Student Replies for Module 4
Discussion: Recent Trends in Business
Analytics
Week 2: Respond to at least two of your classmates’ initial posts. Your
response should be substantive and further the discussion. It is OK to be
critical and use critical thinking. That means you can question what your
classmates say. Do they provide their own critical thinking and logic?
The US Air Force has over 5,400 aircraft in its inventory, with an average airframe
age of 29 years. The Air Force cited in 2017 that “we are at our lowest state of full
spectrum readiness in history.” (Heritage.Org, 2019) The availability of serviceable
aircraft used to conduct combat and training operations, is measured by metrics
such as mission capable (MC) status and aircraft availability. These rates have been
in steady decline across the fleet and in key mission series over the past several
years with an average MC rate in 2018 of 69.97%, down 1.33% from the prior year.
(Losey, 2019) One of the ways that the Air Force is seeking to address this decline is
adopting AI and machine learning to develop predictive maintenance models to
increase aircraft availability rates.
Predictive maintenance uses historical data on part failure rates, mean time
between failure, metadata of the equipment, etc. to accurately predict
maintenance. The goal is to replace components “Just in Time” in order to gain
efficiencies in cost and down time. There are three important considerations for
implementing predictive maintenance models:
1. The problem must be able to be predicted
2. The data record is of sufficient quality and incorporates good and bad
outcomes (both the discrepancy and the repair)
3. Domain experts who understand the issue, processes, and systems that can
assist analysts in structuring the algorithms
Continual model evaluation and optimization is necessary when implementing
predictive maintenance models:
1. Accuracy – the ability of the algorithm to make correct decisions from a
sample data set (and theoretically perform correctly in the future based on
real time data inputs)
2. Precision – the rate of false alarms
3. Recall – correct identification of failures
4. F1 score – average of precision and recall rates from 0 (worst) to 1 (best)

Student Replies for Module - Recent Trends in Business Analytics

Contribute Materials

Related Documents

+13062052269

info@desklib.com