Product Evaluation Prediction Model Based on Multi-Level Deep Feature Fusion

Qingyan Zhou (), Hao Li, Youhua Zhang and Junhong Zheng
Additional contact information
Qingyan Zhou: School of Information and Computer, Anhui Agricultural University, Hefei 230036, China
Hao Li: School of Information and Computer, Anhui Agricultural University, Hefei 230036, China
Youhua Zhang: School of Information and Computer, Anhui Agricultural University, Hefei 230036, China
Junhong Zheng: School of Computer Science and Technology, Zhejiang Sci-Tech University, Hangzhou 310018, China

Future Internet, 2023, vol. 15, issue 1, 1-16

Abstract: Traditional product evaluation research is to collect data through questionnaires or interviews to optimize product design, but the whole process takes a long time to deploy and cannot fully reflect the market situation. Aiming at this problem, we propose a product evaluation prediction model based on multi-level deep feature fusion of online reviews. It mines product satisfaction from the massive reviews published by users on e-commerce websites, and uses this model to analyze the relationship between design attributes and customer satisfaction, design products based on customer satisfaction. Our proposed model can be divided into the following four parts: First, the DSCNN (Depthwise Separable Convolutions) layer and pooling layer are used to combine extracting shallow features from the primordial data. Secondly, CBAM (Convolutional Block Attention Module) is used to realize the dimension separation of features, enhance the expressive ability of key features in the two dimensions of space and channel, and suppress the influence of redundant information. Thirdly, BiLSTM (Bidirectional Long Short-Term Memory) is used to overcome the complexity and nonlinearity of product evaluation prediction, output the predicted result through the fully connected layer. Finally, using the global optimization capability of the genetic algorithm, the hyperparameter optimization of the model constructed above is carried out. The final forecasting model consists of a series of decision rules that avoid model redundancy and achieve the best forecasting effect. It has been verified that the method proposed in this paper is better than the above-mentioned models in five evaluation indicators such as MSE, MAE, RMSE, MAPE and SMAPE, compared with Support Vector Regression (SVR), DSCNN, BiLSTM and DSCNN-BiLSTM. By predicting customer emotional satisfaction, it can provide accurate decision-making suggestions for enterprises to design new products.

Keywords: product evaluation prediction; Depthwise Separable Convolutions; Bidirectional Long Short-Term Memories; attention mechanism; genetic algorithm (search for similar items in EconPapers)
JEL-codes: O3 (search for similar items in EconPapers)
Date: 2023
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/1999-5903/15/1/31/pdf (application/pdf)
https://www.mdpi.com/1999-5903/15/1/31/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jftint:v:15:y:2023:i:1:p:31-:d:1029557

Access Statistics for this article

Future Internet is currently edited by Ms. Grace You

More articles in Future Internet from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().