We used 3 component PCA for purchase data as we did for segmentation data. My question is that why we did not check for the explained variance of 3 components before applying PCA. Could we use higher components to account for more variability?

Hi Dikshit, thanks for reaching out! We looked at the cumulative variance graph and saw that if we keep 3 of the components we keep around 80% of the variance. So, using the rule of thumb we decide to keep the first 3 components. We could also have a 4 components solution, then we'd keep almost 90% of the variance. One of the assignments is to segment the dataset using the 4 components solution. Here's a snip of the graph in question: Hope this helps! Best, 365 Eli