Using Naïve Bayes with Speech Analytics Output to predict First Contact Resolution

What is Naive Bayes?

It is a classification methodology based on Bayes Theorem assuming independence among the predictors. In simple terms, a Naive Bayes classification assumes that presence of features are not correlated to the other features that are present in the data. Being easy to use and understand it can be used with large data sets. Naive Bayes model is known for outperforming other complex and sophisticated classification models.

bayes-rule-formulae

How it all works?

Let’s try and understand the usage of this algorithm on an output generated by a Speech Technology based application. As we all know speech technology (Customer Interaction Analytics applications) helps businesses to categorize call volume and allows its users to use the data to gain additional insights into their business performance. Therefore, stakeholders will be interested to understand the *FCR probability on various call categories generated by a speech application. The output has another layer of metadata that indicates whether a call was surveyed as *FCR or Non-FCR (this metadata element will work as a dependent variable or to be predicted variable).

Below I have the training data set of call categories and corresponding target variable “Survey” (suggesting survey outcomes FCR or No FCR). To be able to classify whether a call will be FCR or Non-FCR by each call category we will perform the under listed steps.

Step 1

Convert the data into frequency table

Step 2

Create likelihood table by finding the probability like Call Category 1 FCR % = 5% and Probability of FCR = 45%

categories

Step 3
Now, that we have the Table of Likelihood we can use the Naïve Bayes to calculate the posterior probability for each class and the highest posterior class’s probability will be the outcome of the prediction

Problem to solve for: Which Call Category will give the highest FCR %

This can be solved by using the above method of posterior probability

P(FCR%)= Call Category FCR% * Overall FCR% / Overall Call Cat%

Here we have P(FCR%) = Call Category 1 FCR% 1/20 = 0.05, Overall FCR % 20/44 = 0.45, Overall Call Category 1 % 5/44 = 0.11

Now, P(FCR%) = 0.05*0.45/0.11 = 0.20 which has a low probability

Conclusion on Naïve Bayes Classifier Algorithm

Naïve Bayes is based on the independence assumption

  • Training is very easy and fast; just requiring considering each attribute in each class separately
  • Test is straightforward; just looking up tables or calculating conditional probabilities with normal distributions

Naïve Bayes is a popular generative classifier model

  • Performance of Naïve Bayes is competitive to most of state-of-the-art classifiers even if in presence of violating the independence assumption
  • It has many successful applications, e.g., spam mail filtering
  • A good candidate of a base learner in ensemble learning
  • Apart from classification, Naïve Bayes can do more…

The real power of data is revealed when Speech Analytic outputs are used in conjunction with these powerful algorithms in a thoughtful and strategic manner.

*FCR = First Contact Resolution in other words One and Done

Advertisements

2 thoughts on “Using Naïve Bayes with Speech Analytics Output to predict First Contact Resolution

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s