For a school project I need to choose a dataset from UCI repository and classify the data with KNN after processing it with "feed forward wrapper" feature selection. Googling for "feed forward wrapper" yields nothing... Can someone explain to me what it is ? And even better, describe me the steps to complete this assignment maybe ? What kind of data should I choose in terms of "data type", "attribute types", "number of attributes" ?
Best, Fatih
My first guess is that your teacher wants you to use a feed-forward neural network (also known as a multilayer perceptron, or MLP), and use the output of that as the input for KNN. It makes some sense, although it is unclear to me how you would train the MLP.
Regarding how to choose the datasets: start with a simple one, like the Iris dataset. That one has 4 dimensions and 3 classes. This means that your algorithm will go through it quite fast, and you should be able to obtain decent performances. After your algorithm performs decently on Iris, you can select larger sets.
Thank you for your response, it was a mistake in the assignment text. Apparently it is simply "forward feature selection" for which Google yields quite a bit of information.