<p>Stock price prediction is an important topic in finance and economics. Owing to the frequent occurrence of singularities in stock prices, the accuracy of stock price prediction is seriously affected. To address this issue, a novel method, which is named personalized similarity regression model based on maximum correntropy criterion (PSR-MCC), is proposed in this paper. Specifically, we first employ a similarity measurement to select series from historical data that are similar to the test series. The personalized similarity measurement embeds Canberra distances in dynamic time warping to attenuate the impact of singularity and coping with time shifts and warping. Afterward, for similar series, a new regression model based on maximum correntropy criterion is proposed. It uses maximum correntropy criterion as the constraint condition of regression models instead of the minimum mean square error for weakening the negative influence of singularities further. Experiments based on real data demonstrate that the proposed method can effectively reduce the impact of abnormal data on prediction accuracy, with small prediction errors and strong robustness.</p>