使用動作模式感知網路的多行為推薦系統

Abstract

本研究提出了APANet 模型，一個使用動作模式感知網路的多行為推薦系統，利用多任務學習演算法將物品序列和動作序列的資訊同時加入推薦系統，以預測用戶的下一個互動物品和其對應動作。在過去 基於會話的多行為推薦算法中存在兩個主要的限制: 首先，現有方法將物品序列和動作序列分開考慮，因此無法有效且準確的描述動作序列和單一物品的依賴關係。其次，大多數現有的模型最終預測僅限於下一個物品，而忽略了預測物品的對應動作。為了克服這兩個限制，本研究設計了一個基於物品之動作模式序列，並使用條件神經網路學習給定物品種類之下的用戶行為意圖。因此，所提出的模型能夠對於一匿名的序列短期資料，預測用戶的下一個互動（包括物品和動作）。在三個公開資料集上的實驗結果證明了所提出的APANet 模型作為基於會話之多行為推薦模型的有效性。

This study proposes APANet, a novel session-based recommendation algorithm with Action Pattern-Aware Networks (APANet) to incorporate both historical item sequences and reformulated item-wise action patterns into the modeling process. Previous multi-behavior-based approaches for session-based recommendation have two major drawbacks: First, the final prediction for most existing multi-behavior-based approaches is limited to the next item, ignoring which action the predicted item is associated with. Second, existing approaches consider item sequences and action sequences individually and thus do not explicitly and accurately model the action dependencies for a single item. To overcome the two limitations, this study designs an itemwise action pattern, and model the action-level intent representation given the next item’s context through a conditional network. Therefore, the proposed model enables us to predict the next-best interaction (i.e., next-best item and its associated action) given a short-term anonymous multi-behavior sequence. Comprehensive experiments on three public benchmark datasets demonstrate the effectiveness of the proposed APANet for multi-behavior session-based recommendation.