ejercicio-de-modelaje.qmd

---
title: "Ejercicio de Modelaje"
format: html
editor: visual
---

## Conectar los datos

```{r}
if(!require(googlesheets4)){
  install.packages("googlesheets4")
  require(googlesheets4)
}
library(googledrive)
library(magick)

my_files <- dir("R",full.names = TRUE)
for(my_file in my_files){
  source(my_file)
}

gs4_deauth()

my_url <- "https://docs.google.com/spreadsheets/d/1SF5DDTEfbQFSNjvYKQnsQKdsBdE7oJeTywCk92pB7f8/edit?usp=sharing"

my_url2 <- 'https://docs.google.com/spreadsheets/d/1SF5DDTEfbQFSNjvYKQnsQKdsBdE7oJeTywCk92pB7f8/edit#gid=1964221805'

my_data <- googlesheets4::read_sheet(my_url2)
```

## Acceder los links

De prueba:

```{r}
dir.create("tmp2")
setwd("tmp2")
tmp_file <- drive_download(my_data$URL[2],overwrite = TRUE)
my_image <- image_read(path =tmp_file$local_path)
my_df <- to_RGB_df(my_image)
my_df %>% 
summarize(
  Red = mean(R),
  Green = mean(G),
  Blue = mean(B),
  Red_sd = sd(R),
  Green_sd = sd(G),
  Blue_sd = sd(B),
  )

```

```{r}
setwd("tmp2")
my_data <- my_data %>% 
  filter(!is.na(URL))
my_new_data <- data.frame()
index <- 1
for(URL in my_data$URL){
  index <- index +1
  
  cat(index, " Analizando: ", URL, "\n")
  tmp_file <- drive_download(URL, overwrite = TRUE)
  my_image <- image_read(path =tmp_file$local_path)
  my_df <- to_RGB_df(my_image)
  my_df %>% 
    summarize(
      Red = mean(R),
      Green = mean(G),
      Blue = mean(B),
      Red_sd = sd(R),
      Green_sd = sd(G),
      Blue_sd = sd(B),
    ) -> tmp
  my_new_data <- rbind(my_new_data, tmp)
}

```

```{r}
my_data2 <- cbind(my_data,my_new_data)
```

```{r}
my_data2
```

# Hacer backup

```{r}
write_rds(my_data2,"./data/fotos_con_analysis_de_colores.rds")
```

# Modelar.

```{r}
my_data2 %>% 
  mutate(Cromatica = as.factor(Cromatica)) %>% 
  glm(Cromatica ~ (Red + Green + Blue)^2,family = "binomial", data=.)-> my_model1
```

```{r}
my_data2 %>% 
  mutate(Cromatica = as.factor(Cromatica)) %>% 
  glm(Cromatica ~ Red + Green + Blue,family = "binomial", data=.)-> my_model2
summary(my_model2)
```

```{r}
predict(my_model2,type = "response") -> porcentajes
```

```{r}
my_data2$prediccion <- porcentajes
```

# Evaluar

```{r}
my_data2 <- my_data2 %>% 
  mutate(pred_dura = ifelse(prediccion>.5, "Frío", "Cálido"))
```

```{r}
sum(my_data2$Cromatica == my_data2$pred_dura)/nrow(my_data2)
```

```{r}
my_data3 <- my_data2
predict(my_model1,type = "response") -> porcentajes

my_data3$prediccion <- porcentajes

my_data3 <- my_data3 %>% 
  mutate(pred_dura = ifelse(prediccion>.5, "Frío", "Cálido"))

sum(my_data3$Cromatica == my_data3$pred_dura)/nrow(my_data3)

```

## Mas interacciones

```{r}
my_data4 <- my_data2
my_data4 %>% 
  mutate(Cromatica = as.factor(Cromatica)) %>% 
  glm(Cromatica ~ (Red + Green + Blue)^3,family = "binomial", data=.)-> my_model4
summary(my_model4)
```

```{r}
predict(my_model4,type = "response") -> porcentajes

my_data4$prediccion <- porcentajes

my_data4 <- my_data4 %>% 
  mutate(pred_dura = ifelse(prediccion>.5, "Frío", "Cálido"))

sum(my_data4$Cromatica == my_data4$pred_dura)/nrow(my_data4)
```

```{r}
coef(my_model4)
```

\$\$

\hat{Y} \sim R + G + B + \epsilon \$\$

\$\$ \hat{Y} = \[Calido\|Frio\] \$\$

\$\$

Calido\|Frio \sim R + G + B \$\$