Outlier Detection Among Influencer Blogs Based on off-Site Web Analytics Data

Abstract : In the current scenario, with the exponential increase in the use of internet, organizations are continuously thriving for visibility on the web. This has opened new avenues in influencer marketing. Several portals encourage these marketers to build content for the purpose of digital marketing. However, the content building process produces a lot of spam within these websites when done in bulk. This is often done in order to establish their presence by using techniques including article spinning and keyword stuffing. This study thus attempts to identify these spam websites using a dataset comprising 2751 websites using bio inspired outlier detection approaches. We use publically available key performance indicators (KPIs) through which websites that create spam content to boost the amount of text in the domain are identified. A hybrid wolf search algorithm (WSA) and bat algorithm (BA) integrated with K-means are used to classify these websites into spam. Findings indicate that metrics including Domain Authority, Page Authority, Moz Rank, Links In, External Equity Links, Spam Score, Alexa Rank, Citation Flow, Trust Flow, External Back Links, Referred Domains, SemRush URL Links and SemRush Hostname Links play an important role in identifying spam. The proposed approach may prove beneficial in segregating spam influencer websites for effective influencer marketing.