h2o_trees.Rmd

---
title: "Pump it Up: Data Mining the Water Table"
author: Jenna Allen
date: 2018-05-22
output: github_document
---

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

```{r}
library(tidyverse)
library(h2o)
```

```{r}
#initialize h20
h2o.init()
```

```{r}
predictors <- names(train_mod_data)[-18]
target <- "status_group"
```

```{r}
training_hex <- as.h2o(train_mod_data, destination_frame = "training.hex")
test_hex <- as.h2o(test_mod_data, destination_frame = "test.hex")
```

```{r}
rf_model <- h2o.randomForest(
  x = predictors,
  y = target,
  training_frame = training_hex,
  model_id = "rf_ntrees100",
  balance_classes = TRUE,
  ntrees = 100, 
  mtries = 10,
  seed = 1234)
```

```{r}
h2o.confusionMatrix(rf_model)
```

```{r}
test_predictions <- as.data.frame(h2o.predict(rf_model, test_hex))
```

```{r}
submission_1 <- test_mod_data %>% 
  select(id) %>% 
  bind_cols(select(test_predictions, predict)) %>% 
  mutate_at(vars(predict), funs(str_replace_all(., "_", " "))) %>% 
  rename(status_group = predict)

write.csv(submission_1, "submission1.csv", row.names = FALSE)
```

Classification rate of 0.8141. The leader has classification rate of 0.8285. This result put me in the top 11% of models.

```{r}
rf_model_imp <- h2o.varimp(rf_model)
```

```{r}
rf_model_1000 <- h2o.randomForest(
  x = predictors,
  y = target,
  training_frame = training_hex,
  model_id = "rf_ntrees1000",
  balance_classes = TRUE,
  ntrees = 1000, 
  mtries = 20,
  seed = 1234)
```

```{r}
h2o.confusionMatrix(rf_model_1000)
```

```{r}
test_predictions_1000 <- as.data.frame(h2o.predict(rf_model_1000, test_hex))
```

```{r}
submission_2 <- test_mod_data %>% 
  select(id) %>% 
  bind_cols(select(test_predictions_1000, predict)) %>% 
  mutate_at(vars(predict), funs(str_replace_all(., "_", " "))) %>% 
  rename(status_group = predict)

write.csv(submission_2, "submission2.csv", row.names = FALSE)
```

```{r}
rf_model_nfold <- h2o.randomForest(
  x = predictors,
  y = target,
  nfolds = 10,
  training_frame = training_hex,
  model_id = "rf_ntrees100_nfold_10",
  balance_classes = TRUE,
  ntrees = 100, 
  mtries = 10,
  seed = 1234)
```

```{r}
h2o.confusionMatrix(rf_model_nfold)
```


```{r}
rf_model_nobal <- h2o.randomForest(
  x = predictors,
  y = target,
  training_frame = training_hex,
  model_id = "rf_ntrees100_nobal",
  ntrees = 100, 
  mtries = 10,
  seed = 1234)
```
```{r}
h2o.confusionMatrix(rf_model_nobal)
```


```{r}
rf_model_100_mtries_20 <- h2o.randomForest(
  x = predictors,
  y = target,
  training_frame = training_hex,
  model_id = "rf_ntrees100_mtries20",
  balance_classes = TRUE,
  ntrees = 100, 
  mtries = 20,
  seed = 1234)
```

```{r}
h2o.confusionMatrix(rf_model_100_mtries_20)
```

```{r}
rf_model_100_mtries5 <- h2o.randomForest(
  x = predictors,
  y = target,
  training_frame = training_hex,
  model_id = "rf_ntrees100_mtries5",
  balance_classes = TRUE,
  ntrees = 100,
  mtries = 5,
  seed = 1234)
```

```{r}
h2o.confusionMatrix(rf_model_100_mtries5)
```

```{r}
rf_model_100_mtries23 <- h2o.randomForest(
  x = predictors,
  y = target,
  training_frame = training_hex,
  model_id = "rf_ntrees100_mtries23",
  balance_classes = TRUE,
  ntrees = 100,
  mtries = 23,
  seed = 1234)
```

```{r}
h2o.confusionMatrix(rf_model_100_mtries23)
```


```{r}
gbm_model <- h2o.gbm(
  x = predictors,
  y = target,
  training_frame = training_hex,
  model_id = "gbm_ntrees100",
  balance_classes = TRUE,
  ntrees = 100,
  seed = 1234)
```

```{r}
h2o.confusionMatrix(gbm_model)
```

```{r}
xgboost_model <- h2o.xgboost(
  x = predictors,
  y = target,
  training_frame = training_hex,
  model_id = "xgboost_ntrees100",
  ntrees = 100,
  seed = 1234)
```

```{r}
h2o.confusionMatrix(xgboost_model)
```