Evaluation of the Recommendation Effect of Individual Interventions in a Recommender System

Collaborative filtering is a recommendation model that evaluates and recommends items match the preferences of each user with high purchase probability. However, some are regularly purchased do not necessarily require recommendation. Thus, identifying effect one challenges when using models. The can be regarded as an intervention in causal inference by considering item intervention. individual users estimated models such counterfactual regression (CFR). Nonetheless, because this only manage type intervention, it cannot easily directly applied to recommender systems many recommended appear In study, we extend estimate effects interventions multiple single CFR combining features form covariates user–item pairs, thereby allowing for all pairs. We demonstrate effectiveness proposed method through experiments artificial data.