Visual Programming Using Orange Tool

In this blog, I will discuss Visual Programming and demonstrate using orange tool. Orange tool is an open source, component-based software written in Python language that works best for machine learning and data mining — namely, visualization.

If you don’t know much about Orange Tool, I’ll recommend one of my blog — Introduction to Orange Tool. This blog is an easy and quickstart to Orange tool.

About Visual Programming

It is a type of programming language which anyone can understand. Anyone can describe the process using illustrations. In text-programming language, a programmer things like a computer whereas in visual programming, programmer describes process which makes sense to a human being.

Here are some tools through which you can practice visual programming — Scratch, Blocky, agentcubes, etc.

Implementation on Orange Tool

Here, I have placed a File widget and selected zoo.tab as my dataset. You can also import your own dataset using browsing or using url. This datset is of type classification and have 16 attirubtes & 101 data instances.

Now, my dataset is ready. I have connected File widget to Data Sampler widget. Data Sampler widget will split data into training & testing set. I have kept 70% of data for training & rest for testing. It is suggested to keep training dataset between 70–90 percent.

Connect Data Sampler widget to Test and Score widget. Also, connect machine learning models to Test and Score widget. I have used 3 — Neural Network, Logistic Regression, SVM algorithms for my demonstration. These 3 algorithms are most suitable for classification. As logistic regression & SVM are specifically designed for such kind of problems. And neural networks is deep learning based algorithm. That’s why I have selected these 3 algorithms.

Here, Data Sampler widget passes 70% of data for training and rest 30% for testing to Test and Score widget.

Now, we compare the 3 algorithms on the basis of AUC, CA, F1 score and we can analyze their respective Precision & Recall. Below image shows the results of training data. We can observe that for the choosen dataset we get 100% precision for all the 3 machine algorithms.

It’s time to test the dataset on training data. We can observe, SVM gives the heighest precision i.e., 97%. All this analysis would be of great help when writing efficient code for classification kind problems.

In this blog we have compared 3 different machine learning algorithms for classification dataset. Using orange tool, analysis become a lot simpler and easy to carry out