An Hotspot Prediction Approach of Scientific Research Based on Autonomously Evolutionary Learner

Abstract

Hotspot prediction of scientific research is a new application in information domain. Big-data analytics technique is a critical way of improving prediction efficiency. Scientific researching data have a high dimension and are from different domains. Just using a single learner is hard to solve the problem. In the paper we adopt several deep neural networks to find the proper data dimension and network structure. The genetic algorithm is introduced as well to fasten the whole prediction. The proposed approach performs well in accuracy ratio and recall ratio compared with the benchmark algorithms. The genetic algorithm and pruning method largely improve its performance as well.