AI Workshop - Orange Data Mining

What is the main user interface style used by Orange Data Mining?

• Text-only command line

• Graphical drag-and-drop workflow canvas

• Spreadsheet only

• Web-browser plug-in

In Orange, what is a “widget”?

• A piece of data stored in memory

• A code snippet

• A visual component representing a step (load data, preprocess, model)

• A separate application

Which of the following file types can Orange accept for input data?

• .csv

• .xlsx

• .png (for image-analytics extension)

• All of the above

When doing a simple image-classification task in Orange, what would you typically do first?

• Train the model before loading any data

• Load images and labels (via “File” widget or image-analytics widget)

• Immediately evaluate the model without preprocessing

• Delete all the widgets

Which widget is commonly used in Orange for evaluating model performance (e.g., accuracy, confusion matrix)?

• File

• Test & Score

• Select Columns

• Scatter Plot

True or False: Orange is only for text-mining and cannot handle images.

• True

• False

In Orange, if you wish to change which variable is used as the “target” (label) in a classification task, which widget might you use?

• Data Table

• Select Columns

• Scatter Plot

• Pivot Table

Why might you want to split your data into training and test sets when using Orange (or any ML tool)?

• To confuse the model

• To hide the data from yourself

• To evaluate how well the model generalises to new/unseen data

• To duplicate the results

Which of the following is not a correct statement about Orange?

• It is open-source and free to install.

• It allows drag-and-drop workflows.

• It can export workflows as fully editable Python scripts.

• It supports preprocessing, modelling and evaluation widgets.

If someone is using Orange for a computer-vision mini-project (e.g., classifying images of animals), which of these is a good practice?

• Use just one image per class and train the model

• Make sure your images are labelled correctly, balanced across classes, resize or normalise them

• Skip any evaluation and trust the first result

• Only use default settings and never explore widgets