SentiTSMixer: A Specific Model for Sales Forecasting Using Sentiment Analysis of Customer
Appropriate forecasting of sales can lead to significant revenue gains for any organization, as it allows them to plan their funding, arrange infrastructure, manage the supply chain, and anticipate profits accordingly. However, sales forecasting depends on various factors, such as product quality, m...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11003952/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Appropriate forecasting of sales can lead to significant revenue gains for any organization, as it allows them to plan their funding, arrange infrastructure, manage the supply chain, and anticipate profits accordingly. However, sales forecasting depends on various factors, such as product quality, market trends, economic conditions, competition, and customer behavior, and it has become even more challenging with the rise of online retailing. In today’s era, especially for online retailing, customer feedback plays a vital role in assessing a product’s quality, as users can express their level of satisfaction through it. Customers can share their opinions using numeric values, such as ratings, and/or through text, such as reviews. Additionally, they can express their views by voting on other reviews they find most helpful, based on their own level of satisfaction. In this research, we have modified the TSMixer model for sales forecasting by amalgamating customer satisfaction levels regarding a specific product. This enhancement allows the model to account for how customer sentiment directly influences sales performance, thereby improving the accuracy of sales forecasting. Experimental results on various types of Amazon data show that, depending on the dataset and the specific error detection techniques used, the proposed model delivers a reduction in error ranging from 65% to 99% compared to established models. |
|---|---|
| ISSN: | 2169-3536 |