A classification model on probabilistic semantic relation for big data: an integrated approach

Data mining is process of analyzing information repositories. As data store took shape of big data, it is difficult to find relevant patterns with current techniques. Existing framework don’t suit integration and analysis of complex scenario. This insufficiency motivates to pro-pose new solutions. The major problem with big data integration and analysis is due to complex interdependence between the changing data granularity, incompatible data models, and data contents. Hence integration and classification model based on probabilistic semantic relation (PSR) of attribute pattern for big data source is proposed. It learns interrelationships and interdependence pattern among data class and data source. This knowledge is utilized to classify probabilistic relation prediction among the pattern and source data which helps in data classification and future analysis. The model implements Data integration and mapping, Construction of knowledge base, and Naive based (NB) PSR approach. An experiment is done over real crime dataset. Measures like Precision, Recall, Fall-out rate and F-measure are calculated to evaluate results. Experiment shows average of 10% increase in precision and recalls as compared to NB classification and an average of 7% improvisation in F-measure. This improvisation suggest that proposed model can be applied to future data class prediction for various prediction task.