soleinik / test-ml

Rust, ML, trading...

A Machine Learning test project

About

The project's goal is to implement start to end file load with polars and use linfa crates to learn and predict financial time series data, with ndarray in between. Project is build from two parts CLI and library. For now data is in local folder that shoud be substituted, in future, with a financial data fetching API (i.e. yahoo_finance_api)

Status

Work-in-progress. At this moment only (list what is implemented)

Currently, only the following Linfa's algorithms implemented

• decision tree (random forest???), flags -Ccdr

• SVM (upport vector machines), flags -Ccss

  works, that can be invoked as follows

$ ./target/debug/ml-cli -Ccdr -h
[r]andom forest learning algorithm

Usage: ml-cli AlgoCategory SL DT {RF|-r}

Options:
  -h, --help  Print help (see more with '--help')

actual run

./target/debug/ml-cli -Ccdr
accuracy:1
predicted:[1, 1, 1]

Progam will use time series data from IBM, 5 years as base, S&P500, 5 years as suplementary and attempt to learn and predict closing price trend for next 3 days.

NOTE: the goal is not actually make correct prediction, rather understand polars, ndarray and linfa interaction

Compiling from source code (via git)

test-ml can be compiled directly from the git repository source code using the following method.

$ git clone https://github.com/soleinik/test-ml.git && cd test-ml
[...]
$ cargo build --release

If successful, this will produce a ml-cli executable located at ./target/release/ml-cli

Executable (for now) expects ./data folder to be available relative to execution root .

Running:

Cargo

$ cargo run -- -h

Demo CLI wrapper

Usage: ml-cli [OPTIONS] <COMMAND>

Commands:
  AlgoCategory, -C  Algorithm's [C]ategory
  help              Print this message or the help of the given subcommand(s)

Options:
  -v, --verbosity...  log level
  -h, --help          Print help (see more with '--help')
  -V, --version       Print version

and follow help prompts... help is context sencitive. The above snippet demonstrates generic help prompt. The following snippet demonstrates help items for Algorithm's category

$ cargo run -- -C -h
Algorithm's [C]ategory

Usage: ml-cli {AlgoCategory|-C} <COMMAND>

Commands:
  SL, -c             [s]upervised learning algorithms class
  USL, -u            [u]nSupervised learning algorithm class
  partial-fit, -f    Partial [f]it ML algorithms class
  pre-processor, -p  [p]re Processors ML algorithms class
  help               Print this message or the help of the given subcommand(s)

Options:
  -h, --help  Print help

Executable ml-cli

When successfuly built, executable will be located in either ./target/debug or ./target/release folder, depends on --debug(default) or --release flag passed to cargo build. Move executable ml-cli together with ./data/* forder into same location. After that invoke it as follows

$ ./ml-cli -h
Demo CLI wrapper

Usage: ml-cli [OPTIONS] <COMMAND>

Commands:
  AlgoCategory, -C  Algorithm's [C]ategory
  help              Print this message or the help of the given subcommand(s)

Options:
  -v, --verbosity...  log level
  -h, --help          Print help (see more with '--help')
  -V, --version       Print version

Development pre-requisites

To build & run this project you need to have the following installed on your system:

• Install Rust (latest stable) – How to install Rust
• Install rustfmt: rustup component add rustfmt
• Install clippy: rustup component add clippy

Format checking (rustfmt)

Make sure your code is well-formatted by running:

cargo fmt

Lint (clippy)

Lint your code (i.e. check it for common issues) with:

cargo clippy