| 摘要 |
PURPOSE:To decrease the number of dimensions of a characteristic vector and to effectively sort the vectors by converting the characteristic vector used for analysis of a hierarchical cluster into another characteristic vector that is more easily clustered. CONSTITUTION:A characteristic vector Vi is obtained to each element Ei included in a gathering. Then a model element is selected for each element Ei of the gathering and the resemblance or the distance is calculated between the model element and a model element Mj selected for each element Ei. Thus the j-th element is obtained for a new characteristic vector Ui of the element Ei. As a result, a secondary characteristic vector is obtained to the element Ei included in the gathering. Then a more suitable sorting operation of vactors is assured with the secondary characteristic vector since the experiential knowledge of the secondary characteristic vector is reflected to a primary characteristic vector. Furthermore the secondary characteristic vector is handled as the analysis of a hierarchical cluster and therefore the number of dimensions of a characteristic vector can be decreased. Then the vectors are effectively sorted. |
