Machine Learning Business Case
These supervised and unsupervised learning techniques will be covered in Week 1. Types of decisions you might explore include:
Should you hire more staff in the next 12 months, and how many more?
Which of your staff are the highest performers?
Which of your customers should you sell a bundled product to?
Which machine is most likely to require maintenance to avoid breaking down within the next week?
Where should you go to eliminate an imported species using aerial data?
What’s the best route for our trucks in logistics companies to take when delivering goods?
Which patients are most likely to be re-admitted to hospital within three days of being released, considering that there are many written notes from clinicians to understand?
Evaluation framework
For each business opportunity you choose, you are required to evaluate the machine learning method by considering the following criteria:
What is the business problem, with an emphasis on what decisions were made?
Who was impacted by the decisions, and what was the impact?
How was a machine learning technique used to inform decisions?
Are there any ethical considerations, positive or negative, from using this machine learning approach (e.g. a machine learning approach that offsets human biases in the decision-making process)?
What made machine learning appropriate for the business question?
How was success measured?
What were the limitations or deficiencies of the machine learning approach?
Where there any trade-offs considered when building the solution, or in choosing between different machine learning approaches?
How could the solution be improved?
How can the decisions be defended? That is, what statistical evidence do you have from the algorithm to prove its accuracy if the decision is challenged by decision-makers or key stakeholders (e.g. past performance of the algorithm in a similar concept)?