Sasin MiM · Unsupervised learning
Find groups in data that has no labels, on our datasets or your own file. The computer does the maths; you decide what the groups mean and whether they are useful.
A dataset is a table: each row is one data point, each column is one piece of information. Here there is no target: nobody has told us which group a data point belongs to.
Features are the columns the computer looks at to decide which data points are alike. Different features give different groups. Nothing is ticked for you: choose the columns that matter for your question.
Number features are put on the same scale first, so that a feature does not count more only because its numbers are bigger (for example income in thousands against age in years).
The method is k-means. You choose the number of groups. Each data point joins the nearest group centre; each centre then moves to the middle of its group; the two steps repeat until nothing changes.
The computer only gives each group a number. What a group is has to come from you: look at how each group differs from the average, then give it a name.
Every grouping you run is logged here so you can compare them. The log disappears when you close or reload the page.