本文為筆者從官網(wǎng)學(xué)習(xí)的代碼實(shí)錄，不對(duì)的地方請(qǐng)多指教克伊！
官網(wǎng)地址：https://www.tidymodels.org/
官網(wǎng)介紹：The tidymodels framework is a collection of packages for modeling and machine learning using tidyverse principles.
本次內(nèi)容為get started部分的pipeline
內(nèi)容目錄如下圖
內(nèi)容目錄
載入需要的包
library(tidymodels) # for the parsnip package, along with the rest of tidymodels
# Helper packages
library(readr) # for importing data
library(broom.mixed) # for converting bayesian models to tidy tibbles
library(dotwhisker) # for visualizing regression results
#讀入測(cè)試數(shù)據(jù)
urchins <-read_csv("https://tidymodels.org/start/models/urchins.csv") %>% #讀入數(shù)據(jù)
setNames(c("food_regime", "initial_volume", "width")) %>% #定義列名稱(chēng)
mutate(food_regime = factor(food_regime, levels = c("Initial", "Low", "High")))#設(shè)定因子變量
glimpse(urchins)
測(cè)試數(shù)據(jù)變量情況
food_regime列為三種不同的喂養(yǎng)策略晰甚，initial_volume為金槍魚(yú)初始的體積，width為金槍魚(yú)最終喂養(yǎng)后的寬度數(shù)據(jù)。研究的主要目的是看看不同的喂養(yǎng)策略對(duì)于金槍魚(yú)最終寬度的影響袁滥。從嘗試可以知道，金槍魚(yú)的初始體積initial_volume也會(huì)影響最終width的結(jié)果灾螃。
#對(duì)數(shù)據(jù)進(jìn)行可視化
ggplot(data = urchins,#數(shù)據(jù)集
aes(x = initial_volume, #全局映射
y = width,
group = food_regime,
color = food_regime)) +
geom_point() + #繪制點(diǎn)圖
geom_smooth(method = lm, se = FALSE) +#繪制平滑曲線
scale_color_viridis_d(option = "plasma", end = .7) #色盲友好顏色
數(shù)據(jù)可視化情況题翻，不同分組對(duì)于最終連續(xù)變量的影響
從圖中可以看出，三組共同的趨勢(shì)為：金槍魚(yú)的初始體積越大腰鬼，最終喂養(yǎng)后的寬度越大嵌赠。不同的喂養(yǎng)策略產(chǎn)生的直線的斜率有點(diǎn)不同
可以看到，因?yàn)楸狙芯康慕Y(jié)局變量為數(shù)值型變量熄赡，所以應(yīng)該用線性回歸模型進(jìn)行擬合分析
選定模型后姜挺，我們還需要對(duì)模型內(nèi)部的engine進(jìn)行選擇，其定義如下：The engine value is often a mash-up of the software that can be used to fit or train the model as well as the estimation method. 個(gè)人認(rèn)為engine的作用主要是確定損失函數(shù)彼硫。
linear_reg()#查看線性回歸默認(rèn)的engine
線性回歸默認(rèn)的engine
linear_reg()的engine可選項(xiàng)目
lm_mod <- linear_reg()#定義需要的模型炊豪，默認(rèn)參數(shù)
lm_fit <- lm_mod %>% #對(duì)模型進(jìn)行擬合
fit(width ~ initial_volume * food_regime, data = urchins)
tidy(lm_fit)#查看模型
模型擬合結(jié)果
#以下對(duì)于估計(jì)值及標(biāo)準(zhǔn)誤進(jìn)行可視化
tidy(lm_fit) %>%
dwplot(dot_args = list(size = 2, color = "black"),
whisker_args = list(color = "black"),
vline = geom_vline(xintercept = 0, colour = "grey50", linetype = 2))+
可視化結(jié)果
#構(gòu)建測(cè)試數(shù)據(jù)
new_points <- expand.grid(initial_volume = 20,
food_regime = c("Initial", "Low", "High"))
new_points
測(cè)試數(shù)據(jù)
#進(jìn)行點(diǎn)數(shù)據(jù)預(yù)測(cè)
m%ean_pred <- predict(lm_fit, new_data = new_points)
mean_pred
均值估計(jì)結(jié)果
#95%CI估計(jì)
conf_int_pred <- predict(lm_fit,
new_data = new_points,
type = "conf_int")
conf_int_pred
95%CI估計(jì)
#構(gòu)建可視化需要的數(shù)據(jù)集
plot_data <-
new_points %>%
bind_cols(mean_pred) %>%
bind_cols(conf_int_pred)
plot_data
可視化需要的數(shù)據(jù)集
#畫(huà)圖
ggplot(plot_data, aes(x = food_regime)) +
geom_point(aes(y = .pred)) +
geom_errorbar(aes(ymin = .pred_lower,
ymax = .pred_upper),
width = .2) +
labs(y = "urchin size")
可視化預(yù)測(cè)結(jié)果
利用其他engine進(jìn)行數(shù)據(jù)擬合及分析
#以下利用貝葉斯模型進(jìn)行數(shù)據(jù)分析擬合
# 設(shè)定數(shù)據(jù)的先驗(yàn)分布，這是后面貝葉斯engine的參數(shù)
prior_dist <- rstanarm::student_t(df = 1)
#設(shè)定種子數(shù)
set.seed(123)
# 定義模型
bayes_mod <-
linear_reg() %>%
set_engine("stan",
prior_intercept = prior_dist,
prior = prior_dist)
# 訓(xùn)練模型
bayes_fit <-
bayes_mod %>%
fit(width ~ initial_volume * food_regime, data = urchins)
print(bayes_fit, digits = 5)#展示模型
貝葉斯模型擬合結(jié)果
tidy(bayes_fit, conf.int = TRUE)
整潔展示
#貝葉斯模型進(jìn)行可視化
bayes_plot_data <-
new_points %>%
bind_cols(predict(bayes_fit, new_data = new_points)) %>%
bind_cols(predict(bayes_fit, new_data = new_points, type = "conf_int"))
ggplot(bayes_plot_data, aes(x = food_regime)) +
geom_point(aes(y = .pred)) +
geom_errorbar(aes(ymin = .pred_lower, ymax = .pred_upper), width = .2) +
labs(y = "urchin size") +
ggtitle("Bayesian model with t(1) prior distribution")
貝葉斯模型結(jié)果
以下網(wǎng)址為tidymodels包提供的可以擬合的模型和engine
https://www.tidymodels.org/find/parsnip/
