Page 10 - QDigitz

Q!Digitz Vol 1 Aug 2019

Instead, we can apply machine learning technique For example, we might get the result like in 15
like Decision Tree. As said earlier, we believe we percent of client dissatisfaction the key
have a reasonable amount of data for us to combination of drivers were, Private Sector >
construct the decision tree. The data shall have aerospace> product development> Germany>
relevant characteristics (Drivers) like Sector Cloud Technology> Time & Material > Incremental
(private vs public), Domain (Healthcare, aerospace, models. In this when the incremental model
etc), Type of service (Application maintenance, changed to Agile, the value is much lesser.
Product development, etc), Region (countries or
states), Technology (Digital, cloud, big data, Such insights about client satisfaction are gold for
mainframe, .Net, etc), year of contract (1st year, any quality or delivery person to work towards
2nd year, etc), Type of contract (Fixed price, Time building a better recipe for developing software
and Material, etc), Method (Agile, DevOps, with the client. The strength of these machine
Incremental, etc) and many more relevant data. learning models is that it can read a volume of
It's always certain there will be few who will look data and correct the learning to give better
for a pattern in every organization, however results.
pattern in a condition of other variables are
di cult for simple visual inspection. It needs
better application models like a decision tree to
provide insights and give results quickly. To know The application of Decision Tree is only an
more on the decision tree, watch the youtube example, like that they are many algorithms exists
video https://youtu.be/DCZ3tsQIoGU. which are better or comparative. The reason we
are talking about it here is, that we as a quality
analyst shall not just baseline client satisfaction
We can use the existing data in the organization for and leave it there. We can predict the behavior
training the decision tree and for this we might split and we can nd the in uencing characteristics
2 parts for training and 1 part of data for testing. which makes the good recipe for client delight.
We can use the Scikit learning library with Python Let's explore A.I for application in Quality Engine.
for supervised and unsupervised learning of
data. The decision tree was established and
visualized with the decision branches. The nodes
from which the branches starts are the drivers
which we need to watch out for and their values
which leads to client dissatisfaction shall be
controlled or actions to be taken to balance it out.

DigitQ.in