ML-Linear-Regression

This Python program implements Linear Regression using Gradient Descent on any data from a .xlsx file.

Files

• linear-regression.py: The python program that implements Linear Regression and Gradient Descent.
• US_GDP_example.xlsx: The default data used if no input file is provided. This data contains the U.S. GDP from 1999-2022 (source).

Usage

To run linear-regression.py, open a command prompt and navigate to the location of the file. Next, run the following command:

python linear-regression.py -l <iterations-limit> -lr <learning-rate> -i <input-file>

There are a few different flags in this command. Here is what they mean:

• -l specifies a limit to the number of iterations the program performs.
  • If no limit is provided, the program will continuously perform the linear regression until the window is closed or the program is killed.
• -lr specifies a learning rate.
  • The right learning rate differs for each dataset.
  • If your learning rate is too high, the slope, y-intercept, and loss will grow to very high values, and the program will yield an error.
  • If your learning rate is too low, the program will take longer to reach the line of best fit.
  • If no learning rate is provided, the program defaults to 0.000001.
• -i specifies an input file.
  • The program takes in .xlsx files containing a column of Y values and a column of X values.
  • US_GDP_example.xlsx is an example of a properly formatted input file.
  • If no input file is provided, the default file is US_GDP_example.xlsx.

To quickly review the usage from the command prompt, enter the following command:

python linear-regression.py -help

Implementation

To implement this program, I used this article as a conceptual explanation of how linear regression and gradient descent works. After understanding the process, I developed my implementation away from the example code. My program ended up focusing more on OOP and visualization than the article's example.