| 摘要 |
<p>PURPOSE:To obtain an appropriate clustering result when plural pairs having maximum affinities exist by updating the affinities by treating the item contained in the most numerous pairs or all of the same conditional pairs containing clusters as one cluster. CONSTITUTION:Where or not input data are a measurement matrix is discriminated and, when they are a measurement matrix, the data are converted into inter-item affinities and each item is treated as one cluster. Then the number of the clusters is checked and, when clusters exist more than two, all of the maximum affinity pairs among clusters are extracted. Thereafter, the most numerously contained item or cluster is extracted. According to the extracted result, all of the pairs having the maximum affinities including the items or clusters, namely, all of the pairs having the same condition are treated as one cluster. Successively, distance updating is repeated until the number of clusters reduced to one.</p> |
