Classiffcation for accuracy and insight: a weighted sum approach

This research presents a classiffer that aims to pro-vide insight into a dataset in addition to achieving classiffcation accuracies comparable to other algo-rithms. The classiffer called, Automated Weighted Sum (AWSum) uses a weighted sum approach where feature values are assigned weights that are summed and compared to a threshold in order to classify an example. Though naive, this approach is scalable, achieves accurate classiffcations on standard datasets and also provides a degree of insight. By insight we mean that the technique provides an appreciation of the in o uence a feature value has on class values, rel-ative to each other. AWSum provides a focus on the feature value space that allows the technique to iden-tify feature values and combinations of feature values that are sensitive and important for a classiffcation. This is particularly useful in ffelds such as medicine where this sort of micro-focus and understanding is critical in classiffcation.