Linear and logistic regression

Hello i have a general question about these topics.
Suppose you want to apply linear regression or logistic regression.

Based on the dataset and data history you have, you represent them in graph and you can determine if those algorithm are fit for your application. For example in logistic regression if the distribution in the graph applies to linear decision boundary.

So you develop it and the application goes live.

A couple of month later you get more history data.
What do you need to do? Represent them again and verify if the distribution changed and change the algorithm? Or it should not change and you don’t need to refactor your application.


Hey @gmazzaglia

When you receive new data, evaluate and plot them to see if the distribution has changed significantly or not (change the algorithm if the distribution has changed a lot).

Then, Evaluate your model’s performance. If the performance has reduced, retrain the model with both the old and new data.