Overview
Clustering is a machine learning algorithm that looks for patterns in your data. It works by grouping similar records together based on comparing their attributes. What makes clustering so powerful over traditional analytics and reporting is that you don’t need to tell the machine learning what patterns to look for; clustering sifts through your data and reveals the patterns. Some applications of clustering are (1) identify automation areas for Virtual Agent or Flow Designer (2) identify issues with existing process.
Lab 3 – Find patterns using clustering
Configure and Training Clustering
1. All > Predictive Intelligence > Clustering > Solution Definitions
2. Select New
3. Fill out the following fields:
Label = HR Automation Clustering
Word Corpus = HR Cases (hit the magnifying glass)
Table = sn_hr_core_case
Fields = short description
Select “Create Cluster Insight Table”
Group By = HR Service.name
Under Purity fields select Assignment group.Name & Priority
4. Hit submit & train in the upper right to train the clustering solution. You cluster may take 15min – 30minutes to train depending on the amount of data.
Review the cluster results
1. Go to all > predictive intelligence > homepage
2. Scroll down to clustering and expand the carrot
3. Look for the cluster solution that you just trained and click on the Version link.
4. Click the Cluster Visualization tab to show the tree map of the clusters. The tree map shows large squares which represent the HR Service and the small squares within the HR Service represent clusters of cases that are similar to one another. Below is a cluster tree map against the short description of my HR case data. Your data will obviously be different.
5. I want to focus on my largest cluster the “Pharmacy Benefits Inquiry”. Click into it to show the clusters within this HR Service.
6. Hover your cursor over the different clusters. The cluster concept are the top words the cases in the cluster have in common, the size represents the number of the records in the cluster, the purity represents how many records below to that priority and assignment group.
7. I move my cursor to one of the smaller clusters, it appears there are some issues around benefits links that take the user outside of the organization.
8. Click into that cluster to get a detailed view of the case in this cluster. Here you can do what you normally do in a list view such as filter, sort, graph, etc. Perusing the short descriptions shows me that we have several cases that have issues because of external links. This insight is something that we can use to improve our HR benefits process, all thanks to Predictive Intelligence.
Congrats you’ve finished learning how to find patterns in your HR case data using clustering.
For more advanced instruction on clustering please go back to article 1 and in the NowLearning courses see the Predictive Intelligence Advanced Topics. There are clustering articles that cover in greater detail all the specifics in clustering.
