How to develop an elastic net model for the data

Hello! I need help in developing an elastic net model for the following data. I am novice to machine learning so i would appreacite your guidance so that i can learn the concept. I have read the relevant information for the model but need guidance so that I can try. Appreciate your help! I have attached the data for reference. Thanks!

Answer 1

[Akshat]

Hi Aishwarya,

I see you are trying to implement an elastic linear model for the data you have attached. Before starting off with my answer, you can refer to the following documentation links for better understanding of the features you can utilise:

Elastic net regularization/ Lasso. This will be the model you are going to train. : https://www.mathworks.com/help/stats/lasso.html

"Readtable" function. This will be used to import your data in CSV to the MATLAB workspace. https://www.mathworks.com/help/matlab/ref/readtable.html

"cvpartition" function. This will be used to split your data into training and testing sets.

https://www.mathworks.com/help/stats/cvpartition.html

Following are the steps to train a model with your data:

1. Get the data in the workspace. You can simply run this command and get the data from your CSV File.

data = readtable('Regmodel1.csv');

2. Judging by the data, you are looking for a model to predict the last column, 'PQ'. The rest are the predictor variables. Also, whenever you are training, you seperate some data out to test the model out. The ratios can be 70:30, 75:25 and 80:20. I have done a 80:20 split, using the 'cvpartition' function. The splitting part is optional, you can skip if you do not want to test.

X = data(:, 1:end-1); 
Y = data(:, end); 

cv = cvpartition(size(X, 1), 'HoldOut', 0.2);
idx = cv.test;
XTrain = X(~idx, :);
YTrain = Y(~idx, :);
XTest = X(idx, :);
YTest = Y(idx, :);

3. Train and test! The 'Alpha' argument in the 'lasso' function will let you decide the elasticity in the regularization. Set it to a number between 0 and 1.

[B, FitInfo] = lasso(XTrain, YTrain, 'Alpha', 0.5); 

Hope this helps!