A Recommendation Algorithm using Adaptive Aggregation of Binary Ratings

Abstract

In this paper, we describe a novel application area of recommendation systems; helping people with disabilities find accessible Point-of-Interest (POI) using binary ratings on various accessibility criteria based on crowd-sourced data. We discuss an adaptive aggregation technique based on time fading aggregation for binary rating stream to predict the current state or confidence of each accessibility criteria for POIs. The confidence along with the user profile is used to calculate personalized accessibility score for the POI. The proposed method can be used with other POI recommendation criteria to recommend accessible places to users. We evaluate our model using synthesized datasets of different size that simulate the change of accessibility confidence over time. On comparison of the results with widely used adaptive, as well as non-adaptive aggregation techniques, we found that the proposed technique significantly improves the accuracy.