 One of the great things about being a data scientist is being able to predict the future. Of course, I'm not talking about an old lady in a crystal ball, but rather predicting from solid data with some degree of certainty. Today I'll show you how to use Predictions Widget to predict class labels for instances in our dataset. This time, I'll be using data on fruits and vegetables. Because I will use this data to train my classifier, I will refer to it as a training dataset. We have nine features in the training set, including the calorie count, proteins, fiber, vitamin and mineral content. Based on these features, we would like to predict whether a plant is a fruit or a vegetable. Of course, we're interested in which features are the most important for our classification. What tells us best if something is a fruit or a vegetable? We'll check this with classification tree viewer. Here, we can nicely visualize which features best split the data to pure subsets where one of the classes prevails. In our case, the most important feature is the calorie count and then the content of vitamin A and proteins. So most likely, these will be the deciding factors in our predictions. Okay, now it's time to make some predictions. Say I have my own free plants for which I would like to know whether they're a fruit or a vegetable. I know their potassium, vitamin and calorie values, so let's write this down in Google Sheets. The one thing I need to be careful of is to use the exact same names for the features in my test dataset so that orange can match them correctly. Now let's load the data in orange and read it in the data table widget. Okay, all of our data is here. Connect the file widget with predictions. Do we see anything yet? Of course not. We need to give the widget some classification model first. Actually, I've already built the model from my training set with the classification tree widget. Now, all I need to do is connect the classification tree to predictions. I can now view the predictions directly in the widget. Seems like two of my plants are a fruit and one is a vegetable. Of course, I can use other classifiers as well. A fast and simple one is logistic regression. I wonder if its predictions will be different. Let's check. Again, I will connect logistic regression widget to the file widget and then pass the predictor to predictions. Seems like logistic regression agrees with predictions from classification tree. As a matter of fact, I have used the actual data for a kiwi, asparagus and raspberry, that is two fruits and a vegetable. Predictions were indeed correct. Today we've learned how to classify our data with classification tree, how to build prediction models and finally how to use them on a new dataset. While we have made the predictions, we have not really evaluated how good these prediction models were. In the upcoming video, we'll talk about model evaluation and scoring.