Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-18 Cooperative journals: 《计算机应用研究》
Abstract: At present, the simplified models for a single building consist of simple walls and roof structures. For building groups, these methods ignore the characteristics of adjoining buildings and the interconnected relationship with the surrounding features. Based on the characteristics of the roof structures, this paper proposed a method which can automatically generate the structural features of the building. It divided the building into different structural parts by the adjacency of the bottom plan. For each structural part of the partitions, it reconstructed the building model by using a top-down projection method that determining the top structural features. Finally, it combined the parts to form a simplified model. The experimental results show that the method can make full use of the topology of buildings and describe the model with the least amount of data for mobile applications, while preserving the top and wall features of buildings.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-24 Cooperative journals: 《计算机应用研究》
Abstract: For the low accuracy of classification caused by the lack of labeled data, and the problem of tedious feature extraction of the traditional time series classification method, this paper analyzed the characteristics of BP neural network and Naive Bayes classifier, it proposed a method based on BP and Naive Bayes. Is used the nonlinear mapping ability of BP neural network and the classification ability of Naive Bayes classifier under a small amount of labeled data, it input into the features extracted from BP neural network Naive Bayes classifier, which could solve the problem of traditional time series classification algorithm. Experimental results show that this model has higher classification accuracy in the classification of time series with fewer labeled data.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-19 Cooperative journals: 《计算机应用研究》
Abstract: Point-of-interest (POI) recommendation is an important service in location-based social networks (LBSNs) . For the current recommendation algorithm exists the problems of the noise data affects the recommended quality and low level of user personalization. Motived by this, this paper proposed a personalized joint recommendation algorithm (JRA) . JRA initially utilized the locality of user activity area to early filter the POIs which are impossible or less likely to be a result. For the received preliminary candidate set, then it also considered consider category factor and the popularity factor of POI, and the social behavior of the user to further improve the user experience. The experiments on real Foursquare check-in dataset demonstrate that the JRA compared with the current advanced algorithm, the accuracy rate increased by 11%, recall rate increased by 8%.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-12 Cooperative journals: 《计算机应用研究》
Abstract: Aiming at the great harm caused by epileptic seizures for patients and leave enough spare time for clinical treatment, the study put forword a system which can predict the seizure in advance for people with epileptic. This method based on 21 epileptic patients and extracted permutation entropy as a feature vector which has lower algorithm complexity. Then the vector was input into the support vector machine (SVM) to train a learning model and identify the ictal samples. Taking full account of patient differences, it used voting mechanism to determine the patient's state. Finally, the method realized a real-time prediction for epileptic. The results show that this method can predict 81% of the seizures with more than 50 minutes before the onset of epilepsy, and it has a low false alarm rate. The method provides a solid foundation for theoretical research of seizure prediction system.