The authors also introduced the iterative graph reconstruction process for inference in relational data, and shown that it leads to improvements in performance. To find the competitors, the authors used machine learning algorithms and probabilistic approaches. They also validate system results and deploy it on the web as a powerful analytic tool for individual and institutional investors. However, the technique has many problems like finding alliances and market demands using the machine learning approach.A formal definition of the competitiveness between two items. Authors used many domains and handled many shortcomings of previous works. In this paper, the author considered the position of the items in the multi-dimensional feature space, and the preferences and opinions of the users. However, the technique addressed many problems like finding the top-k competitors of a given item and handling structured data.
A new online metrics for competitor relationship predicting. This is based on the content, firm links and website log to measure the presence of online isomorphism, here the Competitive isomorphism, which is a phenomenon of competing firms becoming similar as they mimic each other under common market services. Through different analysis they find that predictive models for competitor identification based on online metrics are largely superior to those using offline data. The technique is combined the online and offline metrics to boost the predictive performance.
