Providing accurate predictions is challenging for machine learning algorithmswhen the number of features is larger than the number of samples in the data.Prior knowledge can improve machine learning models by indicating relevantvariables and parameter values. Yet, this prior knowledge is often tacit andonly available from domain experts. We present a novel approach that usesinteractive visualization to elicit the tacit prior knowledge and uses it toimprove the accuracy of prediction models. The main component of our approachis a user model that models the domain expert's knowledge of the relevance ofdifferent features for a prediction task. In particular, based on the expert'searlier input, the user model guides the selection of the features on which toelicit user's knowledge next. The results of a controlled user study show thatthe user model significantly improves prior knowledge elicitation andprediction accuracy, when predicting the relative citation counts of scientificdocuments in a specific domain.
展开▼
