An Efficient Adaptive Graph Time-Variant Classification (AGTVC) Using Roughset Based Online Streaming Feature Selection Algorithm

Adaptive incremental sub-graph classification is a significant tool for examining data with structure dependence. In this paper proposed a novel Adaptive Graph Time-Variant Classification (AGTVC) Using Roughset Based Online Streaming Feature Selection algorithm mines the relevant features from the synthetic real-world social network data sets which improve the graph learning prediction accuracy than previous methods. The proposed AGTVC algorithm divide feature selection and load into memory in a mini-batch manner, which is a important reduction in memory and running time. Experiments results shows the AGTVC algorithms can decrease both the processing time and memory cost. The ROSFS algorithm learns both feature selection and possible results to converts a high dimensional problem into a big constraint problem with respect to feature vector. The experimental results are reanalyzed with several constrains such as number of dimensions versus objective, running time and accuracy. Based on the results generated on this paper, it concludes that AGTCV accuracy and performance increases compared to the previous method of ISJF algorithm.