Final Page

Goal

This is a generalized ML assignment

1. Predict the profit of future products developed at CravenSpeed
2. Using any model (or ensemble) you’d like
3. Evaluated using RMSE on holdout sample (you won’t see)

Submission

• A maximum of 15 slides, without any code, that demonstrates:
  • How and why you created/selected the features used,
  • The choice and design of your model, and
  • Results and insights.
• You must present from the Ford 102 “Teaching Machine” with no login from a public url.

Presentations

• It is a simple matter to create a presentation within Quarto.
• Simply specify “revealjs” format. Read more

The page

final_page.qmd

---
title: "Final Page"
author: "Team $i$"
date: "04/21/2025"
---

# Goal

...

The presentation

final_present.qmd

---
title: "Final Presentation"
author: "Team $i$"
date: "04/21/2025"
format: revealjs
---

# Goal

...

Criteria

• Every group member must participate in the presentation
• Maximum 10 features including interactions

Setup

• You may use any libraries, but tidyverse and caret may be sufficient.
  • If you wish, you may use Python, Julia, or Observable in any manner you see fit and I will figure out how to assess it.
  • Recall - no code on slides! So it won’t matter.

library(tidyverse)
library(caret)

Dataframe

• We use the craven_train dataframe.

fast <- readRDS(gzcon(url("https://github.com/cd-public/D505/raw/refs/heads/master/dat/craven_train.rds")))

• You will necessarily perform some feature engineering as you see fit.
  • Exactly ten (10) features.
• No relation to “Crazy Train”

An “engineer” function

• Besides the presentation
  • Submit a .qmd or .rmd file that includes an “engineer” function
  • It engineers your features over a data frame with the same columns as “craven_train.rds”.

fast <- readRDS("secret.rds") # I have "secret" data
fast <- fast %>% engineer() # I will apply your function.

A bad example

# Engineer 10 features
engineer <- function(df) {
    df |> select(1:10)
}

Setup

• Assessments will be setup as follows:
  • “Profit” is engineered.
  • Note that the first five (10) features are selected.
  • This ensures no more than 10 features are used.
  • “Profit” is incorporated into the data frame.

profit <- fast["Revenue 2019 to present"] - fast["BOM Cost"] * fast["Units Sold"]
fast <- fast %>% engineer()
fast <- fast |> select(1:10) # Max 10 features
fast["Profit"] = profit

Assessment

• Assessments will be evaluated via RMSE over the secret data as follows:

train(Profit ~ .,
      data = fast, 
      method = "lm",
      trControl = trainControl(method = "cv", number = 5))$results$RMSE