Rank aggregation for non-stationary data streams
This addresses rank aggregation for dynamic preference data, which is incremental as it extends existing methods to non-stationary contexts.
The paper tackles the problem of learning the current distribution of rankings from non-stationary data streams, where preferences change over time, by proposing a generalization of the Borda algorithm and weighted voting rules, and provides bounds on sample requirements and parameter settings.
We consider the problem of learning over non-stationary ranking streams. The rankings can be interpreted as the preferences of a population and the non-stationarity means that the distribution of preferences changes over time. Our goal is to learn, in an online manner, the current distribution of rankings. The bottleneck of this process is a rank aggregation problem. We propose a generalization of the Borda algorithm for non-stationary ranking streams. Moreover, we give bounds on the minimum number of samples required to output the ground truth with high probability. Besides, we show how the optimal parameters are set. Then, we generalize the whole family of weighted voting rules (the family to which Borda belongs) to situations in which some rankings are more \textit{reliable} than others and show that this generalization can solve the problem of rank aggregation over non-stationary data streams.