Skip to content

Latest commit

 

History

History
66 lines (48 loc) · 7.1 KB

README.md

File metadata and controls

66 lines (48 loc) · 7.1 KB
Raw

is477-fall2023-final-project

Public repository for IS477 Fall 2023 final project DOI

Overview

  • Purpose: is477-fall2023-final-project is Github repository for the final project of IS 477 Data Management class. The project is about creating the pipeline or the entire process of data management, such as acquisition of data, check integrity of data, data analysis, preservation, legal concerns, and reproduce of data and analysis.
  • Overall project requirements: The project requires basic installation of GitHub, Docker, VScode, and GitHub accounts with repositories for the final project of IS 477.
  • Dataset: The dataset used for this project is from UC Irvine Machine Learning Repository (https://archive.ics.uci.edu/), and the dataset is Wine dataset (https://doi.org/10.24432/C5PC7J). Wine dataset contains total 14 columns that consist of 13 features and 1 target variable for classification purpose. The dataset has total 178 rows, or instances. Wine dataset has Creative Commons Attribution 4.0 International (CC BY 4.0)license, which allow users share and adopt of the dataset with any purpose if appropriate credit is given.
  • Analysis: The purpose of the analysis is figuring out which machine learning classification model showing a best classification of quality of wine that based on 3 classes. The dataframe splitted into target variable (y) and features (X), and then splitted more into train X and y for select classfication model based on metrics scores and test X and y for give final evaluation on the selected classfication model. All models are standard scaled and Grid Searched with 5 fold corss validations to figuring out best parameters with less overfitting for each models.

Contributions

  • Jung, Changho: Creating and finishing the files in data folder, results folder and script folder. Checking and revising the rest of the files in the project.

  • Wen, Dongjing: CO-finishing the script folder and data folder.Creating and finishing the README.md, requirement.txt, snakefile, environment.log, .gitgnore, zenodo.jason, profiling files. Checking and revising the files in results folder.

Analysis

  1. From the applying the train dataset, Logistic Regression showed the best accuracy score: 1. Therefore, Logistic Regression is selected for the final application for classification.
  2. Then, the selected model, Logistic Regression is fitted with test X and y for the final evaluation of the model. To evaluate the model with slightly more detailed, confusion matrix with metrics like accuracy, precision, and recall used. Logistic Regression showed all correct, 100% accuracy, precision, and recall with no negatives or error in confusion matrix.
  • Result: Logistic Regression showed the best performance to classifying wine quality class according to the analysis, compare to Decision Tree and Random Forest. However, since the dataset contains only 178 instances, considering this analysis as no more than performance purposes and may be further works needed for applying to the work place.

Workflow

Workflow DAG graph created by dag.py.

  • The overall workflow consist of 3 steps. Prepare is preparing data, Profile is profiling dataframe from the data from Prepare, and Analyze is analyzing from the data from Prepare.

Reproducing

  • All the code in the program need the python environment of Python 3.11.~ (but Python 3.10 also works fine)
  • First, download or clone this repository from Github and unzip if your downloaded as zip file.
  • Then, reproduce methods can divided into two: Docker and environment.log

If you are using Docker

  1. For acquiring and check integrity of data: docker run --rm -v ${PWD}:/is477 username/is477-fall2023-final-project:v1 python scripts/prepare_data.py
  2. For profiling data: docker run --rm -v ${PWD}:/is477 username/is477-fall2023-final-project:v1 python scripts/profile.py
  3. For analyzing data: docker run --rm -v ${PWD}:/is477 username/is477-fall2023-final-project:v1 python scripts/analysis.py

If you not use Docker

  1. Clone this repository (is477-fall2023-final-project)

  2. Create a virtual environment (recommended), by install "pip3 install virtualenv", create by "python -m venv ", and activate "source .venv/bin/activate".

  3. Install the required Python dependencies: use this command on terminal, "pip install -r requirements.txt", and make sure check enviornmental.log

  4. Run the data preparation script in order: 1. "scripts/prepare_data.py" and unzip adult.zip in data file under is477-fall2023 directory, 2. "scripts/profile.py" or "scripts/analysis.py" at your choice. Running the scripts can be done on the terminal in root directory (for this git clone, is477-fall2023), like python prepare_data.py; python scripts/reproduce_adult.py; python scripts/reproduce_reconstructed.py.

  5. If downloading the UCI Wine dataset will complete after running "python scripts/prepare_data.py", you can run the "python scripts/profile.py" for data profiling and "python scripts/analysis.py" to commence the calssification process. This will add accuracy score results from classfication task of scripts/analysis.py based on dataset from scripts/prepare_data.py, and the result will recorded as txt files in result repository. scripts/profile.py will add data profiling html file on the profiling directory.

License

  • Creative Commons CCZero (CC0-1.0) for the license.
  • Characteristic: No Copyright, Global Applicability, Freedom to Use.
  • It is a very good license to the project, because it provides the neccessary part of it

References