Advertisement
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- datas[-1]
- parsed <- na.omit(datas[-1])
- parsed
- #for(i in 1:length(parsed) ){
- #shapiro.test(parsed[i])
- #}
- shapiro.test(parsed$TTG)
- shapiro.test(parsed$Mochevina)
- shapiro.test(parsed$Ferritin)
- shapiro.test(parsed$VitD)
- shapiro.test(parsed$MKislota)
- shapiro.test(parsed$B12)
- shapiro.test(parsed$CholesterinVP)
- shapiro.test(parsed$GGT)
- shapiro.test(parsed$Cbelok)
- shapiro.test(parsed$Kreatinin)
- shapiro.test(parsed$Albumin)
- shapiro.test(parsed$Trigleceridy)
- shapiro.test(parsed$ALAT)
- shapiro.test(parsed$CholesterinNP)
- shapiro.test(parsed$Age)
- # у всех p-v < 0.05 - ненормально распределены
- #cor.test(parsed$Trigleceridy, parsed$TTG, method="kendall") #p-value = 0.009918
- cor.test(parsed$Trigleceridy, parsed$Mochevina, method="kendall") #p-value = 1.115e-06
- cor.test(parsed$Trigleceridy, parsed$Ferritin, method="kendall") #p-value = 2.626e-16
- #cor.test(parsed$Trigleceridy, parsed$VitD, method="kendall") #p-value = 0.0009767
- cor.test(parsed$Trigleceridy, parsed$MKislota, method="kendall") #p-value < 2.2e-16
- #cor.test(parsed$Trigleceridy, parsed$B12, method="kendall") #p-value = 0.116
- #cor.test(parsed$Trigleceridy, parsed$CholesterinVP, method="kendall") #p-value < 2.2e-16
- cor.test(parsed$Trigleceridy, parsed$GGT, method="kendall") #p-value < 2.2e-16
- cor.test(parsed$Trigleceridy, parsed$Cbelok, method="kendall") #p-value < 2.2e-16
- cor.test(parsed$Trigleceridy, parsed$Kreatinin, method="kendall") #p-value = 3.853e-11
- #cor.test(parsed$Trigleceridy, parsed$Albumin, method="kendall") #p-value = 0.04954
- cor.test(parsed$Trigleceridy, parsed$Cholesterin, method="kendall") #p-value < 2.2e-16
- #cor.test(parsed$Trigleceridy, parsed$Trigleceridy, method="kendall") #p-value < 2.2e-16
- cor.test(parsed$Trigleceridy, parsed$ALAT, method="kendall") #p-value < 2.2e-16
- cor.test(parsed$Trigleceridy, parsed$CholesterinNP, method="kendall") #p-value < 2.2e-16
- cor.test(parsed$Trigleceridy, parsed$Age, method="kendall") #p-value < 2.2e-16
- library(caret)
- fitControl <- trainControl(method="repeatedcv",number=10,repeats=10)
- tr1=train(parsed[,-13],parsed$Trigleceridy,method ="rpart",trControl=fitControl)
- plot(tr1)
- parsedTen <- data.frame(
- Trigleceridy=parsed$Trigleceridy,
- Mochevina=parsed$Mochevina,
- Ferritin=parsed$Ferritin,
- MKislota=parsed$MKislota,
- GGT=parsed$GGT,
- Cbelok = parsed$Cbelok,
- Kreatinin = parsed$Kreatinin,
- Cholesterin = parsed$Cholesterin,
- ALAT = parsed$ALAT,
- CholesterinNP = parsed$CholesterinNP,
- Age = parsed$Age
- )
- parsedTen
- lm.t <- lm(parsedTen$Trigleceridy ~ ., data=parsedTen)
- lm.t
- lm.t <- lm(parsedTen$Trigleceridy ~
- parsedTen$Mochevina +
- parsedTen$Ferritin +
- parsedTen$MKislota +
- parsedTen$GGT +
- parsedTen$Cbelok +
- parsedTen$Kreatinin +
- parsedTen$Cholesterin +
- parsedTen$CholesterinNP
- , data=parsedTen)
- lm.t <- lm(parsedTen$Trigleceridy ~
- parsedTen$Ferritin +
- parsedTen$MKislota +
- parsedTen$Cbelok +
- parsedTen$Kreatinin +
- parsedTen$Cholesterin +
- parsedTen$CholesterinNP
- , data=parsedTen)
- lm.t <- lm(parsedTen$Trigleceridy ~
- parsedTen$Ferritin +
- parsedTen$MKislota +
- parsedTen$Cbelok +
- parsedTen$Cholesterin +
- parsedTen$CholesterinNP
- , data=parsedTen)
- lm.t <- lm(parsedTen$Trigleceridy ~
- parsedTen$Ferritin +
- parsedTen$MKislota +
- parsedTen$Cbelok +
- parsedTen$Cholesterin
- , data=parsedTen) #изменить
- lm.t
- summary(lm.t) # 0.3124 < 1 - маленькая зависимость
- library(MLmetrics)
- mse <- MSE(y_pred = exp(lm.t$fitted.values), y_true = parsedTen$Trigleceridy)
- RMSE(y_pred = exp(lm.t$fitted.values), y_true = parsedTen$Trigleceridy)
- COOA = mean(abs(lm.t$residuals/parsedTen$Trigleceridy)) * 100
- COOA
- parsedOther <- data.frame(
- Trigleceridy = parsed$Trigleceridy,
- TTG=parsed$TTG,
- VitD=parsed$VitD,
- B12=parsed$B12,
- CholesterinVP=parsed$CholesterinVP,
- Albumin=parsed$Albumin
- )
- lm.t2 <- lm(parsedOther$Trigleceridy ~ ., data=parsedOther)
- lm.t2
- summary(lm.t2) # 0.08641 < 1 - маленькая зависимость
- mse <- MSE(y_pred = exp(lm.t2$fitted.values), y_true = parsedOther$Trigleceridy)
- mse
- RMSE(y_pred = exp(lm.t2$fitted.values), y_true = parsedOther$Trigleceridy)
- COOA = mean(abs(lm.t2$residuals/parsedOther$Trigleceridy)) * 100
- COOA
- set.seed(43423)
- tr2=train(parsedTen,parsedTen$Trigleceridy,method ="nnet", trControl=fitControl)
- #nnet.dataset = train(parsedTen$Trigleceridy ~ ., data = parsedTen, method ="nnet")
- tr2$finalModel
- plot(tr2)
- library(nnet)
- library(MLmetrics)
- for(i in 1:10){
- nnet <- nnet(parsedTen$Trigleceridy ~ ., parsedTen, decay=0.1, size=i)
- ind <- sample(2, nrow(parsedTen), replace = TRUE, prob=c(0.7, 0.3))
- trainset = parsedTen[ind == 1,]
- testset = parsedTen[ind == 2,]
- mse <- MSE(y_pred = exp(nnet$fitted.values), y_true = parsedTen$Trigleceridy)
- mse
- RMSE <- RMSE(y_pred = exp(nnet$fitted.values), y_true = parsedTen$Trigleceridy)
- COOA = mean(abs(nnet$residuals/parsedTen$Trigleceridy)) * 100
- print(i)
- print(mse)
- print(RMSE)
- print(COOA)
- }
- nnet1 <- nnet(trainset$Trigleceridy ~ ., trainset, decay=0.1, size=2)
- pr1 <- predict(nnet1, testset)
- pr1
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement