Biochar_ML_Effect.Rmd

---
title: "R Biochar Analysis"
output:
  html_document:
    df_print: paged
---
```{r message=FALSE, warning=FALSE}

```


```{r message=FALSE, warning=FALSE}
#Library
library(plotly)
library(ggplot2)
library(car)
library(MVN)
library(survival)
library(coin)
#install.packages("devtools")
library(devtools)
#install_github("edwindj/ffbase", subdir="pkg")
library(ffbase)
#install_github("mtennekes/tabplot")
library(tabplot)
library(corrplot)
library(ggpubr)

library(randomForestExplainer)
library(magrittr)
library(dplyr)
library(randomForest)
library(datasets)
library(caret)
lib_vect <- c("raster","rpart","PresenceAbsence","AUC","randomForest","ecospat","biomod2","ggplot2","caret","dismo","ncf","tidyr")

sapply(lib_vect,require,character.only=TRUE)

#devtools::install_github("rstudio/tensorflow")
library(tensorflow)

#Load the necessary libraries
library(keras)
library(grid)
library(forestploter)
library(caret)
library(e1071) 
library(neuralnet)
 library(dplyr)
library(nnet)
library(stats)
library(Hmisc)
library(matrixStats)
```



```{r}
SVM_cale <- read.csv("Dataset_biochar_scaled.csv")
```


```{r}
#Binary RF Model 记得把训练使用的数据库每个分别摘取出来
set.seed(2095)
index <- createDataPartition(SVM_cale$category, p=0.75, list=FALSE)
saveforRF <- index

SVM_cale <- read.csv("Dataset_biochar_scaled.csv")
SVM_cale$category <- as.factor(SVM_cale$category)
colnames(SVM_cale)<-c("concentration", "growth_or_not","reproduction_or_not", "survival_or_not","temperature","PAHs","Biochar_EC","Biochar_pH","Biochar_C","Biochar_N","Biochar_Cd","Biochar_Cu","Biochar_Pb","Soil_pH","test_organisms","Biochar_plant_or_not", "Biochar_mix_or_not","Biochar_sludge_or_not", "Biochar_waste_or_not","sand", "silt", "clay","Soil_organic_matter","category")
Y_cate_train <- SVM_cale$category[index]
Y_cate_test <- SVM_cale$category[-index]
X_cate_train <- SVM_cale[index,]
X_cate_test <- SVM_cale[-index,]
regr_cate <- randomForest(category~., ntrees = 3000, nodesize = 11, mtry=2,data = X_cate_train, localImp = TRUE)
predict_cate_x <- predict(regr_cate,X_cate_test)
predict_cate_train <- predict(regr_cate, X_cate_train)
result_matrix <- confusionMatrix(table(predict_cate_x, X_cate_test$category))
train_matrix <- confusionMatrix(table(predict_cate_train, X_cate_train$category))
train_matrix #Confusion matrix for training set,  binary random forest
result_matrix #Confusion matrix for test set,  binary random forest

RF_BINARY <- regr_cate
```

```{r}
#Binary NN model
set.seed(32)

#32 
nn_index <- createDataPartition(SVM_cale$category, p=0.75, list=FALSE)
nn_train <- SVM_cale[nn_index,]
nn_test <- SVM_cale[-nn_index,]

 
nnet1 <- nnet(category~., nn_train, size = 12, linout = FALSE)

 
ypred <- predict(nnet1, nn_test, type = "class") 
trainpred <- predict(nnet1, nn_train, type = "class")

confusionMatrix(as.factor(ypred), nn_test$category) #Confusion matrix for test set,  binary neural network

confusionMatrix(as.factor(trainpred), nn_train$category) #Confusion matrix for training set, binary neural network
NN_BINARY <- nnet1
```

```{r}
# LINEAR SVM BINARY
set.seed(2095)
svm_model2 <- svm(formula = category~.,data = X_cate_train, type = 'C-classification', kernel = 'linear')
print(svm_model2)
pred_svm2 <- predict(svm_model2, X_cate_test)
tr_svm2 <- predict(svm_model2, X_cate_train)
confusionMatrix(data = pred_svm2, X_cate_test$category) #Confusion matrix for test set,  binary linear SVM
confusionMatrix(data = tr_svm2, X_cate_train$category) #Confusion matrix for training set,  binary linear SVM
LINEAR_SVM_BINARY <- svm_model2
```
```{r}
# Gaussian SVM BINARY
set.seed(2095)
svm_model2 <- svm(formula = category~.,data = X_cate_train, type = 'C-classification', kernel = 'radial')
print(svm_model2)
pred_svm2 <- predict(svm_model2, X_cate_test)
tr_svm2 <- predict(svm_model2, X_cate_train)
confusionMatrix(data = pred_svm2, X_cate_test$category) #Confusion matrix for test set, binary gaussian SVM
confusionMatrix(data = tr_svm2, X_cate_train$category) #Confusion matrix for test set, binary gaussian SVM
GAUSSIAN_SVM_BINARY <- svm_model2
```

```{r}
#quan_rf
set.seed(2095)

train_rownames2 <- read.csv("train_rownames.csv")
test_rownames2 <- read.csv("test_rownames.csv")


train_rownames <- as.numeric(train_rownames2$train_rownames)
test_rownames <- as.numeric(test_rownames2$test_rownames)

rf_quan1 <- read.csv("Dataset_biochar_scaled_witheffect.csv")
rf_quan <- rf_quan1[,-1]
colnames(rf_quan) <- c("effect","concentration", "growth_or_not","reproduction_or_not", "survival_or_not","temperature","PAHs","Biochar_EC","Biochar_pH","Biochar_C","Biochar_N","Biochar_Cd","Biochar_Cu","Biochar_Pb","Soil_pH","test_organisms","Biochar_plant_or_not", "Biochar_mix_or_not","Biochar_sludge_or_not", "Biochar_waste_or_not","sand", "silt", "clay","Soil_organic_matter")
index <- createDataPartition(rf_quan$effect, p=0.75, list=FALSE)
rf_x_train <- rf_quan[ train_rownames, ]
rf_x_test <- rf_quan[test_rownames, ]
rf_y_train <- rf_quan$effect[index]
rf_y_test<-rf_quan$effect[-index]
regr_rf <- randomForest(effect~., ntrees = 3000, nodesize = 11, mtry=2,data = rf_x_train, localImp = TRUE)

predictions_train <- predict(regr_rf, rf_x_train)
predictions_test <- predict(regr_rf, rf_x_test)

print(paste0('R2: ' ,caret::postResample(predictions_train, rf_x_train$effect)['Rsquared'] )) #Training set R square
print(paste0('R2: ' ,caret::postResample(predictions_test, rf_x_test$effect)['Rsquared'] )) #Test set R square
 
train_rmse1 <- sqrt(mean((predictions_train - rf_x_train$effect)^2))
test_rmse1 <- sqrt(mean((predictions_test - rf_x_test$effect)^2))
train_rmse1 #Training set root mean square error
test_rmse1 #Test set root mean square error

QUAN_RF <- regr_rf
```

```{r}
#Ternary Linear SVM Model
three_category <- read.csv("three_category.csv")
SVM_cale$category <- three_category$rf_three.Category
SVM_cale$category <- as.factor(SVM_cale$category)
set.seed(2095)
svm_index <- createDataPartition(SVM_cale$category, p=0.75, list=FALSE)
train_svm <- SVM_cale[svm_index,]
test_svm <- SVM_cale[-svm_index,]
train_c <- trainControl(method = "cv", number = 100)
svm_model_three <- caret::train(category~.,data = train_svm, method = "svmLinear", trControl = train_c)
print(svm_model_three)
pred_svm <- predict(svm_model_three, test_svm)
trains_svm <- predict(svm_model_three, train_svm)
confusionMatrix(data = pred_svm, test_svm$category) #Confusion matrix for test set, ternary linear SVM
confusionMatrix(data = trains_svm, as.factor(train_svm$category)) #Confusion matrix for training set, ternary linear SVM
TERNARY_SVM_LINEAR <- svm_model_three
```



```{r}
#Ternary Gaussian SVM Model
set.seed(2095)
svm_index <- createDataPartition(SVM_cale$category, p=0.75, list=FALSE)
train_svm <- SVM_cale[svm_index,]
test_svm <- SVM_cale[-svm_index,]
train_c <- trainControl(method = "cv", number = 100)
svm_model_three <- caret::train(category~.,data = train_svm, method = "svmRadial", trControl = train_c)
print(svm_model_three)
pred_svm <- predict(svm_model_three, test_svm)
trains_svm <- predict(svm_model_three, train_svm)
confusionMatrix(data = pred_svm, test_svm$category) #Confusion matrix for test set, ternary gaussian SVM
confusionMatrix(data = trains_svm, as.factor(train_svm$category)) #Confusion matrix for training set, ternary gaussian SVM
TERNARY_SVM_Gaussian <- svm_model_three
```

```{r}
#Ternary RF model
set.seed(2095)

cccccc1 <- read.csv("Dataset_biochar_colnamed.csv")
rf_three_use <- cccccc1
rf_three_use$category <- as.factor(rf_three_use$category)

index <- createDataPartition(rf_three_use$category, p=0.75, list=FALSE)
Y_cate_train <- rf_three_use$category[index]
Y_cate_test <- rf_three_use$category[-index]
X_cate_train <- rf_three_use[index,]
X_cate_test <- rf_three_use[-index,]
regr_cate_three <- randomForest(category~., data = X_cate_train, localImp = TRUE)

predict_cate_x <- predict(regr_cate_three,X_cate_test)
predict_cate_train <- predict(regr_cate_three, X_cate_train)
result_matrix_three_RF <- confusionMatrix(table(predict_cate_x, X_cate_test$category))
train_matrix_three_RF <- confusionMatrix(table(predict_cate_train, X_cate_train$category))
train_matrix_three_RF #Confusion matrix for training set, ternary random forest
result_matrix_three_RF #Confusion matrix for test set, ternary random forest
TERNARY_RF <- regr_cate_three


```

```{r}
#Ternary NN model
set.seed(2095)
rf_three_use$category <- as.factor(rf_three_use$category)
index <- createDataPartition(rf_three_use$category, p=0.75, list=FALSE)
X_cate_train <- rf_three_use[index,]
X_cate_test <- rf_three_use[-index,]
regr_cate_three_NN <- nnet(category ~ . , X_cate_train, size = 12,linout = FALSE)
predict_cate_x <- predict(regr_cate_three_NN,X_cate_test, type = "class")
predict_cate_train <- predict(regr_cate_three_NN, X_cate_train, type = "class")
result_matrix_three_NN <- confusionMatrix(table(predict_cate_x, X_cate_test$category))
train_matrix_three_NN <- confusionMatrix(table(predict_cate_train, X_cate_train$category))
train_matrix_three_NN #Confusion matrix for training set, ternary neural network
result_matrix_three_NN #Confusion matrix for test set, ternary neural network
TERNARY_NN <- regr_cate_three_NN
rf_scale <- read.csv("Dataset_biochar_original.csv")
rf_scale$category <- as.factor(rf_scale$category)
```

```{r}
#optimized RF
set.seed(2095)
index <- createDataPartition(rf_scale$category, p=0.75, list=FALSE)
rf_here <- rf_scale[,-1]


Y_cate_train <- rf_here$category[index]
Y_cate_test <- rf_here$category[-index]
X_cate_train <- rf_here[index,]
X_cate_test <- rf_here[-index,]



subset_rf <- rf_scale[rf_scale$effect > 10 | rf_scale$effect < -10, ]
subset_rf <- subset_rf[,-1]

set.seed(2095)
index <- createDataPartition(subset_rf$category, p=0.75, list=FALSE)
X_cate_train_sub <- subset_rf[index,]
X_cate_test_sub <- subset_rf[-index,]
regr_op_test <- randomForest(category~., data = X_cate_train_sub, localImp = TRUE)





predict_cate_x <- predict(regr_op_test,X_cate_test)
predict_cate_train <- predict(regr_op_test, X_cate_train)
result_matrix_op_RF <- confusionMatrix(table(predict_cate_x, X_cate_test$category))
train_matrix_op_RF <- confusionMatrix(table(predict_cate_train, X_cate_train$category))
train_matrix_op_RF #Confusion matrix for training set, optimized random forest
result_matrix_op_RF # Confusion matrix for test set, optimized random forest
OPTIMIZED_RF_BINARY <- regr_op_test
```


```{r}
#List of final models
#BINARY MODELS
# RF_BINARY
# NN_BINARY
# LINEAR_SVM_BINARY
# GAUSSIAN_SVM_BINARY
#QUANTITATIVE Random forest
# QUAN_RF
#TERNARY MODELS
# TERNARY_SVM_LINEAR
# TERNARY_SVM_Gaussian
# TERNARY_RF
# TERNARY_NN
#Optimized RF
# OPTIMIZED_RF_BINARY
```


```{r}

```


1
Add a new chunk by clicking the *Insert Chunk* button on the toolbar or by pressing *Cmd+Option+I*.

When you save the notebook, an HTML file containing the code and output will be saved alongside it (click the *Preview* button or press *Cmd+Shift+K* to preview the HTML file). 

The preview shows you a rendered HTML copy of the contents of the editor. Consequently, unlike *Knit*, *Preview* does not run any R code chunks. Instead, the output of the chunk when it was last run in the editor is displayed.