The matrix of transition probabilities between rating classes is a popular approach for predicting the future credit rating. This paper instead attempts to predict the future credit rating using a non-Markovian model. The prediction is done via the probability of the future credit rating given the ratings in the present and previous quarters. The estimation of the conditional probability of future credit rating is carried out by means of simulation after fitting the data with a multivariate power-normal distribution. The results based on the quarterly credit ratings of ten companies over 15 years taken from the database of the Taiwan Economic Journal indicate the need of extending the Markovian model to the non-Markovian model.