Heterogeneous Weighted Voting-Based Ensemble (HWVE) for Root-Cause Analysis

Related Works

Root-cause identification had been a very recent technique, gaining prominence with the increased usage of the Web and its related components. However, sentiment analysis or opinion mining and aspect identification which are major components of root-cause identification architecture have been studied by the research community for a while. This section presents some of the recent and prominent contributions in these domains.

The earliest and most prominent contribution in the domain of aspect extraction was by Qiu (2011). The method called Double Propagation (DP) proposed in this work is one of the most prominent works in this domain. This work proposes a rule based technique based on bootstrapping, that enables effective identification of aspects from the given text. A summarization based model proposed by Hu (2004) has been another earlier works, that enables identification of principle components in text. This model was further improved by Popescu (2005) to extract product features and their corresponding opinions from product reviews. Another similar product feature extraction model proposed by Scaffifi (2007) uses language model to identify product features. A deep learning model for aspect identification was proposed by Poria (2016). This model uses convolutional neural networks to map language intricacies and provide predictions. An analysis of the effectiveness of aspect extraction in the process of sentiment analysis was analyzed in the work proposed by Rana (2016). This work was followed up by Rana et al. a two-fold rule based model for aspect extraction (Rana, 2017).