Subscribe to DSC Newsletter

Market Basket Analysis - An Interesting Use Case of Speech Analytics Output used with RapidMiner

Association Rule creation using Speech Analytics Data

There are ways and techniques to extract information out of the day to day data that an organization produces and looks at. To make the most out of the data any organization should be geared towards utilizing these tools to dig “Golden Nuggets” out of it.

Today we will talk about how to use a speech analytics word cloud or call category data to produce an association algorithm to predict what are the chances of a customer discussing a particular topic under the umbrella of Association Rule using R? This will help us to create an outcome that can be understood as “If this stated, then that will be mentioned”

Application

So what kind of discussions or items we are talking about?

There could be numerous applications of association:

  • Member sentiment analysis: When a Member talks about Claim Denial while talking about Claim Status he/she gets confused. This provides a clear opportunity for any healthcare payer to ensure that up to date information is provided to the members related to their claim status. This will not only help them to reduce member effort but will also reduce the call volume.
  • Website Enhancement associations: When a Member talks about Web Portal in conjunction with finding information they tend to state phrases related to repeats. This is an indicative of website navigation related improvement opportunities

The Member Call Reason Data Set

Imagine there are 500,000 Member conversations uniquely identified by the speech analytics application with call reasons about which a Member contacted the call center. These conversations represents the Customer’s Cart or Basket and therefore we are using “Market Basket Analysis” technique to understand the next probable discussion topic.

This is what the Member conversation data set includes; a collection of mediafiles representing 1 call type that a member discusses. Each mediafile is called as Member conversation and each row represents a mediafile.

 

Let’s understand the Maths behind it this algorithm:

We can represent the Member Conversations as Item in an item set:

                                                                                                   I = {i1,i2…..in}

Therefore the Member Conversation is represented as a:

                                                                                                   tn = {ia,ib…..in}

The above method gives us the rules represented as:

                                                                                                   tn = {i1,i2} => {ia}

The above can be read as “If a Member mentions an item in a an item set on the left hand side, then the Member is likely to mention the item on the right hand side too. To increase the readability and understanding of the above thought here is an example:

                                                                                              {Benefits, Coverage} => {x-rays}

If there are mentions of benefits and coverage, then that member conversation is likely to have conversations related to X Rays as they are not sure if the X Rays are covered under their plan or not.

The above helps us to understand 3 important ratios within the Member conversations:

  1. Support – The fraction of which a conversation item is occurring in the datbase
  2. Confidence – Probability of the correctness of the rule for new conversations within the item set
  3. Lift – The ration by which the confidence of the rule exceeds the expected confidence.

Point to remember: Lift 1 indicates the independence of items on the right and left (connotes: the items are independent)

Now that we understand the working of the Association Rules / Market Basket Analysis lets understand its use with various applications to mine the insights. Though there are multiple applications and methodologies that are available to perform this analysis (R, Rattle, Rapid Miner etc.). However for this demonstration we will talk the basic workflow within R and RapidMiner.

Market Basket Analysis Flow in RapidMiner

Output (Clearly Outlines what to expect when a particular conversation happens):

Conclusion

When speech analytics outputs are used in conjunction with these powerful Data Analysis and Mining Tools the insights gained out of these methodologies can help the organizations to not only improve their current processes and performance but also take the customer satisfaction to the next level.

Views: 2492

Videos

  • Add Videos
  • View All

© 2019   Data Science Central ®   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service