Multi-level Network Construction Based on Intelligent Big Data Analysis

Abstract

In this work, we present a hybrid miner-network analyzer (HMNA) system includes three main stages; the first stage called preparing and preprocessing stage that includes building initial network from citation file and Find Keywords through apply Rake and cleaning. The second stage involves building classification model including parameters detection and apply LDA for find topics, final stage, add the topics of document into initial network to construction multi-layer network, where each level represent community related of that topic. We can summarize the main points of HMNA system as: (i) It deals with real, complex, huge database of papers ‘citation’. (ii) The preprocessing stage involves retrieve keywords from corpus using Rake after add constructions on it and cleaning without using any feature selection method. (iii) It building digital corpus that combines with dictionary to clustering the clean text into multi groups based on LDA model. (iv) It reconstructed the initial network by add the labeling of topic results from above step to it. (v) It builds multi communities (multi-level network), each level in that network represent single communities, (vi) It computes the main characteristics measures for each level or communities to determine the similarity in it structure with cluster in citation network.