The turning points of housing prices play a significant role in the real estate market and economy. However, because multiple factors impact the market, the prediction of the turning points of housing prices faces significant challenges. To solve this problem, in this study, a historical data-based model that incorporates a multi-population genetic algorithm with elitism into the log-periodic power law model is proposed. This model overcomes the weaknesses of multivariate and univariate methods that it does not require any external factors while achieving excellent interpretations. We applied the model to the case study collected from housing prices in Wuhan, China, from December 2016 to October 2018. To verify its reliability, we compared the results of the proposed model to those of the log-periodic power law model optimized by the standard genetic algorithm and simulated annealing, the results of which indicate that the proposed model performs best in terms of prediction. Efficiently predicting and analyzing the housing prices will help the government promulgate effective policies for regulating the real estate market and protect home buyers.

Fine-grained subjective partitioning of urban space using human activity flows reveals actual human activity spaces with high resolution, which has great implications for the development and validation of planning strategies. This paper presents a new method for fine-grained subjective partitioning of urban space based on the combination of network analysis and human interactions from social media. Three main procedures are involved in this method: 1) a cut-off point for hierarchical partitioning is determined by fitting the probability distribution function of human activity patterns; 2) based on this cut-off point, improved hierarchical weighted-directed spatial networks are constructed by incorporating a gravity model into conventional spatial networks to take into account the importance of the attraction of nodes in shaping urban space; and 3) the hierarchical and fine-grained partitioning results, which reveal the actual human activity spaces with high resolution at multiscale are obtained by implementing a spatial community detection algorithm in these networks. A case study, using a real-world dataset from the capital of China validates the effectiveness of the proposed method. By analyzing the results from Beijing, we concluded that the social media, a gravity model, and the hierarchical subjective communities detected from the hierarchical human activity networks are all outstanding contributors to the fine-grained subjective partitioning of urban spaces.INDEX TERMS Gravity model, hierarchy, network analysis, social media, subjective partitioning.

No abstract