Issue No. 03 - March (1984 vol. 6)
James E. Goin , Geometric Data, 999 West Valley Rd., Wayne PA 19087.
The k-nearest neighbor classifier has been used extensively in pattern analysis applications. This classifier can, however, have substantial bias when there is little class separation and the sample sizes are unequal. This classification bias is examined for the two-class situation and formulas presented that allows selection of values of k that yields minimum bias.
James E. Goin, "Classification Bias of the k-Nearest Neighbor Algorithm", IEEE Transactions on Pattern Analysis & Machine Intelligence, vol. 6, no. , pp. 379-381, March 1984, doi:10.1109/TPAMI.1984.4767533