Analysis of Thai text from social media for mass customization

Abstract

Knowledge extraction from social media data is a crucial problem when the derivatives of Thai text are concerned, such as using the wrong word, miss-spelling, and using various symbols. Due to the speed and competitiveness of business operations and the dynamic change of customer behaviour, extracting the core and optional attributes from customers was the research objective. One solution for supporting mass customization marketing (MC) is to identify the core attributes and optional attributes from customers posted via social media since the customers should express their needs, wants and demand in their message, which led to the factual data. To analyse these data, the principle of Text Mining and Regular Expression was used at the word level. Additionally, conditional probability was used to identify the significant words for supporting MC on the core and optional product components. The samples were collected from 1,400 posted data in the year 2019 from the Facebook fan page of Swensen’s ice cream because the brand is a popular brand, and there are many branches in Thailand. According to the objective, the results showed that the related words to MC production with the conditional probability of the season were as high as 0.66 and 0.48 for summer and winter products, respectively. The core attribute “durian ice cream” had a probability of 0.548, while the optional attributes for ice cream in winter were strawberry and sticky rice in the summer. Finally, the firm can use the findings for supporting product design as customized style.