Iterative weighted k-NN for constructing missing feature values in Wisconsin breast cancer dataset

This paper presents a new approach for constructing missing feature values based on iterative nearest neighbors and distance metrics. The proposed approach employs weighted k nearest neighbors´ algorithm and propagating the classification accuracy to a certain threshold. The proposed method showed improvement of classification accuracy of 0.005 in the constructed dataset than the original dataset which contain some missing feature values. The maximum classification accuracy was 0.9698 on k=1. This work is a component from a research for an automated diagnosing for breast cancer. The main aim of the current paper is to prepare the dataset for mining process. Future work includes applying the proposed method on more datasets.