Repository which is an attempt to solve the Titanic ML competition with Rust.
You need an environment that can build LightGBM for lightgbm-rs.
# linux
apt install -y cmake libclang-dev libc++-dev gcc-multilib
# OS X
brew install cmake libomp
Download titanic data from kaggle, or use Kaggle API to download the data as below.
$ kaggle competitions download -c titanic -p .
$ mkdir -p input
$ unzip titanic.zip -d input/titanic
$ rm titanic.zip
Install cargo-make as a task runner.
cargo install --force cargo-make
$ cd crates/expXXX && cargo run
If you installed cargo-make, you can execute as below.
$ cargo make run expXXX
Fold 1
Accuracy: 0.8435754189944135
Fold 2
Accuracy: 0.8324022346368715
Fold 3
Accuracy: 0.8426966292134831
Fold 4
Accuracy: 0.8539325842696629
Fold 5
Accuracy: 0.751412429378531
CV Accuracy: 0.8249158249158249
Feature names: ["pclass", "sex", "age", "sibsp", "parch", "fare", "embarked", "title", "family_size"]
Feature importances: [108.4, 57.0, 697.6, 43.6, 30.4, 819.2, 96.8, 118.8, 93.6]
Public Score: 0.75119
See the output of cargo make --list-all-steps.
cargo make --list-all-steps
[cargo-make] INFO - cargo make 0.36.3
[cargo-make] INFO - Build File: Makefile.toml
[cargo-make] INFO - Task: default
[cargo-make] INFO - Profile: development
Build
----------
build - cargo build
run - cargo run
Development
----------
clean - cargo clean
format - cargo fmt
Hooks
----------
init - Initial task to validate the number of arguments
Test
----------
check - cargo check
clippy - cargo clippy
coverage - cargo llvm-cov
coverage-html - cargo llvm-cov --html
coverage-open - open coverage report
test - cargo nextest