[1I3-GS-2] Machine learning: Market analysis

[1I3-GS-2-03] A study on ordered logit model considering the diversity of data

Mana Takaya¹, 〇Haruka Yamashita¹ (1. Sophia University)

Keywords:ordered logit model, probabilistic latent class model, Marketing

The ordered logit model is applied in various fields as a model for estimating the weight of the explanatory factor for objective variable represented by an ordered scale. When this method is applied to large-scale data, it is sometimes difficult to represent by one model due to the diversity of the population. In this study, we focus on probabilistic latent semantic analysis (PLSA), and proposes a method for estimating parameters simultaneously with ordered logit model. As a result, when analyzing complex data that is difficult to express with a single ordered logit model, the characteristics of the data will be properly represented. In this study, the proposed model is applied to a real-world data. We analyze the data of customer's evaluation data of one item, and clarify the characteristic of customer and the factor of the change of the mental state.