---
title: "SpeedML"
author: "Jia-Shen Tsai, Wendy Wen, Zhengqi Jiao, Miaojun Pang, Alexander Yoshizumi"
date: "2024-03-02"
last updated: "2024-03-12"
output: html_document
---
# Speed ML V1
#### Author: Jia-Shen Tsai, Wendy Wen, Zhengqi Jiao, Miaojun Pang, Alexander Yoshizumi
When run, the script create the random forest model and results for the given dataset.
```{r setup, message = FALSE, warning = FALSE}
rm(list=ls())
library(ggplot2)
library(dplyr)
library(randomForest)
# devtools::install_github("MI2DataLab/randomForestExplainer")
library(randomForestExplainer)
```
```{r reading files, message = FALSE, warning = FALSE}
df_all <- read.csv("../01_Data/02_Processed/Speed_Model_R.csv")
df_east <- read.csv("../01_Data/02_Processed/Speed_Model_East.csv")
df_west <- read.csv("../01_Data/02_Processed/Speed_Model_West.csv")
str(df_all)
dim(df_east)
dim(df_west)
```
```{r create a forest, message = FALSE, warning = FALSE}
set.seed(2025)
forest <- randomForest(Speed_mph ~ ., data = df_all, localImp = TRUE)
```
```{r calling forest, warning = FALSE}
forest
```
# Distribution of minimal depth
```{r Create Minimal depth for each variable, warning = FALSE}
# min_depth_frame <- min_depth_distribution(forest)
# save(min_depth_frame, file = '../01_Data/02_Processed/randomForest/min_depth_fram.rda')
load('../01_Data/02_Processed/randomForest/min_depth_fram.rda')
head(min_depth_frame, n=10)
```
```{r Minimal Depth Distribution Plot , warning = FALSE, fig.width=10}
plot_min_depth_distribution(min_depth_frame)
```
```{r Minimal Depth Distribution Plot without missing values, warning = FALSE, fig.width=10}
plot_min_depth_distribution(min_depth_frame, mean_sample = "relevant_trees", k = 15)
```
# Various variable importance measures
```{r Importance Frame, warning = FALSE}
#importance_frame <- measure_importance(forest)
#save(importance_frame, file = "../01_Data/02_Processed/randomForest/importance_frame.rda")
load("../01_Data/02_Processed/randomForest/importance_frame.rda")
importance_frame
```
```{r Multi-Way Importance Plot, warning = FALSE}
# plot_multi_way_importance(forest, size_measure = "no_of_nodes") # gives the same result as below but takes longer
plot_multi_way_importance(importance_frame, size_measure = "no_of_nodes")
```
```{r Multi-way importance plot with p-value, warning = FALSE}
plot_multi_way_importance(importance_frame, x_measure = "mse_increase", y_measure = "node_purity_increase", size_measure = "p_value", no_of_labels = 5)
```
## Compare measures using ggpairs
```{r Relations between measures of importance, warning = FALSE}
# plot_importance_ggpairs(forest) # gives the same result as below but takes longer
plot_importance_ggpairs(importance_frame)
```
## Compare different rankings
```{r Relations between rankings according to different measures, message = FALSE, warning = FALSE}
# plot_importance_rankings(forest) # gives the same result as below but takes longer
plot_importance_rankings(importance_frame)
```
# Variable Interactions
## Conditional minimal depth
```{r Extract the names of 5 most important variables according to both the mean minimal depth and number of trees, warning = FALSE}
# (vars <- important_variables(forest, k = 5, measures = c("mean_min_depth", "no_of_trees"))) # gives the same result as below but takes longer
(vars <- important_variables(importance_frame, k = 5, measures = c("mean_min_depth", "no_of_trees")))
```
[1] "Curve_Max_Following_250ft_degrees"
[2] "Curve_Max_Preceding_250ft_degrees"
[3] "Distance_to_Key_Nodes_FOL"
[4] "Distance_to_Key_Nodes_PRE"
[5] "Elevation_Smoothen_ft"
[6] "Station"
```{r Interaction frame with interactions, warning = FALSE}
# interactions_frame <- min_depth_interactions(forest, vars)
# save(interactions_frame, file = "../01_Data/02_Processed/randomForest/interactions_frame.rda")
load("../01_Data/02_Processed/randomForest/interactions_frame.rda")
head(interactions_frame[order(interactions_frame$occurrences, decreasing = TRUE), ])
```
```{r Mean minimal depth for 30 most frequent interactions, warning = FALSE, fig.width=10}
# plot_min_depth_interactions(forest) # calculates the interactions_frame for default settings so may give different results than the function below depending on our settings and takes more time
plot_min_depth_interactions(interactions_frame)
```
```{r Mean minimal depth for 30 most frequent interactions without missing values, warning = FALSE, fig.width=10}
#interactions_frame <- min_depth_interactions(forest, vars, mean_sample = "relevant_trees", uncond_mean_sample = "relevant_trees")
# save(interactions_frame, file = "../01_Data/02_Processed/randomForest/interactions_frame_relevant.rda")
load("../01_Data/02_Processed/randomForest/interactions_frame_relevant.rda")
plot_min_depth_interactions(interactions_frame)
```
## Predict of the forest on a grid
```{r Prediction of the forest by interactions, warning = FALSE}
plot_predict_interaction(forest, df_all, "Station", "Distance_to_Key_Nodes_FOL")
```
Sign in with Wallet
Connect another wallet