Article Preview
Top1. Introduction
With the ever-increasing growth of real estate industry and the diversity of house owners’ personalized preferences, people are apt to decorate their houses sufficiently to meet their respective aesthetic standards (Yuan, 2013; Barua, 2019; Yu, 2018; Chen, 2017; Beneðeld, 2019). For example, young house owners often prefer to decorate their houses according to modern decoration styles such as landscape painting style and cartoon style, while elder people may prefer great man style instead of other complicated ones. Today, house decoration has become a crucial and important industry all over the world and has provisioned people with various and diverse house decoration styles to meet the personalized requirements from different individuals.
With the gradual popularity of Web (Xue, 2020; Srinivasan, 2021; Kou, 2020; Sridharan, 2021; Wu, 2021), people are apt to publish their house decoration cases on the Web for the purpose of case sharing and attraction of readers. On one hand, such shared house decoration cases offer new house owners more case references or decoration experiences, and on the other hand, place a heavy burden on the decoration style or art selection, because there are so many available house decoration cases generated from past decoration behaviors of historical users. For example, Fig.1 presents three kinds of decoration art works with distinct decoration styles (e.g., Space Formation with different elements, A Single Boat floating on a river, Night with static moon and stars).
Figure 1. Different decoration styles or arts
In this situation, a challenge is raised to help new house owners to find out their interested or preferred decoration arts or styles accurately and quickly. Recommender systems have been proven a promising way to realize the above goal. While existing recommendation solutions seldom optimize recommendation accuracy and privacy-preservation simultaneously. Inspired by this challenge, a novel decorative art pattern mining and discovery approach (abbreviated as DPMD) is put forward in this paper, which is mainly based on the Locality-Sensitive Hashing technique (LSH) (Qi, 2020; Qi et al, 2020) that is widely adopted and welcome in the domain of information retrieval. Generally, the major contributions in this research are two-fold:
- (1)
We introduce a novel LSH-based similar house owners clustering approach DPMD, to achieve time-efficient and privacy-aware decorative art work recommendations. This way, we can balance user privacy and recommendation accuracy well.
- (2)
We design a set of experiments to validate the effectiveness and efficiency of our proposal in terms of making decorative art pattern mining and discovery. Through comparisons with related approaches, we prove the feasibility of our proposed decorative art pattern mining and discovery solution for house owners.
The reminder of this paper is organized as follows. Section 2 presents a motivating example to illustrate the research value and significance of this paper. In Section 3, we elaborate the concrete steps of our proposed decorative art pattern mining and discovery solution based on the historical house decoration cases and ratings on various decoration elements. A group of experiments are designed and executed in Section 4, through which we demonstrate the effectiveness and efficiency of our proposed resolution. At last, in Section 5, we conclude this research work and point out the possible research topics in the future work.
Top2. Motivation
As Fig.1 shows, each decorative art work is often comprised of a series of decorative elements such as image, color, texture and material. Two users (i.e., house owners), i.e., Tom and Judy rated some elements of decorative arts, whose rating values are shown in Fig.2 (e.g., 5* and 4*; the larger, the better). In this situation, these rating values towards different decorative elements from historical users provide a promising basis to evaluate the decorative preferences of Tom and Judy. In this situation, to judge or evaluate whether Tom and Judy have the same decorative preferences towards houses, we need to calculate the similarity degree of Tom and Judy according to their ratings towards different elements.