Research On Outlier Detection Of Apparent Temperature Based On Crowd Sensing

Crowd Sensing is a new theory based on the concepts of wireless sensor networks.Through the incentive playing to the advantage of more people more power, crowd sensing converts the part of sensing needing lots of manpower and material resources to unconscious daily activities, and then people extract useful information from these large amounts of inexpensive data. Now crowd sensing is widely applied in many areas,such as detection of tra?c condition, detection of the surrounding environment and social friends, and it is the focus of the current international scienti?c research.Apparent temperature is closely related to people’s lives, and it makes people feel more realistic compared to temperature provided by meteorological observatory. So it is very important to know how to use the advantages of crowd sensing, inexpensive to collect a large number of apparent temperature data, and how to use the knowledge of data collection in wireless sensor networks, to transfer large amounts of data to the database, and how to ?lter these not very reliable data, to get a more realistic apparent air temperature data. According to the problems mentioned above, the contributions of this thesis are as follows:The collection of apparent temperature data part: On the one hand, the apparent temperature is analyzed from temperature, humidity, wind speed and solar irradiance.Because apparent temperature is affected by multiple factors, it is more di?cult to describe than temperature. On the other hand, based on crowd sensing using mobile APP we designed the collection of data. In the design of APP, it is more considered the problem of user sencing.The outlier detection of apparent air temperature part: Because the source of the data which is in the initial apparent air temperature database is mobile devices users that do not have professional training, and the data in the transmission process of network experienced many links, abnormal data can hardly be avoided exist in the database. In this paper, through the comparison of clustering data in the database, and use DTW technology to solve the similarity between two unequal length time series,?nd out the abnormal data, to improve the accuracy of the apparent air temperature database, and make the related users to get more accurate information.