anthonymorast / cppart

A parallel implementation of CART algorithm with delayed split selection using C++.

Anthony Morast

About Me

• 💼 I am C++ Software Engineer at Sterling Trading Tech
• 📖 I have a blog on my personal website and Medium
• 📗 I'm reading An Introduction to Mathematical Finance with Applications by Arlie O. Petters and Xiaying Dong.
• 💬 Ich lerne Deutsch.
• 👷 I'm currently working on:
  • A fluid dynamics simulation playground in Rust using the Lattice Boltzmann method
  • Improving my website by working through the submitted bugs
  • Option pricing via binomial models in Rust

Education

• 📃 Applied and Compuational Mathematics, B.S.
• 📃 Computer Science, B.S.
• 📃 Computational Sciences and Robotics, M.S.

---

Things I Really Like to Use

Some More Things I Can Use

Connect With Me

📧 Email 💻 Consulting

---

cppart's Issues

Verify Gini implementation.

lmean/rmean calculations causing differing directions

lmean and rmean in giniSplit in rpart are using 1-nclasses for their mult, in cppart we used the class number

for wine this is 2 4 5 6 and 7 or something like that, likely the cause for splits going opposite of rpart occasionally.

Redo PCA

Only do PCA on the training data (after centering the data [demeaning])
- first check if using the corerlation matrix centers the data around the origin

Translate the Training data with the principal components

Switch delayed to only use improvement

High improvement <=> low SS for anova, switch to only using the improvement for the delayed instead of SS (saves lots of computation) (rpart only uses improvement).

Same for gini, just need to find highest improvement.

Implement Gini Logic

For classification problems use Classification Error

How does rpart determine if it's classification or regression?

Verify prediction errors with Rpart

Run test data down some trees.