 All of our previous datasets came from Orange's dataset server. Remember, so far we've been using the datasets widget. We loaded the data, say on countries and their socioeconomic indices, and started off by checking it out on a data table. But what if I wanted to use some of my own data? As it turns out, Orange can't read tabular data from various sources, including Excel files. So, let's start out by constructing one such dataset. My spreadsheet will include some data on students. First, I add a column of names, let's say Jill, Anna, John, Stacy, Mike, and Fred. Next, I add a column with their age, gender, and eye color. And then, I add their times for a 100 meter run. Not that this data makes any sense, it's just an example of how to prepare a tabulated dataset for any machine learning. Now, I store the dataset I just created to the desktop in a file. Let's just call it fitness. Now, I open Orange and place a file widget on the Orange canvas. Like with other widgets, I use the right click to find this particular widget. Orange comes with some preloaded datasets that I can access from this widget, but right now we're interested in our fitness data. So, I click on the three dots button and find my Excel file. Here it is. The file widget reports on the data columns and shows the names of the features. Notice that Orange also detects the feature types. For instance, age is a numerical feature, as it stores numbers. The color of the eyes is categorical. Categorical and numerical features profile the students and we can use them, for instance, to compute distances. We can't do anything computationally with names, sort of textual meta features, though. Meta features provide additional information about the data instances, for example, to identify them in a scatter plot, but Orange doesn't use them in any modeling. As a side note, Orange does actually provide a way to analyze text in the text mining add-on, but I'll skip that for now and go back to our dataset. The first thing to do after loading the data is to check it out in a spreadsheet. Let's open it in data table. Here, we find our data just like we set it up in Excel. Let's say I don't need information on eye color. I can use the select columns widget and move the eyes from the list of features to the ignored box. Checking the data in a new data table, I indeed find no eye colors here. Good. The select columns widget is a great widget to manually edit the data domain and play with including and excluding features. We'll use it from time to time in our next videos. Moving on, I would now like to save my edited dataset. This can be easily achieved with the save data widget. I just click save as. Choose to save my data in an Excel file and rename the file to something like fitness no eyes, as creepy as that sounds. This file can also just go on my desktop for now. Taking a look at the file, I can double click it to view it in Excel. Here, notice that apart from the feature names, Orange also stored some extra data on the feature types in the second and third rows. This extra information can be toggled on or off in the save file widget. Great. Now we know how to repair some of our own data for Orange. And we also know how to export data from Orange. Note again that every time we repair our own datasets, it's just good practice to first check it out in data table. In my running example, I used Excel, but Orange also loads tab or comma separated datasets, as well as datasets in a bunch of other formats.