Wine recommendation algorithm based on partitioning and stacking integration strategy for Chinese wine consumers

This study tries to propose a wine recommendation algorithm based on partitioning and Stacking Integration Strategy for Chinese wine consumers. The approaches follow the idea of partitioning, decomposing traditional recommendation task into several subtasks according to wine attributes, using neural...

Full description

Saved in:
Bibliographic Details
Published in:Italian journal of food science Vol. 34; no. 2; pp. 98 - 109
Main Authors: Mu, Weisong, Feng, Yumeng, Shu, Haojie, Wang, Bo, Tian, Dong
Format: Journal Article
Language:English
Published: Pinerolo Codon Publications 01.01.2022
Subjects:
Accuracy
Algorithms
Aroma
Artificial intelligence
Classification
Consumers
Customer satisfaction
Customer services
Decision trees
Education
Integration
Machine learning
Neural networks
Packaging
Partitioning
Preferences
Regression models
Stacking
Support vector machines
Sweetness
Variables
Wines
China
ISSN:1120-1770, 1120-1770, 2239-5687
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study tries to propose a wine recommendation algorithm based on partitioning and Stacking Integration Strategy for Chinese wine consumers. The approaches follow the idea of partitioning, decomposing traditional recommendation task into several subtasks according to wine attributes, using neural network, support vector machine (SVM), decision tree, random forest, optimized random forest, Adaboost and XGBoost as recommendation models. Then, based on Stacking integration method, five models are screened out for each recommendation index as the base classifier, and the decision tree or logistic regression model is selected as the meta-learner to construct a two-layer Stacking integration framework. Finally, the optimal recommendation algorithm be built for recommendation subtasks according to the prediction accuracy. The result showed that the Stacking integrated recommendation model was suitable for the recommendation of eight attributes including colour, sweetness, foamability, mouthfeel, aroma type, year, packaging and brand, while SVM model was suitable to recommend aroma concentration and price, and the XGboost model was most appropriate for origin. This study would subserve consumers to choose the wine more easily and conveniently and provide support for wine companies to improve customer satisfaction with consumer services. The study expands the approach of concerning research and proposes a specific multi-model recommendation strategy based on artificial intelligence models to recommend multiattribute commodities.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1120-1770
1120-1770
2239-5687
DOI:10.15586/ijfs.v34i2.2209