Housing Price Forecast With Attention Recurrent Neural Network

The urban economy is growing rapidly with the development of information technology.this thesis starts from the housing price prediction that residents are most concerned about as the research object,and provides support for realizing the balanced development of urban housing prices under the premise of high accuracy prediction.There are many influencing factors on housing prices,which are mainly divided into enterprises,consumers,housing,economy,policies and other major factors.This paper mainly uses the research and application of cluster analysis,non-parametric tests and other dynamic screening methods to influence the impact The indicators of Shenzhen's real estate prices were selected,and based on the selected indicators,the gray correlation model was used to quantitatively analyze the relevant monthly data of the city.The experimental results show that the main factors of housing prices are not only related to corporate factors,but also the property of the house itself to a large extent determine price changes.At present,most of the algorithms for housing price prediction are still implemented on the basis of traditional machine learning prediction or statistical analysis.This thesis researches on the latest deep learning algorithms and proposes an attention recurrent neural network to improve the accuracy for the prediction of housing price,which has highly practical value in certain aspects.In order to further improve the prediction accuracy,this work proposed the improved models that adds an attention mechanism to Long Short-Term Memory(LSTM)algorithm to increases the extraction of feature attribute values.Further,in order to improve the accuracy of the prediction algorithm,a Sliding Attention-BILSTM is proposed.At the same time,through experiments Perform verification to get the best parameters and improve the prediction accuracy of the algorithm.The data source of this paper is to obtain relevant information from professional home sales Websites as the data source.Based on the model of the Scrapy framework,data collection is carried out.Furthermore,the comparison of various deep learning algorithms is performed in experiment section,the final result shows our proposed algorithm has a better accuracy on multiple data sets.