25949117
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
The what, the how, and the why
There is one last thing to note about the use of predictive analytics to solve today’s business problems.
Sometimes, it is just as important to focus on what you are really trying to predict versus how you are
trying to predict it.
For example, predicting whether a person’s income is greater than $50,000 per year is good.
Predicting whether that individual will purchase a given item is a much better prediction and is highly
desired to improve marketing effectiveness. The key is to focus on the “actionable” part of the
prediction process.
Predictive models are commonly deployed to perform real-time calculations during customer
interactions, such as product recommendations, spam filtering, or scoring credit risks. These models are
deployed to evaluate the risk or opportunity of a given customer or transaction. The ultimate goal is to
help guide a key user decision in real time.
The key to success is to focus on creating prediction models that will ultimately help drive better
operational decisions. Examples of this would be the use of agent modeling systems to help simulate
human behavior or reactions to given stimuli or scenarios. Taking it one step further, predictive models
could even be tested against synthetic human models to help improve the accuracy of the desired
prediction.
The notion of synthetic human models was exactly the strategy that was used to train the Xbox
Kinect device to determine human body movements and interactions. Initially, humans were used to
record basic physical body movements; trainers wore sensors attached to arms, legs, hands, and fingers
while recording devices captured the movements. Once the basic human physical movements were
initially captured, computer-simulated data could then be extrapolated and synthetically generated
many times over to account for variations in things like the size of physical appendages, objects in the
room, and distances from the Kinect unit.
Summary
Azure Machine Learning provides a way of applying historical data to a problem by creating a model
and using it to successfully predict future behaviors or trends. In this chapter, we learned about the
high-level workflow of Azure Machine Learning and the continuous cycle of predictive model creation,
model evaluation, model deployment, and the testing and feedback loop.
The good news is that a working knowledge of data science theories and predictive modeling
algorithms is highly beneficial—but not absolutely required—for working with Azure Machine Learning.
The primary predictive analytics algorithms currently used in Azure Machine Learning are classification,
regression, and clustering.
36