M²: Mixed Models With Preferences, Popularities and Transitions for Next-Basket Recommendation

Next-basket recommendation considers the problem of recommending a set of items into the next basket that users will purchase as a whole. In this paper, we develop a novel mixed model with preferences, popularities and transitions (<inline-formula><tex-math notation="LaTeX">$\mathop {\mathtt {M^2}}\limits$</tex-math><alternatives><mml:math><mml:msup><mml:mi mathvariant="monospace">M</mml:mi><mml:mn mathvariant="monospace">2</mml:mn></mml:msup></mml:math><inline-graphic xlink:href="ning-ieq1-3142773.gif"/></alternatives></inline-formula>) for the next-basket recommendation. This method models three important factors in next-basket generation process: 1) users&#x2019; general preferences, 2) items&#x2019; global popularities and 3) transition patterns among items. Unlike existing recurrent neural network-based approaches, <inline-formula><tex-math notation="LaTeX">$\mathop {\mathtt {M^2}}\limits$</tex-math><alternatives><mml:math><mml:msup><mml:mi mathvariant="monospace">M</mml:mi><mml:mn mathvariant="monospace">2</mml:mn></mml:msup></mml:math><inline-graphic xlink:href="ning-ieq2-3142773.gif"/></alternatives></inline-formula> does not use the complicated networks to model the transitions among items, or generate embeddings for users. Instead, it has a simple encoder-decoder based approach (<inline-formula><tex-math notation="LaTeX">$\mathop {\mathtt {ed\text{-}Trans}}\limits$</tex-math><alternatives><mml:math><mml:mrow><mml:mi mathvariant="monospace">ed</mml:mi><mml:mtext mathvariant="monospace">-</mml:mtext><mml:mi mathvariant="monospace">Trans</mml:mi></mml:mrow></mml:math><inline-graphic xlink:href="ning-ieq3-3142773.gif"/></alternatives></inline-formula>) to better model the transition patterns among items. We compared <inline-formula><tex-math notation="LaTeX">$\mathop {\mathtt {M^2}}\limits$</tex-math><alternatives><mml:math><mml:msup><mml:mi mathvariant="monospace">M</mml:mi><mml:mn mathvariant="monospace">2</mml:mn></mml:msup></mml:math><inline-graphic xlink:href="ning-ieq4-3142773.gif"/></alternatives></inline-formula> with different combinations of the factors with 5 state-of-the-art next-basket recommendation methods on 4 public benchmark datasets in recommending the first, second and third next basket. Our experimental results demonstrate that <inline-formula><tex-math notation="LaTeX">$\mathop {\mathtt {M^2}}\limits$</tex-math><alternatives><mml:math><mml:msup><mml:mi mathvariant="monospace">M</mml:mi><mml:mn mathvariant="monospace">2</mml:mn></mml:msup></mml:math><inline-graphic xlink:href="ning-ieq5-3142773.gif"/></alternatives></inline-formula> significantly outperforms the state-of-the-art methods on all the datasets in all the tasks, with an improvement of up to 22.1&#x0025;. In addition, our ablation study demonstrates that the <inline-formula><tex-math notation="LaTeX">$\mathop {\mathtt {ed\text{-}Trans}}\limits$</tex-math><alternatives><mml:math><mml:mrow><mml:mi mathvariant="monospace">ed</mml:mi><mml:mtext mathvariant="monospace">-</mml:mtext><mml:mi mathvariant="monospace">Trans</mml:mi></mml:mrow></mml:math><inline-graphic xlink:href="ning-ieq6-3142773.gif"/></alternatives></inline-formula> is more effective than recurrent neural networks in terms of the recommendation performance. We also have a thorough discussion on various experimental protocols and evaluation metrics for next-basket recommendation evaluation.