HW5(NEW)
b06502078 謝承軒
library(dplyr, warn.conflicts = FALSE)
library(ggplot2)
library(magrittr)
library(e1071)
library(scales)
library(reshape2)
library(stats)
library(jpeg)
library(factoextra)## Welcome! Related Books: `Practical Guide To Cluster Analysis in R` at https://goo.gl/13EFCZPART 1 part about PL(英格蘭超級足球聯賽)
分析各種進攻數據與聯賽排名的相關性
pl = read.csv("C:/Users/admin/Desktop/R/pl.csv")plCor<-cor(pl[,c(1:7)])
plMelt<-melt(plCor,varnames = c("x","y"),value.name = "Correlation")
plMelt<-plMelt[order(plMelt$Correlation),]
ggplot(plMelt,aes(x=x,y=y))+
geom_tile(aes(fill=Correlation))+
scale_fill_gradient2(low="red",mid="white",high="darkblue",guide=guide_colorbar(ticks=FALSE,barheight=10),limits=c(-1,1))+
theme_minimal()+
labs(x=NULL,y=NULL)
RANK = 聯賽排名 SCORE = 得分 ASIST = 助攻
SHOT = 射門 SHOT ON TARGET = 射正(朝球門射出的球)
SQUARE PASS = 橫傳 CORNER KICK = 角球
可看出 傳球多寡對於勝負的結果影響不大 甚至可以說是極低
PART 2 part about samsung gaming
分析今年英雄聯盟世界冠軍隊隊伍數據
這支隊伍於團隊表現上面是目前五個聯賽裡最優秀的
SSG = read.csv("C:/Users/admin/Desktop/R/SSG.csv")SSGCor<-cor(SSG[,c(1:4,6:8)])
SSGMelt<-melt(SSGCor,varnames = c("x","y"),value.name = "Correlation")
SSGMelt<-SSGMelt[order(SSGMelt$Correlation),]
ggplot(SSGMelt,aes(x=x,y=y))+
geom_tile(aes(fill=Correlation))+
scale_fill_gradient2(low="red",mid="white",high="darkblue",guide=guide_colorbar(ticks=FALSE,barheight=10),limits=c(-1,1))+
theme_minimal()+
labs(x=NULL,y=NULL)
大致上各項數據都符合邏輯
可以看出這個遊戲的擊殺助攻與死亡並沒有太大的相關性(畢竟dota遊戲)
而這支隊伍由於團隊溝通(隊友配合)較佳 因此擊殺與助攻相關性頗大
由普遍來說最為直觀的數據 擊殺死亡助攻(kda)判斷比賽勝負
sec 1
由擊殺與死亡判斷
train_data1 <- SSG[1:22,c(1,2,4)]
test_data1 <- SSG[23:37, c(2,4)]
svm_fit1 = svm(as.factor(W.L) ~ ., data = train_data1,
kernel = "polynomial",
cost = 20, scale = FALSE)
plot(svm_fit1, train_data1)
SSG_predicted1 <- predict(svm_fit1, test_data1)result1 = table(SSG_predicted1,SSG[23:37,1])
print(result1)##
## SSG_predicted1 0 1
## 0 3 5
## 1 0 7準確率為0.667 偏低
sec 2
加入時間因素 由分鐘擊殺與死亡判斷
train_data2 <- SSG[1:22,c(1,6,8)]
test_data2 <- SSG[23:37, c(6,8)]
svm_fit2 = svm(as.factor(W.L) ~ ., data = train_data2,
kernel = "polynomial",
cost = 20, scale = FALSE)
plot(svm_fit2, train_data2)
SSG_predicted2 <- predict(svm_fit2, test_data2)result2 = table(SSG_predicted2,SSG[23:37,1])
print(result2)##
## SSG_predicted2 0 1
## 0 3 1
## 1 0 11準確率為 0.933
sec 3
由分鐘擊殺與死亡判斷 並將training data增加
train_data3 <- SSG[1:27,c(1,6,8)]
test_data3 <- SSG[28:37, c(6,8)]
svm_fit3 = svm(as.factor(W.L) ~ ., data = train_data3,
kernel = "polynomial",
cost = 20, scale = FALSE)
plot(svm_fit3, train_data3)
SSG_predicted3 <- predict(svm_fit3, test_data3)result3 = table(SSG_predicted3,SSG[28:37,1])
print(result3)##
## SSG_predicted3 0 1
## 0 1 0
## 1 0 9