CSE 8803 / ME 8883 - Materials Informatics Course - Fall 2016

Discoverability
Visible
Join Policy
Invite Only
Created
20 Aug 2016

Nanocrystals: Blog Post 8 - Regression Model

Extracting morphological changes in nanocrystals using in situ liquid cell microscopy’

Regression Models

Now that we have the first three pc scores for all the microstructures, we can start building a model. For simplicity, we will start with a polynomial regression model and if it proves to be sufficient we will keep it. The first thing to decide when building a polynomial regression model is the order of the polynomial. We created second-order through seventh-order models and calculated several metrics for error to compare. Below are plotted the various error metrics as a function of polynomial order for the first pc score.

rsquare.JPG

mae.JPG

maestd.JPG

For each metric, the 5th order polynomial performed the best. This includes for the second pc score which is not shown. Therefore, we choose the 5th order polynomial for our model for the first two scores. For the third score, a 6th polynomial was best, as shown below.

pc3r2.jpg

pc3mae.jpg

Created on , Last modified on