Importance-Performance Analysis of Product Attributes Using Explainable Deep Neural Network From Online Reviews

2021 ◽  
Author(s):  
Junegak Joung ◽  
Harrison Kim
Keyword(s):  
Neural Network ◽  
Performance Analysis ◽  
Deep Neural Network ◽  
Online Reviews ◽  
Product Attributes

Importance-Performance Analysis of Product Attributes Using Explainable Deep Neural Network From Online Reviews

2020 ◽  
Author(s):  
Junegak Joung ◽  
Harrison M. Kim
Keyword(s):  
Neural Network ◽  
Performance Analysis ◽  
Network Architecture ◽  
Deep Neural Network ◽  
Online Reviews ◽  
Product Attributes ◽  
Neural Network Architecture ◽  
Single Neural Network

Abstract Importance-performance analysis (IPA) is a technique used to understand customer satisfaction and improve the quality of product attributes. This study proposes an explainable deep-neural-network-based method to carry out IPA of product attributes from online reviews for product design. Previous works used shallow neural network (SNN)-based methods to estimate importance values, but it was unclear whether the SNN is an optimal neural network architecture. The estimated importance has high variability by a single neural network from a training set that is randomly selected. However, the proposed method provides importance values with a lower variance by improving the importance estimation of each product attribute in the IPA. The proposed method first identifies the product attributes and estimates their performance. Then, it infers the importance values by combining explanations of the input features from multiple optimal neural networks. A case study on smartphones is used herein to demonstrate the proposed method.

Approach for Importance–Performance Analysis of Product Attributes From Online Reviews

2021 ◽  
Vol 143 (8) ◽  
Author(s):  
Junegak Joung ◽  
Harrison M. Kim
Keyword(s):  
Neural Network ◽  
Performance Analysis ◽  
Sentiment Analysis ◽  
High Performance ◽  
Latent Dirichlet Allocation ◽  
Online Reviews ◽  
Product Attributes ◽  
Ibm Watson ◽  
Promising Source

Abstract The importance–performance analysis (IPA) is a widely used technique to guide strategic planning for the improvement of customer satisfaction. Compared with surveys, numerous online reviews can be easily collected at a lower cost. Online reviews provide a promising source for the IPA. This paper proposes an approach for conducting the IPA from online reviews for product design. Product attributes from online reviews are first identified by latent Dirichlet allocation. The performance of the identified attributes is subsequently estimated by the aspect-based sentiment analysis of IBM Watson. Finally, the importance of the identified attributes is estimated by evaluating the effect of sentiments of each product attribute on the overall rating using an explainable deep neural network. A Shapley additive explanation-based method is proposed to estimate the importance values of product attributes with a low variance by combining the effect of the input features from multiple optimal neural networks with a high performance. A case study of smartphones is presented to demonstrate the proposed approach. The performance and importance estimates of the proposed approach are compared with those of previous sentiment analysis and neural network-based method, and the results exhibit that the former can perform IPA more reliably. The proposed approach uses minimal manual operation and can support companies to take decisions rapidly and effectively, compared with survey-based methods.

scholarly journals AN AUTOMATED METHOD TO CONDUCT IMPORTANCE-PERFORMANCE ANALYSIS OF PRODUCT ATTRIBUTES FROM ONLINE REVIEWS - AN EXTENSION WITH A CASE STUDY

2021 ◽  
Vol 1 ◽  
pp. 417-426
Author(s):  
Kangcheng Lin ◽  
Harrison Kim
Keyword(s):  
Performance Analysis ◽  
Latent Dirichlet Allocation ◽  
Product Information ◽  
Online Reviews ◽  
Product Attributes ◽  
Attribute Importance ◽  
Online Data ◽  
Automated Method ◽  
Ibm Watson

AbstractWith the growth of online marketplaces and social media, product designers have been seeing an exponential growth of data available, which can serve as an extremely valuable source of information communicated from customers without geographical limitations. The data will reveal customers’ preferences, which can be expensive and slow to obtain via traditional methods such as survey and questionnaires. While existing methods in the literature have been proposed to extract product information and make inference from online data, they have limitations, especially in providing reliable results and in dealing with data sparsity. Therefore, this paper proposes a method to conduct an Important-performance analysis from online reviews. The major steps of this method involve using latent Dirichlet allocation (LDA) to identify product attributes, using IBM Watson Natural Language Understanding tool to perform aspect-based sentiment analysis, and using XGBoost model to infer product attribute importance from the collected dataset. In our case study, we have collected over 150,000 text reviews of more than 3,000 laptops from Amazon.

Sign in / Sign up

Export Citation Format

Share Document