Objectives

The aim of this project is to practice best practices in data science workflows as well as some newly-acquired supervised machine learning techniques.

In this project, I implement the k-nearest neighbours algorithm and the decision tree algorithm on the Wine Data Set. In the end, I get the accuracy score of both algorithms when predicting the type of wine for a new input.

I present the accuracy scores as percentages. I am interested in whether there will be a dramatic difference in the accuracy scores of these algorithms and if there is, which one will be higher.

Data

The data used in this project is from UC Irvine Machine Learning Repository. It consists of 178 observations, 13 attributes and 3 classes of wine.

The data is also available in current repository as wine_data.csv.

System Requirements

• Python 3.6 and packages:
  • scikit-learn==0.18.1
  • pandas==0.20.1
  • numpy==1.12.1
  • argparse==1.4.0
  • matplotlib==2.0.2

Dependency Diagram

Reproducing the Analysis

Clone this repository or download it. Then, cd to the project directory on your computer. The project directory already has the intermediate files of the analysis that has been run before. In order to clean the analysis and re-run it, first, run make clean.

You can use two options to run the analysis. The directions for both are explained as follows:

Using conda environment:

Run the command below.

conda env create -f environment.yml

This will create the python environment required for the analysis. Then run the command below to carry out the analysis from top to bottom.

make all

Using docker image:

If you have Docker installed on your computer, you can run the command below that will tell automatically download/pull the Docker image required for this analysis. Don't forget to modify the VOLUME_ON_YOUR_COMPUTER part with the appropriate path to the project directory on your computer.

docker run --rm -it -v VOLUME_ON_YOUR_COMPUTER:/home/wine_classification nazliozum/wine_classification /bin/bash

Now, your prompt should change to look something like this:

root@1fc309a08883:/#

Then cd into the /home/wine_classification directory and then run make all.

Now, the whole analysis will run from top to bottom.

You can use the command exit to exit the container and go back to your regular Shell.

Author

Nazli Ozum Kafaee