A contributor article from Ana Martin at iBoske.
Think about it, there are topics which require you to study a great deal: history, medicine, economics, computers… But when you’re just looking for a solution, is it really needed to spend hours, days or years!? No, you don’t want to be a subject matter expert, you just want to make the best decision possible.
Decision trees are great for this kind of situations. They can be used in two different scenarios:
- To provide a solution based on existing data
- To provide a solution based on the current situation
The first one, the one that is based on data, is mainly used in business to analyse data. Tree based learning algorithms are widely used for supervised learning methods. Tree based methods empower predictive models with high accuracy, stability and ease of interpretation.
The second one, which helps people based on their current situation are also known as flowcharts. They’re really easy to follow as they’re essentially a graph, and people loves graphs!
Nowadays it’s really difficult to attract people. Just two examples on how important graphics and images are for people:
- You get up to 94% more views if you add graphics to your content.
- Infographic search has increased by over 800% in just over 2 years
As you can see, if Internet content is king, graphs/images are the queen.
Why should you consider decision trees for your business?
Employee retention is a big issue in many sectors. Teaching new employees, or new situations, in fast changing environments such as help desks and support centers requires a special effort. That’s the perfect scenario for decision trees, because they can adapt problem solving to the situation the client is facing.
Decision trees have been used also many times for troubleshooting, do you remember Clippy, the Microsoft Office Assistant? Based on several questions it could provide you the best solution. But we also remember that many people didn’t like poor Clippy. I think it was because the graphical interface wasn’t attractive enough, remember, people like graphs, and a decision tree without a graph isn’t as attractive.
But also when complex procedures have to be followed, decision trees come to help you. For example, in a town hall, when they have to screen if someone is eligible for a public service.
In ecommerce shops decision trees are very useful also, as they allow the customer to find the right product for their needs. Imagine buying a HP printer… laser or ink, b&w or color, network connection or not…
Also, in those environments where you need to solve something really fast, maybe without prior knowledge, decision trees can save lives, for example in a hospital.
Decision trees drawbacks
There are three main drawbacks:
First, as I explained at the beginning, decision trees are not a good solution when you want to explain or teach some subject, because it may require several written pages and, as you know, decision trees should be concise.
Second, you have to provide a solution. Yes, I know that it sounds elemental, but creating a decision tree is not like writing a blog post, you can’t be beating about the bush, and you can’t talk about something without giving a solution to that situation. Because of that, decision tree creators are usually subject matter experts. And at the same time this makes decision trees more reliable, and their creators more trustworthy.
Third, they can become huge. It’s true, a simple decision may require to consider many different inputs, because they all influence in the right choice. But at the same time it allows the decision tree creator to see the empty branches and realize gaps, which avoids forgetting important elements. Once again, a drawback becomes a benefit. And don’t worry, nowadays decision tree tools like iBoske.com makes it really easy to create a decision tree or flowchart.
I can’t finish this guest blog post without greeting Martin Butler for this opportunity to share with you all my experience with decision trees, and without remembering Sheldon Cooper on Big Bang Theory with his friendship algorithm.
Ana Martín
iBoske.com CoFounder