---
title: "Generalized Linear Model Theta Forecast with attention Pt.3"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Generalized Linear Model Theta Forecast with attention Pt.3}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r fig.width=7.5}
library(forecast)
library(ahead)

attention_types <- c("dot_product", "scaled_dot_product", 
  "cosine", "exponential", 
  "gaussian", "linear",
  "value_based", "hybrid", 
  "parametric")

results <- data.frame(matrix(0, ncol=length(attention_types), nrow=8))

colnames(results) <- attention_types

rmse <- function(obj, y_test)
{
  sqrt(mean((obj$mean - y_test)**2))
}

y <- AirPassengers

y_split <- misc::splitts(y, split_prob = 0.9)

y_train <- y_split$training

y_test <- y_split$testing

h <- length(y_test)

for (attention_type in attention_types)
{
   misc::debug_print(attention_type)
  
   (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::glm.nb, attention = TRUE, attention_type = attention_type)))
  print(obj)
  results[1, attention_type] <- rmse(obj, y_test)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=stats::glm, attention = TRUE, attention_type = attention_type)))
  results[2, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::rlm, attention = TRUE, attention_type = attention_type)))
  results[3, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::lqs, attention = TRUE, attention_type = attention_type)))
  results[4, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj) 
  
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::glm.nb, attention = FALSE, attention_type = attention_type)))
  print(obj)
  results[5, attention_type] <- rmse(obj, y_test)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=stats::glm, attention = FALSE, attention_type = attention_type)))
  results[6, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::rlm, attention = FALSE, attention_type = attention_type)))
  results[7, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::lqs, attention = FALSE, attention_type = attention_type)))
  results[8, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj) 

}

print(results)

results1 <- results
```

```{r fig.width=7.5, eval=FALSE}
library(forecast)
library(ahead)

attention_types <- c("dot_product", "scaled_dot_product", 
  "cosine", "exponential", 
  "gaussian", "linear",
  "value_based", "hybrid", 
  "parametric")

results <- data.frame(matrix(0, ncol=length(attention_types), nrow=8))

colnames(results) <- attention_types

rmse <- function(obj, y_test)
{
  sqrt(mean((obj$mean - y_test)**2))
}

y <- AirPassengers

y_split <- misc::splitts(y, split_prob = 0.9)

y_train <- y_split$training

y_test <- y_split$testing

h <- length(y_test)

for (attention_type in attention_types)
{
   misc::debug_print(attention_type)
  
   (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::glm.nb, attention = TRUE, attention_type = attention_type, attention_method = "historical")))
  print(obj)
  results[1, attention_type] <- rmse(obj, y_test)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=stats::glm, attention = TRUE, attention_type = attention_type, attention_method = "historical")))
  results[2, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::rlm, attention = TRUE, attention_type = attention_type, attention_method = "historical")))
  results[3, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::lqs, attention = TRUE, attention_type = attention_type, attention_method = "historical")))
  results[4, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj) 
  
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::glm.nb, attention = FALSE, attention_type = attention_type, attention_method = "historical")))
  print(obj)
  results[5, attention_type] <- rmse(obj, y_test)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=stats::glm, attention = FALSE, attention_type = attention_type, attention_method = "historical")))
  results[6, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::rlm, attention = FALSE, attention_type = attention_type, attention_method = "historical")))
  results[7, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj)
  
  (obj <- suppressWarnings(ahead::glmthetaf(y_train, h=h, fit_func=MASS::lqs, attention = FALSE, attention_type = attention_type, attention_method = "historical")))
  results[8, attention_type] <- rmse(obj, y_test)
  print(obj)
  plot(obj) 

}

print(results1)
print(results)
```