Weakly Supervised Critical Nugget Finding Algorithm for Improving Classification Accuracy

Accuracy is more important in data classification. An elimination of noisy data, attribute and property discovery is a major consideration in the proposed research. From the overall given population the system predicts the nuggets effectively. The subpopulation and exceptional property pair which is known as outliers. With the aim of effective critical nuggets detection, the proposed WS-CNF algorithm applies a provisional model which identifies the exceptional property pair with the scoring method implementation. Several outlier detection methods have been introduced with certain domains and applications, but they were more generic and affected by subset detection problem. The proposed concept effectively implements DBD (Data Boundary Detection) model based approach which is used for improving the classification accuracy by extending the boundary values by various iterations, the collection of these have named as Weakly Supervised Critical Nugget Finding algorithm and primary direction algorithm for the detection of sub population scores for both numerical and categorical datasets. Also the system performs the classification method in order to find best class based on the score and label. Finally, the proposed algorithm can reduce the computation cost and lack of accuracy problem by applying best data mining and suitable pruning techniques. The experiments and the results provide the mild and extreme outlier ranges with score values. Keywords—Outliers, Outlier detection, data mining, classification accuracy, Principle component analysis.