Induction algorithm based on statistics theory with Delphi

Analysis by way of the experiment, compared with ID3 algorithm, there is a large difference between SD-CA algorithm in this thesis and decision tree algorithm originated from ID3 And the classification rule is also different from the practice. It relates to the data containing middling, the proportions are all 0.5. Then, its results to classification are much more related to other attributes; some attributes´ values have the determinative effect. For instance, only if dressing index is in the attribute of normal, it must be positive. Of course, the others are of different effects. So we´d better think over entirely before come to the final result. According to the statistic probability, while the training set is increasing much more, the proportion of attribute´s positive and negative will be stable and the precise of the classification will be higher.