Public transport, a vital part of transportation systems, plays an essential role in our society to reduce many social issues. however, the issue of mismatching in operation and management in public transport causes passenger dissatisfaction and waste of energy. Due to the use of modern technologies in data collection and data analysis, it is possible to make online predictions on real-time observations to respond to upcoming situations. The online demand prediction is important in facilitating effective operation and management to improve the services in public transport. However, data quality issue impacts the accuracy of results for practical deployment. Most recent studies propose accurate online prediction models relying on data preprocessing instead of enhancing the robustness to resist noises. This paper put forward a Pattern Recognition Prediction based on PCA (PRP-PCA) model which delivers high robustness with stable eigen demand images. In the case study, a comparative analysis is performed with data in Stockholm on the PRP-PCA models at different prediction intervals, and other two benchmark models: Clustering based PCA method and history average method in terms of accuracy, transferability and robustness under different designed experimental settings. The result demonstrated that PRP-PCA model outperformed the other benchmark models in different scenarios and the predictions were very close to the real demands. In terms of robustness, PRP-PCA model could tolerate a certain extent of noise. For instance, the model was robust to resist the missing data up to 50% missing with the noise level between 10%-20%. Moreover, the thesis found that the eigen demand images had a relationship with the station distributions and the Covid-19 did not influence the hidden patterns of a transport system. Overall, the study could drive many applications in online prediction problems with data quality issues to improve transportation operation and management and the approach could be used for other predictions, such as traffic flows and travel times.