df2 <- subset(cleanDF, cleanDF$tokenAmount >= x1 & cleanDF$tokenAmount < x2  , select=c(newDate))
  freqCountDF <- data.frame(table(df2$newDate))
  
  if(nrow(freqCountDF) > 2){
    
    layerNum <- layerNum +1
    
    # Counting the frequency of dates in the first dataset
    colnames(freqCountDF) <- c("newDate", "Freq")
  
    # merging two datasets together to find Closing amount in a given date
    mergeDF <- subset(merge(freqCountDF,df_set2,by="newDate"), select=c(Freq, Close))
    
    # Counting the corrolation between closing amount and frequency of transaction per date
    resultVec[layerNum] <- cor(mergeDF$Freq, mergeDF$Close, method = "pearson")
    
    xlabel[layerNum] <- paste0(layerNum , ": ", x1, "< tokenAmount < " , x2 , "  Cor=", resultVec[layerNum] )
  }
}
barplot( resultVec, names.arg=c(1:length(resultVec)), xlab = "Layer", ylab = "Correlation")
print(xlabel)

```

```{r, echo = TRUE}
# In this part, we used frequncy of transaction per date as a feature. Then, we found the corrolation of this feature and the amount of close date. 

minimum <- min(cleanDF$tokenAmount, na.rm=TRUE)
maximum<-max(cleanDF$tokenAmount, na.rm=TRUE)

y <- maximum
x2 <- minimum
layerNum <- 0
resultVec <- c()
xlabel <-c()

while ( x2*10<maximum) {
  

  x1 <- x2
  x2 <- x2 * 10
  
  df2 <- subset(cleanDF, cleanDF$tokenAmount >= x1 & cleanDF$tokenAmount < x2  , select=c(newDate))
  freqCountDF <- data.frame(table(df2$newDate))
  
  if(nrow(freqCountDF) > 2){
    
    layerNum <- layerNum +1
    
    # Counting the frequency of dates in the first dataset
    colnames(freqCountDF) <- c("newDate", "Freq")
  
    # merging two datasets together to find Closing amount in a given date
    mergeDF <- subset(merge(freqCountDF,df_set2,by="newDate"), select=c(Freq, Close))
    
    # Counting the corrolation between closing amount and frequency of transaction per date
    resultVec[layerNum] <- cor(mergeDF$Freq, mergeDF$Close, method = "pearson")
    
    xlabel[layerNum] <- paste0(layerNum , ": ", x1, "< tokenAmount < " , x2 , "  Cor=", resultVec[layerNum] )
    
  }
}
barplot( resultVec, names.arg=c(1:length(resultVec)), xlab = "Layer", ylab = "Correlation")
print(xlabel)
```
### Cumalitive Correlation 


```{r, echo = TRUE}
# In this part, we used frequncy of transaction per date as a feature. Then, we found the corrolation of this feature and the amount of close date. 

minimum <- min(cleanDF$tokenAmount, na.rm=TRUE)
maximum<-max(cleanDF$tokenAmount, na.rm=TRUE)

x2 <- minimum
layerNum <- 0
resultVec <- c()

xlabel <-c()

while ( x2*2<maximum) {
  
  x1 <- x2
  x2 <- x2 * 2
  
  df2 <- subset(cleanDF, cleanDF$tokenAmount < x2  , select=c(newDate))
  
  
  freqCountDF <- data.frame(table(df2$newDate))
  
  if(nrow(freqCountDF) > 2){
    
    layerNum <- layerNum + 1
    
    
    # Counting the frequency of dates in the first dataset
    colnames(freqCountDF) <- c("newDate", "Freq")
  
    # merging two datasets together to find Closing amount in a given date
    mergeDF <- subset(merge(freqCountDF,df_set2,by="newDate"), select=c(Freq, Close))
    
    # Counting the corrolation between closing amount and frequency of transaction per date
    resultVec[layerNum] <- cor(mergeDF$Freq, mergeDF$Close, method = "pearson")
    
    xlabel[layerNum] <- paste0(layerNum , ": tokenAmount < " , x2 , "  Cor=", resultVec[layerNum] )
    print(xlabel[layerNum])
  }
}
barplot( resultVec, names.arg=c(1:length(resultVec)), xlab = "Layer", ylab = "Cumilitive Correlation")
print(xlabel)
```


```{r, echo = TRUE}
# In this part, we used frequncy of transaction per date as a feature. Then, we found the corrolation of this feature and the amount of close date. 

minimum <- min(cleanDF$tokenAmount, na.rm=TRUE)
maximum<-max(cleanDF$tokenAmount, na.rm=TRUE)

y <- maximum
x2 <- minimum
layerNum <- 0
resultVec <- c()

xlabel <-c()

while ( x2*10<maximum) {
  
  x1 <- x2
  x2 <- x2 * 10
  
  df2 <- subset(cleanDF, cleanDF$tokenAmount < x2  , select=c(newDate))
  
  freqCountDF <- data.frame(table(df2$newDate))
  
  if(nrow(freqCountDF) > 2){
    
    layerNum <- layerNum + 1
    
    
    # Counting the frequency of dates in the first dataset
    colnames(freqCountDF) <- c("newDate", "Freq")
  
    # merging two datasets together to find Closing amount in a given date
    mergeDF <- subset(merge(freqCountDF,df_set2,by="newDate"), select=c(Freq, Close))
    
    # Counting the corrolation between closing amount and frequency of transaction per date
    resultVec[layerNum] <- cor(mergeDF$Freq, mergeDF$Close, method = "pearson")
    
    xlabel[layerNum] <- paste0(layerNum , ": tokenAmount < " , x2 , "  Cor=", resultVec[layerNum] )
  }
}
barplot( resultVec, names.arg=c(1:length(resultVec)), xlab = "Layer", ylab = "Cumilitive Correlation")
print(xlabel)
```