Customer Behaviour Hidden Markov Model

Abstract

In this work, the Customer behaviour hidden Markov model (CBHMM) is proposed to predict the behaviour of customers in e-commerce with the goal to forecast the store income. The model consists of three sub-models: Vendor, Psychology and Loyalty, returning probabilities used in the transition matrix of the hidden Markov model, deciding upon three decision-states: “Order completed”, “Order uncompleted” or “No order”. The model outputs are read by the Viterbi algorithm to estimate if the order has been completed successfully, followed by the evaluation of the forecasted store income. The proposed CBHMM was compared to the baseline prediction represented by the Google Analytics tracking system mechanism (GA model). The forecasted income computed using CBHMM as well as the GA model followed the trend of real income data obtained from the store for the year 2021. Based on the comparison criteria the proposed CBHMM outperforms the GA model in terms of the R-squared criterion, giving a 5% better fit, and with the PG value more than 3 dB higher.