Kalyan Banga Staff asked 5 years ago
Hi, I’m a beginner of analytics. I have a transaction dataset: Sold, Decription Item*, Sottocategory*, Category*, Segment*.
The attribute * are hierarchical. I want understand a good granularity for analysis, In order to reduce dimensionality, and so perform the analysis according the choice granularity.
Is there any technique?
Kalyan Banga Staff answered 8 years ago
It seems that you may be unaware of clustering methods. Since this seems to be an introductory project for you I’d suggest looking at some of the methods offered by R’s cran ‘cluster’ package. https://cran.r-project.org/web/packages/cluster/cluster.pdf
With proper clustering you can reduce your hierarchical dimensionality by multiple factors (based on your choice of number of clusters) while still retaining the majority or pertinent information (or extrapolate later to refill the information).
r-tutor.com has created a decent introductory tutorial for a hierarchical clustering method, which you can find here: http://www.r-tutor.com/gpu-computing/clustering/hierarchical-cluste…, or check out a more advanced technique posted by Vincent Granville on this website which uses R-Shiny and D3.js http://www.joyofdata.de/blog/hierarchical-clustering-with-r/ (from his comprehensive list of data science resources http://www.datasciencecentral.com/group/resources/forum/topics/comp…)
Good luck on your project!