Human elicited features in retail site analytics

Abstract

Location selection is indispensable for a company or industry to survive for a long term. A strategically positioned retail location can increase the profitability and draw more customers for a company. Conventionally, a good location decision is associated with the relevant and significant location factors. However, integrating human subjective opinion as part of feature engineering process can be a challenge; there is no guarantee that these features can be optimum. In this light, this paper aims to investigate the impact of additional human elicitation features on the retail site selection model. This research focuses on retail site analytics to predict the sale of a telecommunication company in Malaysia. Apart from features such as geographical information, demographics and economics, this paper also includes the features determined by domain experts to investigate whether the human elicited features could improve the accuracy of sales estimation given a specific location. The findings of current work show that the additional of human elicited features successfully increase the model accuracy by 18.22%.