Anomaly Detection for Online Visiting Traffic as a Real-Estate Indicator: The Case of HomeBuyer

Real-estate development involves a large amount of cash flow, yet the overall process takes a long period of time to complete, leading to a high risk of on-going demand change, competitor, and society. Since many buyers these days preview properties from online sources, the total user amount viewing each project and the market segment can imply a current trend for purchasing demand. Instead of monitoring a user logging file manually every day, we develop an auto-alarm system to detect anomaly events. In particular, we apply a seasonal auto-regressive integrated moving average (SARIMA) model to the number of user views varying in a seasonal manner from week to week. We then use Bollinger Bands, a widely used statistical indicator, to draw a boundary for any incident sporadically deviating from the expected. This system can alarm the real-estate developers whether their target customer interest is still staying on their properties or moving towards different areas or competitors so that they could abruptly adjust their strategy in time.