Efficient urban transportation systems are widely accepted as essential infrastructure for smart cities, and they can highly increase a city°s vitality and convenience for residents. The three core pillars of smart cities can be considered to be data mining technology, IoT, and mobile wireless networks. Enormous data from IoT is stimulating our cities to become smarter than ever before. In transportation systems, data-driven management can dramatically enhance the operating efficiency by providing a clear and insightful image of passengers° transportation behavior. In this article, we focus on the data validity problem in a cellular network based transportation data collection system from two aspects: Internal time discrepancy and data loss. First, the essence of time discrepancy was analyzed for both automated fare collection (AFC) and automated vehicular location (AVL) systems, and it was found that time discrepancies can be identified and rectified by analyzing passenger origin inference success rate using different time shift values and evolutionary algorithms. Second, the algorithmic framework to handle location data loss and time discrepancy was provided. Third, the spatial distribution characteristics of location data loss events were analyzed, and we discovered that they have a strong and positive relationship with both high passenger volume and shadowing effects in urbanized areas, which can cause severe biases on passenger traffic analysis. Our research has proposed some data-driven methodologies to increase data validity and provided some insights into the influence of IoT level data loss on public transportation systems for smart cities.