Abstract: The rapid increase in the number of elevators in China demonstrates the indispensable role in daily life. We propose a prediction model based on the residual error correction model and the Markov chain for forecasting the number of elevator failures. First, we developed a traditional grey model (GM) (1,1) by using data from January to October; we then used the residual errors to correct the GM(1,1), thereby creating an error-correction GM (C-GM). The experimental results indicated that after two corrections, this model achieved a variance ratio C was 0.26, and a small error probability P was 1.0. Finally, we created a Markov-C-GM by combining the C-GM with the Markov chain to predict the number of elevator failures for the subsequent 2 months. We also compared this model with the autoregressive integrated moving average (ARIMA) model regarding their ability to predict the number of elevator failures in December; the results demonstrated that the Markov-C-GM outperformed the ARIMA model in terms of its ability to process small-sample data.