“It is in your moments of decision that your destiny is shaped.” – Tony Robbins
Part 1 – Why you should include decision trees in your business solution
Decision trees are a graphical representation of a decision path. They illustrate potential solutions and selection criteria when making decisions.
It is a tool used in the majority of areas that involve decision making. As examples of questions that could be answered by a decision tree, we can think of: what treatment should a doctor choose for his patient? Which customer should be solicited by an insurance broker? What kind of product launch should a corporate decision maker prioritize?
As a human being, one of the biggest limitations to decision making is that it is difficult to choose beyond known alternatives. Decision trees help to formalize the brainstorming process and identify a multitude of possible solutions based on a large amount of data already collected. The decision that emerges naturally is not always the best; relying on verifiable data and facts is sometimes more profitable than following one’s instinct.
Why turn to a decision tree?
Why would you benefit from learning more about decision trees? To optimize your time and maximize your performance! When a virtual assistant uses a decision tree, it is possible to automate a portion of the thinking when making a decision. In today’s digital age, this type of solution is becoming increasingly popular and allows managers to focus on higher value tasks and reduce opportunities for errors. Choosing a software with a decision tree is thus a differentiator for a company that allows it to stand out from the competition, helping, for example, managers optimize their operations or increase their sales.
Let’s keep going!
Now that you are convinced of the need to learn about decision trees, here’s a summary of what you need to know:
A decision tree is able to be built by itself when provided with a training set. In the field of machine learning, therefore, the decision tree belongs to the category of supervised learning. Each entry contained in the training set includes a list of features and also provides the response value that is to be predicted. It is this combination of features / response that will be used to build the tree. Concretely :
Name Status Age Nb of kids Income Accepted our new life insurance
M. Tremblay Married 58 5 55 000 Yes
In this way, we can maximize the rate of positive responses by concentrating on those clients who are most likely to accept one or other option, thereby maximizing their time and increasing their income. Since a tree is as accurate as the data it relies upon, it is crucial that your set be as accurate as possible.
Types of decision trees
A decision tree can be deployed in several forms according to the criteria that will be assigned to it. In practice, two main types of trees stand out.
- Classification : The tree is used to predict the class to which the data belongs. The response variable is qualitative. e.g : Yes / No, Buy / Sell / Keep
- Regression : The tree is used to predict a certain value that the data will have. The response variable is quantitative. e.g : 1, 10, 50 …
How to interpret the decision tree?
A decision tree can simply be considered as presenting the possible paths of a decision. Each possible path of the tree, from the initial node to the leaves, can be considered as a decision rule.
Here are the different parts of the tree:
- Nodes (rectangle) : The condition that a decision faces. These are the independent variables, and are the features listed above.
- Branches (arrow) : The decisions taken at each node of the tree.
- Leaves (ellipse) : The response variable that we are trying to predict. This is the final decision provided by the tree.
In the example above, considering a married customer, 58 years old and who has more than two children, it would be justified to contact him to offer him a new life insurance.
Advantages and disadvantages
- Simplicity : Very visual and easy to interpret. Provides more visibility and more intuitive than a neural network, for example.
- Realism : Imitate the way humans make decisions. Give an accurate algorithm that anyone can follow.
- Quick preparation : Very little preparation and cleaning of the data bank to be performed (no standardization required, processing of missing data or creation of dummy variables). Moreover, no hypothesis is necessary to create the tree.
- Versatility: The model takes into account both quantitative and qualitative variables.
- Instability : A slight change in the data can cause a change in the tree’s training and a repercussion of this change on the prediction. Having the most recent tree ensures that it represents all the data.
- Statism : A decision tree is not autonomous if it is left to itself (mainly because of its instability). It needs to be maintained regularly, especially in changing areas. However, some tools allow the automation of the process and thus, the facilitation and acceleration of the construction of a tree.
- Accuracy : The method is not as accurate in terms of predictive efficiency as other methods of learning. It is easy to increase the efficiency of the prediction thanks to a random forest of decision trees.
- Over-learning : The tree is not immune to complexity and tends to be over-trained, which can cause the tree to not generalize beyond the training data. The best way to overcome over-learning is to prune the tree.
Decision tree in business
A decision tree can be added to virtually any business solution. Whether in your CRM, your transport and logistics software or in your financial system, a decision tree will find its place and will certainly improve and standardize your decision-making process. Thereby, your company will stand out from others by its performance and efficiency.
To get the maximum benefits, it is essential that this decision tree is alive and well built. Be assured that Arcbees is able to give you the best advice and support your business needs by allowing you to exceed your own goals.
Feel free to contact us directly by email or leaving a comment below. It is with great pleasure that we will discuss it with you. Our next article will guide you through each of the steps to move from a set of data to a functional tree. Don’t forget to follow Arcbees’ blog not to miss anything!