An Integrated Data-Driven Procedure for Product Specification Recommendation Optimization with LDA-LightGBM and QFD

Abstract

E-commerce and social media have become increasingly essential and influential for sustainable business growth, particularly due to the COVID-19 pandemic, which has permanently altered the business landscape. The vast amount of consumer data available online holds significant potential and value. The strategic utilization of this information can expedite the research and development of new products, leading to shorter product cycles and increased innovation. This study explores the effectiveness of employing the latent Dirichlet allocation (LDA) method and various deep learning technologies to predict Amazon consumer ratings. We propose a product service system that utilizes natural language analyses of online sales data and user reviews, enabling industries to quickly identify and respond to market demands. We present a data-driven procedure for the customer-to-manufacturer (C2M) business model, specifically focusing on sustainable data-driven business models based on knowledge and innovation management. This procedure analyzes user comments on online shopping platforms to match product requirements and features, optimize product values, and address issues related to product specifications and new product development planning. The results of the business verification demonstrate that this procedure accurately evaluates product specifications under different demands, facilitates effective product planning, and enhances research and development decision making. This approach, based on sustainable data-driven business models and knowledge and innovation management, expands market opportunities for the sector and improves overall production efficiency, starting from the research and development stage.