Pemodelan Poisson Hidden Markov pada Infeksi Nosokomial

Abstract

Nosocomial infections are infections from the hospital. Development and spread of the infections is caused by several factors such as the immunity of microorganisms in the body of antibiotics, the spread of microorganisms between patients, the lack of hygiene in medical personnel and medical devices. If we assume that the cause of nosocomial infection is not observed directly and form a Markov chain then the number of nosocomial infections can be modeled by a Poisson hidden Markov. Poisson hidden Markov model is characterized by its parameters. Model parameters are estimated using maximum likelihood method and the expectation maximization algorithm. Poisson hidden Markov model is then applied to the data of nosocomial infections for estimating the number of nosocomial infections that occur. To simplify the process of estimating the parameters, the computation conducted by using the software Mathematica 10.0. After the model parameters are derived, observation data can be generated. The best model obtained is 3-state Poisson hidden Markov model with root mean square error .1.2953