Quality-Aware Neural Complementary Item Recommendation Yin Zhang , - - PowerPoint PPT Presentation

Quality-Aware Neural Complementary Item Recommendation

Yin Zhang, Haokai Lu, Wei Niu, James Caverlee

Department of Computer Science and Engineering Texas A&M University, USA

ACM RecSys’18 :: October 3rd, 2018

Item-to-Item Recommendation

? ?

substitutes

Complementary Item Recommendation:

items that might be purchased together

Filters? Lens? Others? Bags?

Complementary Item Recommendation: Ground Truth

Bought-Together

Amazon item relationship dataset: McAuley, Julian, et al. "Image-based recommendations on styles and substitutes." SIGIR, 2015.

Also-Bought

Complementary Item Recommendation: Challenges

• 2. How to balance quality vs. complementary relationship?
• 1. How to define “complementary" distance?
• 3. How to model complex interactions?

Complementary relationship Item Quality “Bella Ladies” hoodie

Recommendation

1 star 2 star 3 star 4 star 5 star 10 20 30 40 1 star 2 star 3 star 4 star 5 star 100 200 300 400

• Previous methods rely on a single source to detect item

relationships— images [McAuley SIGIR 2015] or text [Wang WSDM 2018].

Melville, Prenn, et, al. “Content-boosted collaborative filtering for improved recommendations.” AAAI, 2002 McAuley, Julian, et al. "Image-based recommendations on styles and substitutes." SIGIR, 2015. Wang, Zihan, et al. “A Path-constrained Framework for Discriminating Substitutable and Complementary Products in E-commerce.” WSDM. 2018.

• Potential non-linear relationships between items features and quality.

Our Solution: ENCORE

ENCORE: Neural COmplementary item REcommendation Complement threshold

• 1. Detect Complementary

Items

• 2. Quality-Aware

Recommendation

• 3. Transform via

Neural Model

• 1. Detecting Complementary Items
• Influence factors

Visual Textual

USB Battery USB Flash Drive Mac Air Charger Mac Pro Charger

• Style-Based Complements:
• Functional Complements:

d j|i

(cm)(Ii,I j) =|| (mi − mj)T EM ||2 2

d j|i

(ct)(Ii,I j) =|| (ti − t j)T ET ||2 2

• Idea: Embed Style + Function

Image Feature Vector Learned Low-ranked Embedding for image Word2Vec Learned Low-ranked Embedding for text

Complementary relationship between items is influenced by style (image) and function (text) and this influence varies by items.

(a) (b) (c) (d)

Mac Pro

• 2. Quality-Aware Recommendation

Users may not choose the nearest complementary items but the highest- quality complementary items.

• Complement relationship vs Item Quality

“Bella Ladies” hoodie

(A)(B) Bella Ladies pants: the nearest complement items (C) Spandex pants

• Item Quality Estimation

Posterior Distribution

Item 1 Item 2

• 3. Neural Complementary Item Recommendation

Visual Textual Quality

Vary by Categories Items

Image Input Text Input User Ratings Image Input Text Input Image Input Text Input User Ratings Image Embedding Text Embedding

Image Distance Text Distance

Image Embedding Text Embedding

Image Distance

Non-linear Layers Non-linear Layers

Text Distance

Asymmetric Quality- aware Recommendation Asymmetric Quality- aware Recommendation

• ENCORE Framework
• Relationship

Complementary item recommendation is influenced by the complex interactions of item visual, textual and quality information.

Experiments

• How well does ENCORE perform versus baselines?
• What impact do the design choices of ENCORE have? (images,

textual information, Non-linearity)

Dataset*: Six categories in Amazon (Electronics, Cell Phones &

Accessories (C & A), Clothing, Books, Digital Music, and Movies)

* McAuley, Julian, et al. "Image-based recommendations on styles and substitutes." Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information Retrieval. ACM, 2015.

Experimental Setup: Baselines

• LRA: Logistic Regression with Average Rating
• LRB: Logistic Regression with Bayesian Rating
• WNN: Weighted Nearest Neighbor
• FNN: Feedforward Neural Network
• LMT: Low-rank Mahalanobis Transform [McAuley SIGIR 2015]
• Monomer [He ICDM 2016]
• Variations of ENCORE (see paper)

Same Inputs

Metrics: Accuracy, Precision at top-k

* McAuley, Julian, et al. "Image-based recommendations on styles and substitutes." SIGIR ACM, 2015. * He, Ruining, Charles Packer, and Julian McAuley. "Learning compatibility across categories for heterogeneous item recommendation." Data Mining (ICDM), 2016.

Experiments: Recommendation Effectiveness

ENCORE outperforms state-of-the-art methods in accuracy, precision@5 and precision@10 for most situations, especially for Electronics and Clothing categories.

Also-Bought Bought-Together

Experiments: Case Study

Query Items

Complementary Items Recommended by ENCORE

(a) (b) (c)

IdeaPad U430

Conclusions and Future Work

• Complementary relationships vary for different items. Items visual and textual

information can help find complement items.

• Users may not choose the nearest complementary items but the highest-quality
• nes. Modeling item rating distribution by Bayesian inference can improve the

accuracy and precision for complementary recommendation.

• Neural network structure in ENCORE provides improvement to the accuracy

and precision of complement item recommendation

• Future work:
• Personalized complementary item recommendation.
• Effectively model textual information to improve the quality of recommendation.

Quality-Aware Neural Complementary Item Recommendation

Yin Zhang, Haokai Lu, Wei Niu, James Caverlee

Department of Computer Science and Engineering Texas A&M University, USA

ACM RecSys’18 :: October 3rd, 2018

Thank you !