No abstract

The recent applications of data mining such as biological, scientific, financial and others are changing data regularly, which is uncertain and incomplete. For finding tendency in these data up-to-date, we need to modify existing data mining algorithms with dynamic characteristics. Soft computing methods are suitable for finding changes in uncertain data. In order to adopt change in data we can apply any of two approaches, update algorithm by ignoring earlier state or update with respect to earlier state. In this paper, we have framed two fuzzy clustering methods based on these approaches and implementation done using R software with comparison.

No abstract

The term “data-drift” refers to a difference between the data used to test and validate a model and the data used to deploy it in production. It is possible for data to drift for a variety of reasons. The track of time is an important consideration. Data mining procedures such as classification, clustering, and data stream mining are critical to information extraction and knowledge discovery because of the possibility for significant data type and dimensionality changes over time. The amount of research on mining and analyzing real-time streaming data has risen dramatically in the recent decade. As the name suggests, it’s a stream of data that originates from a number of sources. Analyzing information assets has taken on increased significance in the quest for real-time analytics fulfilment. Traditional mining methods are no longer effective since data is acting in a different way. Aside from storage and temporal constraints, data streams provide additional challenges because just a single pass of the data is required. The dynamic nature of data streams makes it difficult to run any mining method, such as classification, clustering, or indexing, in a single iteration of data. This research identifies concept drift in streaming data classification. For data classification techniques, a Labelled Classifier with Weighted Drift Trigger Model (LCWDTM) is proposed that provides categorization and the capacity to tackle concept drift difficulties. The proposed classifier efficiency is contrasted with the existing classifiers and the results represent that the proposed model in data drift detection is accurate and efficient.

No abstract