---
title: "Boston House Prices"
author: "Eric A. Suess"
format: 
  html:
    embed-resources: true
---

## Predicting house prices: A regression example



```{r}
library(keras)

boston <- dataset_boston_housing()

c(c(train_data, train_targets), c(test_data, test_targets)) %<-% boston
```

Normalize the data.

```{r}
mean <- apply(train_data, 2, mean)

sd <- apply(train_data, 2, sd)

train_data <- scale(train_data, center = mean, scale = sd)

test_data <- scale(test_data, center = mean, scale = sd)
```

Build the network.

```{r}
build_model <- function() {

  model <- keras_model_sequential() %>%
    layer_dense(64, activation = "relu") %>%
    layer_dense(64, activation = "relu") %>%
    layer_dense(1)

  model %>% compile(optimizer = "rmsprop",
                    loss = "mse",
                    metrics = "mae")
  model
}

```

```{r}
k <- 4
fold_id <- sample(rep(1:k, length.out = nrow(train_data)))
num_epochs <- 100
all_scores <- numeric()

for (i in 1:k) {
  cat("Processing fold #", i, "\n")

  val_indices <- which(fold_id == i)

  val_data <- train_data[val_indices, ]
  val_targets <- train_targets[val_indices]

  partial_train_data <- train_data[-val_indices, ]
  partial_train_targets <- train_targets[-val_indices]

  model <- build_model()

  model %>% fit(
    partial_train_data,
    partial_train_targets,
    epochs = num_epochs,
    batch_size = 16,
    validation_split = 0.2,
    verbose = 1
  )

  results <- model %>% evaluate(val_data, val_targets, verbose = 0)
  all_scores[[i]] <- results[['mae']]
}


```

```{r}
all_scores

mean(all_scores)
```

```{r}
num_epochs <- 500
all_mae_histories <- list()
for (i in 1:k) {
  cat("Processing fold #", i, "\n")

  val_indices <- which(fold_id == i)
  val_data <- train_data[val_indices, ]
  val_targets <- train_targets[val_indices]

  partial_train_data <- train_data[-val_indices, ]
  partial_train_targets <- train_targets[-val_indices]

  model <- build_model()
  history <- model %>% fit(
    partial_train_data, partial_train_targets,
    validation_data = list(val_data, val_targets),
    epochs = num_epochs, batch_size = 16, verbose = 0
  )
  mae_history <- history$metrics$val_mae
  all_mae_histories[[i]] <- mae_history
}

all_mae_histories <- do.call(cbind, all_mae_histories)

```

```{r}
average_mae_history <- rowMeans(all_mae_histories)

plot(average_mae_history, xlab = "epoch", type = 'l')

truncated_mae_history <- average_mae_history[-(1:10)]
plot(average_mae_history, xlab = "epoch", type = 'l',
     ylim = range(truncated_mae_history))
```

```{r}
average_mae_history <- rowMeans(all_mae_histories)

plot(average_mae_history, xlab = "epoch", type = 'l')
```

```{r}
predictions <- model %>% predict(test_data)

predictions[1, ]
```