DATA ANALYTICS BASED TOPIC MODELLING COMBINED WITH KANSEI ENGINEERING FOR PRODUCT DEVELOPMENT

Abstract

Customer review has major impact to the reviewed product maker. Some of which convince customer whether to buy the product or not. This review also contains a lot meaning which can be exhausting from the perspective of product’s maker, to understand it one by one. One of the mediums to access this review is Amazon, which includes many items to be reviewed, either to check the review or to buy a product. One of the products is Samsung S9 smartphone, which is discussed by a lot of users on the website. Later, these reviews are downloaded and analysed by using LDA (Latent Dirichlet Allocation). Kansei engineering is also employed on the analysis to provide product improvement guideline. These processes will also be held on other reviews dataset which involve other smartphone brands, to be used as the comparison to the Samsung S9 review dataset in terms of the findings. The result shows that LDA can summarize the features reviewed from both dataset such as Bixby feature, Battery life, Picture, Quality, Screen, Camera, Speaker, etc. While, the Kansei words findings based on the word2vec library implementation showcase some significant reviews based on the co-occurrence percentage towards the features. Some examples like the Bixby button feature has Kansei words based on the reviews, such as annoying, large, edge, etc. The final result is summarized and visualized with product improvement guidelines which comes from the features and Kansei words analysis from both dataset and comparison analysis from both results.