ft_linear_regression

Final grade

Intro

This project is the first step into AI and Machine Learning under 42 curriculum at Hive Helsinki. Basic machine learning algorithm and program that predicts the price of a car by using a linear regression function with a gradient descent algorithm.

Plot Preview

Requirements

Installed rust is required and the plotting crate also requires following cmake and other dependencies. Everything can be installed with make install and removed with make uninstall. Or manually:

• rust curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
• plotters dependencies sudo apt-get install cmake libfreetype6-dev libfontconfig1-dev xclip

Usage

The basic usage is to run train binary on your data. Then running predict binary supplying previously trained model.

Simple

This is a showcase run, that uses provided dataset data/subject_data.csv. The train program uses 4/5 of the dataset to train the model and remaining 1/5 to test its precision.

Simple make command will run following steps

• tests
• build
• train
• predict

You'll be promped to input a mileage that is then used to estimate a car's value using the pretrained model.

Building

make build

Two bineries are created and located in bin/ directory.

Training

USAGE:
    train [OPTIONS] --file <datafile>

FLAGS:
    -h, --help       Prints help information
    -V, --version    Prints version information

OPTIONS:
    -a, --alpha <alpha>        α - Learning rate 
    -f, --file <datafile>      Input data file
    -i, --iterations <iter>    Number of iterations to run, this will overwrite TD limit
    -o, --out <outfile>        Path to output file (model)
    -r, --ratio <ratio>        Distribution between test and train set ratio
    -s, --seed <seed>          Randomness seed for data splitting to train & test sets
        --stats <stats>        Path to a directory where plots and statistics should be saved
    -t, --tdlimit <tdlimit>    Temporal difference limit (amout of change per iteration). How accurate local minima is.

Predicting

USAGE:
    predict [OPTIONS] --modelfile <model>

FLAGS:
    -h, --help       Prints help information
    -V, --version    Prints version information

OPTIONS:
    -k, --key <key>            Key to use in value estimation, using trained linear regression model.
    -f, --modelfile <model>    Path to trained linear regression model

Testing

make test

Runs cargo test

Documentation

make doc

Compiles and opens documentation