Live Data Stream Classification for Reducing Query Processing Time

The problem of data analysis and making decisions are increases with the volume of data. In other words, processing of large data requires large resources to process and providing the final response. The big data is environment which is used for the large data processing and their analytics. But when the traffic is slow and block size of data is larger than the query response is generated with the significant amount of delay. In order to optimize the delayed response, need to make some effort for improving the performance of the big data systems. In this presented work a new technique for solving this delayed data response a streamed data mining based technique is proposed. The proposed technique contributes for demonstration of the live twitter stream gathering, pre-processing of data and transformation of the unstructured data into the structured data features, classification of data streams using the ensemble learning concept for streamed text data and finally the performance study of the developed data model. By using the given model the system improves the query processing time and produces response in less time even when a single pattern is appeared for the query processing. The technique's implementation requires twitter data API, Strom infrastructure implementation, ensemble learning technique, Hadoop technology and the java technology. The outcomes of the given model using the experimental observations are concluded in two different parts first based on the classification performance of streamed data and second the utilized time for generating the query response. According to both the scenarios the proposed technique is found optimal and their performance is adoptable for classifying different kinds of text data on live streams.