Machine Learning and Statistics - Project

Machine Learning and Statistics module, Data Analytics Course, GMIT 2020-2021
Lecturer: Dr. Ian McLoughlin

Author: Andrzej Kocielski
Github: andkoc001
Email: G00376291@gmit.ie

---

Introduction

This repository houses my assignment project for the Machine Learning and Statistics module, Galway-Mayo Institute of Technology, 2020.

Project objectives

The objective of the project is to develop a web service to make predictions using Machine Learning (ML) paradigm.

The goal of the project is to produce a model or models that, based on the provided dataset power production, and through applying the appropriate ML techniques, predict power output generated by wind turbine from the wind. The power output predictions should be generated in response to wind speed values to be obtained as HTTP requests.

The detailed project instruction can be found in the pdf file.

Assignment delivery

The project is about development of a machine learning models in Jupyter Notebook environment, using existing algorithms and tools, such as Keras, TensorFlow or Scikit learn packages. The applied models are subsequently assesst on their accuracy.

In the Jupyter Notebook Powerproduction_ML.ipynb, I have incorporated the research and described the project progress. It is illustrated the machine learning concepts and applied methods together with relevant code snippets. The notebook includes also the calculated outputs and plots with the accompanying descriptions.

A web app is built upon the Flask framework. The local server hosts the selected machine learning models used in the project. An attractive front-end web page has been designed to present the results in a user-friendly way. The app returns predicted values of power output, upon user input wind speed.

The web app is also available in a virtualised container on Docker platform.

---

Accessing the Project

Viewing Notebook

For viewing the notebook online, it is recommended to use Jupyter Notebooks viewer, nbviewer. Paste the link to the notebook to be inspected into the provided field.

Accessing the Web App

In order to open the web app localy, clone or download this github repository (keep the directory structure intact).

Option 1) Virtual environment

The following steps depend on the operating system.

A) Linux

Within the directory, create a new virtual environment, typing in the terminal

python -m venv venv

Activate the virtual environment:

source venv/bin/activate 

Install all the required libraries listed in the requirements.txt file:

pip install -r requirements.txt

Start the web app using the command:

export FLASK_APP=app.py

To run the server program, type:

python -m flask run

This will activate localhost server at http://127.0.0.1:5000/.

To stop the server running, press ctrl+c in terminal.

In order to leave the virtual environment:

deactivate

B) Windows

Within the directory, create a new virtual environment, typing in the terminal

python -m venv venv

Activate the virtual environment:

\venv\Scripts\activate.bat

Install all the required libraries listed in the requirements.txt file:

pip install -r requirements.txt

To run the server program, type:

set FLASK_APP=app.py

To run the server program, type

python -m flask run

This will activate localhost server at http://127.0.0.1:5000/.

To stop the server running, press ctrl+c in terminal.

In order to leave the virtual environment:

deactivate

Option 2) Docker

Install Docker on your computer, typing in the terminal:

pip install docker

To create a new docker image, inside the directory holding the repository type:

docker build . -t wind-power-app

Alternatively to creating a new one, pull the image from the dockerhub, using the command:

docker pull andkoc001/wind-power-app

In order to create and start the docker container, execute the command:

docker run -i -t -p 5000:5000 --rm wind-power-app 

This will activate localhost server at http://127.0.0.1:5000/.

To stop the server running, press ctrl+c in terminal. The container will be automaticly removed.

---

References

General, high-level, reference sources are listed below. References to specific problems are included in the Notebook.

Regarding the project

• Project details project brief.

Regarding Python environment and tools

• Python - the programming language used for the project
• Jupyter Notebook - interactive python code editor
• NumPy - scientific computing with n-dimensional arrays
• Pandas - data analysis and manipulation tool
• Scikit-learn - machine learning library
• Matplotlib - plotting library
• Seaborn - data visualisation

Regarding web app deployment and virtualisation

• Flask - web development framework
• Docker - OS-level virtualisation platform

---

Andrzej Kocielski, October 2020 - January 2021