A decision tree is a tree-shaped diagram that shows statistical probability or determines a course of action. It shows the analysts and by that, the decision-makers which steps they have to take and how different choices could affect the whole process. Therefore, it is a suitable tool for research analysis, probability calculation, risk management, or strategic planning as well.
A decision tree has three main parts: a root node, leaf nodes, and branches. The root node is the target value that we are seeking to reach. The leaf nodes contain the information about criteria. The branches connect the nodes and show the route through the leaves to the target value.
Decision or Prediction
Decision trees are often mixed with prediction trees. Although they do show many similarities in form and function, on decision trees the terminal nodes are labeled with a decision, based on a prediction, like: “seek more information” or “investigate this”.
On the other hand, prediction trees are more specified, as the terminal nodes are labeled with the prediction of a variable, like “Chance of [event] is..” or “average of [variable] is”.
In AnswerMiner we are using prediction trees for the decision-making processes. This gives the advantage to determine all the actions based on collected and processed data.
Creating a Decision Tree with AnswerMiner
At first sight, it may look difficult to create a decision tree from raw data. The good news is that AnswerMiner builds the tree automatically from the dataset. There are only a few steps to take for making a prediction tree.
First, open up your AnswerMiner account and choose a dataset or upload your own one. In this article, we are going to use the Employee Satisfaction data set as an example.
To use the decision tree feature or Prediction Tree in AnswerMiner, simply click the icon on the left sidebar.
For making the tree, you have to choose a target variable. You have to know, what you want to predict.
Let’s suppose, you want to predict workforce fluctuation as an HR specialist.Your target variable will most likely be a logical variable, in this case, the Left value. This means you want to know the factors related to employee attrition. Feel free to choose other variables as a target by clicking on the Target box.
Tree in Action
Now is the time to fine-tune your decision tree. What if you want to know which department has the highest fluctuation? Or what is the problem related, in general? Below the Target value, you can see the different Predictors as well. Click on the columns to highlight them and click again if you want to deselect and add filters if needed.
After picking the right variables, you will be able to see the automatically generated Prediction Tree. No coding needed, the platform itself computes the results.
See the Confusion Matrix under the tree? This indicates the accuracy of the predictions made by the system. This pops up only when you chose a logical or ordinal variable as a target. If you change the cutoff values in this mini-feature, you can see how accuracy alters too.
Now, if you want to change the colors of the legends, click on the column with a red-blue gradient left to the tree. Here you can see two main options: narrow and full. The full version of the gradient coloring takes into account the full range of the data, while the narrow coloring option is based on the values showed on the tree.
In order to improve the performance of the app, in case of a high amount of values AnswerMiner uses a representative sampling. This can be turned On & Off in the settings. Please note that if you turn representative sampling off, the calculation may require more time.
Finally, you can even see explanations related to the predictions on the right side of your screen. This helps you to understand the logical background of the calculations made by the algorithm. Also, by clicking on the nodes you can find more insights about the different subsegments. Save, download, and share your Decision Tree and create utilizable reports.
Decision trees are powerful visualization tools that help decision-makers to make the right moves at the right time. As we are living in a data-driven world, more and more sectors are using them nowadays. To optimize strategies, predict the outcomes or the probability of events, prediction trees can be utilized in many different ways to reach the wanted goals, or to avoid pitfalls.
How much time do you need to analyze and understand your data?
With AnswerMiner, making predictions is much faster and easier, and no coding is required.