Pemodelan Transaksi Pelanggan Menggunakan Model Hidden Markov

Abstract

A customer transaction is one of the activities needed to establish a market. This event may happen at any time in a long time series and may be repeated in the future. If the causes of customer transactions are assumed not to be observed directly and they form a Markov chain, then the customer transactions can be modeled by using the hidden Markov model. The hidden Markov model is characterized by some parameters of initial distribution, transition, and transactions. The parameters in this final assignment are estimated by using the Rabiner method, which consists of forward-backward, Viterbi, and Baum-Welch algorithms. The hidden Markov model for customer transactions is applied to transaction data of Anisa Cell’s cellular company. By using the transactions data, a computer program is created using Mathematica 7.0 and LINGO 11.0 to estimate the parameters of the hidden Markov model. The results show that the hidden Markov model can model the company’s transactions well. This can be seen from the estimation’s accuracy, which is above 96%.