FAO_models.Rmd

---
title: "wfp-clean"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(tidyverse)
library(readr)
library(broom)
```

```{r}
devtools::load_all("/home/agar2/Documents/FAOSTATpackage/FAOSTAT")
fao <- FAOsearch(full=T)
```
3,5,48,57,74

BC=3=crop_stock
CC=5=crop_supply
OA=48=crop_production
QC=57=population
TP=74=crop_trade

```{r, "https://github.com/muuankarski/faobulk/blob/master/R/get_data.R"}
get_data <- function(DatasetCode = "QP"){

  datas <- fao
  urli <- datas[datas$DatasetCode == DatasetCode,]$FileLocation
  fly <- tempfile(fileext = ".zip")
  download.file(url = urli, destfile = fly)
  dat <- read_csv(fly)
  names(dat) <- tolower(sub(" ", "_", names(dat)))
  return(dat)
}
```

```{r}
population <- get_data("OA")[,c(2,4,6,8,9,10)]

crop_stock <- get_data("BC")[,c(2,4,6,8,9,10)]
crop_supply <- get_data("CC")[,c(2,4,6,8,9,10)]
crop_production <- get_data("QC")[,c(2,4,6,8,9,10)]
crop_trade <- get_data("TP")[,c(2,4,6,8,9,10)]
```


```{r}
crop_total <- rbind(crop_production, crop_stock, crop_supply, crop_trade)
rm(crop_production, crop_stock, crop_supply, crop_trade)
crop_total <- crop_total[!crop_total$value==0,]
crop_total <- crop_total[complete.cases(crop_total),]
write.csv(crop_total, "crop_total.csv")
```

```{r}
crop_total <- read_csv("crop_total.csv", )
crop_total <- crop_total[,-1]
crop_unique <- apply(crop_total[,c(1:5)], 2, unique)
```

# Create nested models - element, area, item: year vs (value == unit)
```{r}
# Create nested tibble
crop_production <- crop_total %>% filter(element == "Production") %>% select(-element) %>% group_by(area, item, unit) %>% nest()
# Create model and Add coefficient from models
crop_model <- crop_production[c(1:10),] %>% mutate(models = lapply(data, function(df) lm(value ~ year, data = df))) %>% mutate(coef=unlist(map(models, coef))['year'])


top <- crop_production$item %>% table() %>% sort(decreasing = T)
barplot(top[top > 240])

crop_model_fruit <- crop_production %>% filter(item == "Fruit Primary") %>% mutate(models = lapply(data, function(df) lm(value ~ year, data = df))) %>% mutate(coef=unlist(map(models, coef))['year']) %>% arrange(desc(coef))

crop_production %>% filter(item == "Fruit Primary") %>% unnest() %>% pivot_wider(names_from = year, values_from = value)

crop_production %>% filter(item == "Fruit Primary") %>% unnest() %>% ggplot(aes(year, value, colour = area)) + geom_line() 


# plot growth rate vs year of country per filter(item) 
crop_model_fruit$data[[1]]
plot(data.frame(crop_model_fruit$data[[1]]$year, crop_model_fruit$data[[1]]$value / min(crop_model_fruit$data[[1]]$value)))
data.frame(crop_model_fruit$data[[1]]$year, crop_model_fruit$data[[1]]$value / min(crop_model_fruit$data[[1]]$value))

plot(data.frame(crop_model_fruit$data[[1]]$year, scale(crop_model_fruit$data[[1]]$value)))

scale(crop_model_fruit$data[[1]]$value)


```
Create shiny dashboard

    

```{r}
# Extracted coefficient
data.frame(crop_model, "coef"= unlist(unname(as.data.frame(map(crop_model$models, coef))['year',])))

```