An Efficient Probabilistic Multi Labeled Big Data Clustering Model for Privacy Preservation Using Linked Weight Optimization Model

Abstract

An unsupervised analysis of the classification and clustering of data is one of the most powerful and insightful data mining approaches used in different disciplines to identify homogenous groups of objects based on similarities. In machine learning with the increased generation of data, classification continues to be a key subject. While several literary works are interested in classifying the single label, the enormous dimensions of the data require a new approach. Multi-label clustering has therefore gained considerable attention in the testing community in recent years. This method involves a data instance with different labels and it is useful for many fields, e.g. image analysis, text classification and Big Data privacy security. In this case the classification of the single label is expanded. The high dimensionality of the distributed system needs an efficient and effective data management. Multi Label Classifier divides one or more labels in a set of labels of a particular instance. Multi-label classification is one of the leading data collection methods, where a set of labels is annotated in the data collection for each single instance. In one instance, the nature of multiple labels requires more computer power than classified one-label tasks. A multi-label grouping is often simplified by the method of splitting into one label classification, which avoids the distinction between labels. A Multi-Label Big Data Clustering with Privacy Protection Probability Linked Weight Optimization (MLBDC-PP-LWO) model is provided in this paper. In this proposed work, after the identification of sensitive data from data clusters, sensitive information is protected or generalized. The models proposed are compared to existing models and the findings show that the proposed model privacy preservation levels are more than the traditional methods.