The prediction of travel times is challenging because of the sparseness of real-time traffic data and the intrinsic uncertainty of travel on congested urban road networks. We propose a new gradient–boosted regression tree method to accurately predict travel times. This model accounts for spatiotemporal correlations extracted from historical and real-time traffic data for adjacent and target links. This method can deliver high prediction accuracy by combining simple regression trees with poor performance. It corrects the error found in existing models for improved prediction accuracy. Our spatiotemporal gradient–boosted regression tree model was verified in experiments. The training data were obtained from big data reflecting historic traffic conditions collected by probe vehicles in Wuhan from January to May 2014. Real-time data were extracted from 11 weeks of GPS records collected in Wuhan from 5 May 2014 to 20 July 2014. Based on these data, we predicted link travel time for the period from 21 July 2014 to 25 July 2014. Experiments showed that our proposed spatiotemporal gradient–boosted regression tree model obtained better results than gradient boosting, random forest, or autoregressive integrated moving average approaches. Furthermore, these results indicate the advantages of our model for urban link travel time prediction.

There is a growing interest in the interoperability between building information models (BIM) and geographic information systems (GIS). BIM can provide rich information for the generation of a GIS model, while the analysis tools in GIS can be used by BIM applications, such as the management of material supply chains and site selection. A comprehensive integration of the BIM and GIS is critical for interoperability between the two domains. However, many integration methods including manual inspection of complex and heterogeneous model files are time-consuming and have weak generalisation capability. This paper proposes a new linguistic-based semantic mapping method that uses the letter trigram-based word hashing method (WHSMM) to construct the feature vector, and then maps the Industry Foundation Classes (IFC) and City Geography Markup Language (CityGML) schemas through text mining techniques at schema level. The semantic mapping precision and recall obtained using WHSMM are compared with those of another linguistic-based method. The mapping result shows that performance of WHSMM is superior to that of another method in most cases. Therefore, the mapping candidates of WHSMM are used to transform the IFC datasets into CityGML models. The transformation results show that most of mapping candidates of the WHSMM can be used to practical data integration. The semantic mapping method presented in this paper can help map the IFC and CityGML schemas by effectively reducing the search space of the manual mapping method.

The capsule network (Caps) is a novel type of neural network that has great potential for the classification of hyperspectral remote sensing. However, the Caps suffers from the issue of gradient vanishing. To solve this problem, a powered activation regularization based adaptive capsule network (PAR-ACaps) was proposed for hyperspectral remote sensing classification, in which an adaptive routing algorithm without iteration was applied to amplify the gradient, and the powered activation regularization method was used to learn the sparser and more discriminative representation. The classification performance of PAR-ACaps was evaluated using two public hyperspectral remote sensing datasets, i.e., the Pavia University (PU) and Salinas (SA) datasets. The average overall classification accuracy (OA) of PAR-ACaps with shallower architecture was measured and compared with those of the benchmarks, including random forest (RF), support vector machine (SVM), 1-dimensional convolutional neural network (1DCNN), two-dimensional convolutional neural network (CNN), three-dimensional convolutional neural network (3DCNN), Caps, and the original adaptive capsule network (ACaps) with comparable network architectures. The OA of PAR-ACaps for PU and SA datasets was 99.51% and 94.52%, respectively, which was higher than those of benchmarks. Moreover, the classification performance of PAR-ACaps with relatively deeper neural architecture (four and six convolutional layers in the feature extraction stage) was also evaluated to demonstrate the effectiveness of gradient amplification. As shown in the experimental results, the classification performance of PAR-ACaps with relatively deeper neural architecture for PU and SA datasets was also superior to 1DCNN, CNN, 3DCNN, Caps, and ACaps with comparable neural architectures. Additionally, the training time consumed by PAR-ACaps was significantly lower than that of Caps. The proposed PAR-ACaps is, therefore, recommended as an effective alternative for hyperspectral remote sensing classification.