Machine Learning-Based Color Recommendation Engines for Enhanced Customer Personalization

Abstract

The aim of this research is to explore advanced color recommendation engines to enhance customer personalization based on machine learning. As e-commerce continues to grow, understanding the customer is the most difficult task for the owners. The color match value among garment products is essential in deciding whether products are displayed together, which also affects customer loyalty and purchase intention massively. Moreover, retail owners usually do not take advantage of their historical data to effectively find out what color might better suit their customer. In this research work, three main algorithms of color matching are developed. The effectiveness of the proposed color matching approaches is evaluated via several experiments on real-world datasets.
In the recommendation process, a currently purchased item of a user could match or match better deference with respect to its property colors. Nevertheless, many e-commerce platforms still treat apparel colors as none the like other properties, inappropriately display products independently, which may confuse customers and therefore lose both of sales and brand loyalty. Accurate prediction of mismatch garment colors enhances customer satisfaction and timely profitable product display. A data-driven personalized smart lighting recommender system, employing both supervised and unsupervised learning, is proposed. Two recommendation approaches are developed: color scheme recommendation and routine recommendation. The unsupervised recommendation of routine for lighting is modeled as a cluster problem given a large number of behavioral log data. Moreover, by analyzing users' daily patterns, geographical location, temporal and usage information, the color preference of users is predicted.