Direct mining of discriminative patterns for classifying uncertain data

An Associative Classifier for Uncertain Datasets

“…Recently, a considerable amount of studies in machine learning are directed toward the uncertain data classification, including: TSVC [7] (inspired by SVM), DTU [8] (decision tree), UNN [9] (based on Neural Network), a Bayesian classifier [10], uRule [11] (rule based), uHARMONY [12] and UCBA [13] (based on associative classifiers). However, models suggested by the previous work do not capture some possible types of uncertainty.…”

“…and gives a close estimation of the classifier performance in the real world problems. We selected the same datasets as in [12] to compare our method with the results reported in their paper for uHARMONY, uRule and DTU. This also ensures that we did not choose only the datasets on which our method performs better.…”

“…Table 1 provides a comparison between UAC and other existing rule based methods in terms of accuracy. The reported accuracies for uHARMONY (#3), DTU (#4) and uRule (#5) are reproduced from [12]. We applied UAC (#2) to the same datasets generated by the same procedure of adding uncertainty as [12] to make the comparison meaningful.…”

“…The reported accuracies for uHARMONY (#3), DTU (#4) and uRule (#5) are reproduced from [12]. We applied UAC (#2) to the same datasets generated by the same procedure of adding uncertainty as [12] to make the comparison meaningful. Value N/A, existing in the experiments reported by [12], shows that the classifier has run out of resources in their experiments.…”

“…We applied UAC (#2) to the same datasets generated by the same procedure of adding uncertainty as [12] to make the comparison meaningful. Value N/A, existing in the experiments reported by [12], shows that the classifier has run out of resources in their experiments. In Table 1, uncertainty level is U10@4 meaning that datasets have 10 percent uncertainty, where only four of the attributes with the highest information gain are uncertain.…”