Author(s)

D. CIRNE

Abstract

In this paper, we explore an algorithm to determine the relevance of each item in a finite set of items in reference to each other to address an item you have to first go through a convergence or proxy item. If we imagine a media streaming company (convergence item) and all its available genres for playback (items in a finite set), how relevant is each music genre at different moments in time? Or with a sports broadcasting company and the covered sports, how does the relevance of each sport changes throughout the year as sports seasons begin and end?

As the algorithm gets developed in the paper, we introduce an artificial node to the relationship graph, which brings a disproportional weight in importance. Later, we show how to remove the artificial node from the final rank vector to obtain a ranking of items in the set without any distortions.

In addition to the ranking of a single point in time, the algorithm expands to analyze a sequence of consecutive convergence rankings, infers the behavior of trends and allows for the forecasting of near-future or cyclical ranks.

The applications of this algorithm are immediate and plentiful  in possibilities. In its essence, this algorithm can help to understand the usage behavior of services by its users based on non-invasive simple data collection. From its results, it is possible to better plan the allocation of resources.

Keywords

algorithm, big data, convergence, eigenvector, graph, math, rank, reduce, stochastic