[1] "versão 4"Linguagem R
Além do #rstats
Thiago Pires
Cientista de Dados (CIO/IBM)
Formação:
- Estatística (UERJ)
- MSc. Epidemiologia (ENSP/FIOCRUZ)
- DSc. Engenharia Biomédica (COPPE/UFRJ)
Linguagem R
- A linguagem R foi derivada da linguagem S.
- Foi lançada em 1993 pelos dois estatísticos Ross Ihaka e Robert Gentleman da Universidade de Auckland, Nova Zelândia
- Atualmente na versão 4.2
Linguagem R
Sintaxe
Tidyverse e RStudio
Tidyverseé uma coleção de pacotes desenvolvidos pela RStudio e projetados para ciência de dados. Todos estes pacotes tentam compartilhar uma mesma sintaxe e estrutura de dados.
ggplot2:: utilizado na visualização de dados
Percentual de mortos e vivos, segundo sexo e classe
library(ggplot2)
titanic::titanic_train |>
ggplot() +
aes(Sex, ..count../sum(..count..),
group = Survived,
fill = as.factor(Survived)) +
geom_bar(position = "fill") +
facet_grid(~Pclass) +
scale_y_continuous(labels = scales::percent) +
scale_fill_manual(values = c("#003f9a", "#5099f4"),
labels = c("não", "sim")) +
labs(x = "Sexo", y = "", fill = "Sobreviveu") +
theme_light() +
theme(text = element_text(size = 24))
stringr:: e forcats::
ggplot2:: na visualização de casos de COVID 19
covid19 |>
dplyr::mutate(moving_average = zoo::rollmean(value, 7, align = "right", fill = NA)) |>
ggplot2::ggplot() +
ggplot2::aes(date, moving_average) +
ggplot2::geom_line() +
ggplot2::theme_minimal() +
ggplot2::labs(x = "Data",
y = "Média móvel casos COVID 19") +
theme(text = element_text(size = 22))
tidymodels:: visualizando o modelo
newdata <-
expand.grid(Pclass = c("1st", "2nd", "3rd"),
Sex = c("male", "female"))
pihat <-
(lr_fit |> predict(newdata, type = "prob")) |>
pull(.pred_sim)
newdata |> mutate(Pihat = pihat) |>
ggplot(aes(Sex, Pihat,
group = Pclass,
colour = Pclass)) +
geom_line() + geom_point() +
labs(x = "Sex", y = expression(pi(Survived == sim)),
colour = "Ticket Class") +
theme_minimal() +
theme(text = element_text(size = 22))
gtrendsR:: Google trends
dataStocks <- function(code, from, to) {
stock <-
quantmod::getSymbols(code,
src = "yahoo",
from = from,
to = to,
auto.assign = FALSE)
stock <-
stock |>
as.data.frame() |>
tibble::rownames_to_column(var = "Date") |>
dplyr::as_tibble() |>
dplyr::select(dplyr::matches("Date|Adjusted")) |>
dplyr::mutate(Date = Date |> lubridate::as_date()) |>
dplyr::rename(Adjusted = 2)
return(stock)
}
dataGtrends <- function(query, from, to) {
data_gtrends <-
gtrendsR::gtrends(keyword = c(query),
geo = "BR",
time = glue::glue("{from} {to}"),
onlyInterest = TRUE)
data_gtrends$interest_over_time |>
dplyr::as_tibble() |>
dplyr::mutate(Date = date |> lubridate::as_date()) |>
dplyr::select(Date, Hits = hits)
}
plotStocksGtrends <- function(stock, query, from, to) {
data_stocks <- dataStocks(stock, from, to)
data_gtrends <- dataGtrends(query, from, to)
coeff <- 50/data_stocks$Adjusted |> mean(na.rm = TRUE)
data_stocks |>
dplyr::mutate(Week = Date |> lubridate::week()) |>
dplyr::left_join(
data_gtrends |>
dplyr::mutate(Week = Date |> lubridate::week()) |>
dplyr::select(Week, Hits), by = "Week"
) |>
ggplot2::ggplot(ggplot2::aes(x = Date)) +
ggplot2::geom_smooth(ggplot2::aes(y = Adjusted, colour = "Stock Price")) +
ggplot2::geom_line(ggplot2::aes(y = Adjusted), colour = "gray") +
ggplot2::geom_smooth(ggplot2::aes(y = Hits / coeff, colour = "Google Trends")) +
ggplot2::scale_y_continuous(
name = "Adjusted Price",
sec.axis = ggplot2::sec_axis(~.*coeff, name = "Hits")) +
ggplot2::theme_minimal() +
ggplot2::labs(colour = "") +
ggplot2::theme(text = element_text(size = 20))
}
plotStocksGtrends("PETR4.SA", "Petrobras", "2020-10-05", "2021-10-05")
shiny:: app para obter preços de ações
library(shiny)
library(shinythemes)
source("dataStocks.R")
start_value <- Sys.Date() - 31
end_value <- Sys.Date()
ui <- fluidPage(
theme = shinytheme("sandstone"),
navbarPage("STOCK PRICE"),
sidebarLayout(
sidebarPanel(
textInput("code", "Stock code", value = "PETR4.SA"),
dateRangeInput("date", "Date",
start = start_value,
end = end_value)
),
mainPanel(
plotly::plotlyOutput("plot")
)
)
)
server <- function(input, output, session) {
data_stock <- reactive({
validate(
need(input$code, "Please type a stock code!"),
need(input$date[1], "Please choose a start date!"),
need(input$date[2], "Please choose a end date!"))
dataStocks(input$code, input$date[1], input$date[2])
})
output$plot <- plotly::renderPlotly({
plotly::plot_ly(data_stock(),
x = ~Date,
y = ~Adjusted,
color = "orange",
mode = "lines",
type = "scatter") |>
plotly::rangeslider()
})
}
shinyApp(ui, server)rayshader:: mapas em 3D
source("../src/define_image_size.R")
source("../src/get_arcgis_map_image.R")
image_size <- define_image_size(bbox, major_dim = 600)
# Fetch overlay image
overlay_file <- "../data/rj_overlay.png"
get_arcgis_map_image(bbox, map_type = "World_Imagery", file = overlay_file,
width = image_size$width, height = image_size$height,
sr_bbox = 4326)
overlay_img <- png::readPNG(overlay_file)
# Calculate rayshader layers
watermap <- rayshader::detect_water(elev_matrix)
# Plot 2D with texture
elev_matrix |>
rayshader::sphere_shade(texture = "imhof4") |>
rayshader::add_water(watermap, color = "imhof4") |>
rayshader::add_overlay(overlay_img, alphalayer = 0.5) |>
rayshader::plot_map()
rayshader:: mapas em 3D
# Plot 3D
zscale <- 30
rgl::clear3d()
elev_matrix |>
rayshader::sphere_shade(texture = "imhof4") |>
rayshader::add_water(watermap, color = "imhof4") |>
rayshader::add_overlay(overlay_img, alphalayer = .8) |>
plot_3d(elev_matrix, zscale = zscale, windowsize = c(1500, 1200),
water = TRUE, soliddepth = -max(elev_matrix)/zscale, wateralpha = 1,
theta = 25, phi = 30, zoom = 0.65, fov = 60)
rayshader::render_snapshot("../data/rj-3d.png")
tidytext:: e wordcloud::
data_decoded_review <- readRDS("../data/data_decoded_review.RDS")
set.seed(1984)
data_decoded_review |>
tidyr::unnest(words) |>
dplyr::anti_join(tidytext::stop_words, by = c("words" = "word")) |>
dplyr::filter(! words %in% c("?", "br", "1", "2",
"3", "4", "7", "8",
"9", "10")) |>
dplyr::count(words, y, sort = TRUE) |>
dplyr::mutate(y = ifelse(y == 0, "Negativo", "Positivo")) |>
reshape2::acast(words ~ y, value.var = "n", fill = 0) |>
wordcloud::comparison.cloud(colors = c("#E24664", "#1D6CE6"),
max.words = 200,
title.size = 4,
scale = c(5, 2))
Vias como linhas baseado no osmdata
if (!exists("has_internet_via_proxy")) {
assign("has_internet_via_proxy", TRUE, environment(curl::has_internet))
}
# Consulta
query <- osmdata::opq(bbox = unlist(bbox)) |>
osmdata::add_osm_feature(key = "highway") |>
osmdata::osmdata_sf()
# Retornando linhas
query_lines <- query$osm_lines
# Visualizando o objeto
query_lines |>
sf::st_geometry() |>
ggplot2::ggplot() +
ggplot2::geom_sf(colour = "gray") +
ggplot2::theme_minimal() +
ggplot2::theme(text = element_text(size = 15))
Convertendo as vias em um objeto sfnetwork
poly_to_lines <- sf::st_cast(query$osm_polygons, "LINESTRING")
query_lines <- dplyr::bind_rows(query_lines, poly_to_lines)
highway_net <- sfnetworks::as_sfnetwork(query_lines, directed = FALSE)
# Remove nodes that have only two edges connected
highway_simple <- tidygraph::convert(highway_net, sfnetworks::to_spatial_smooth)
# Filter connected components
connected_net <- highway_simple |>
tidygraph::activate("nodes") |>
dplyr::filter(tidygraph::group_components() == 1)
# Plot
ggplot2::ggplot() +
ggplot2::geom_sf(data = sf::st_as_sf(connected_net, "edges"), col = "gray") +
ggplot2::geom_sf(data = sf::st_as_sf(connected_net, "nodes"), col = "gray") +
ggplot2::theme_minimal() +
ggplot2::theme(text = element_text(size = 15))
Problema de otimização de rota de frota
São 3 veículos e preciso selecionar a melhor rota. Eles devem sair e retornar a um depósito.
set.seed(711)
index_visits <- sample(1:603, 9)
visits <- connected_net |>
tidygraph::activate("nodes") |>
dplyr::slice(index_visits)
depot <- connected_net |>
tidygraph::activate("nodes") |>
dplyr::slice(23)
ggplot2::ggplot() +
ggplot2::geom_sf(data = sf::st_as_sf(connected_net, "edges"), col = "gray") +
ggplot2::geom_sf(data = sf::st_as_sf(connected_net, "nodes"), col = "gray") +
ggplot2::theme_minimal() +
ggplot2::theme(text = element_text(size = 15)) +
ggplot2::geom_sf(data = sf::st_as_sf(visits, "nodes"),
ggplot2::aes(col = "Visits"), size = 6) +
ggplot2::geom_sf(data = sf::st_as_sf(depot, "nodes"),
ggplot2::aes(col = "Depot"), size = 6) +
ggplot2::labs(x = "", y = "", colour = "") +
ggplot2::scale_color_manual(values = c("Visits" = "blue",
"Depot" = "orange"))
Projetos pessoais com R
- Autenticação de aplicações shiny com App ID: github.com/IBM/AppIdR
- Autenticação de aplicações shiny com W3ID: github.ibm.com/Thiago-Pires/w3idr
- Carbon Design System em Shiny apps: github.com/th1460/shiny-carbon-web
