Location-aware time-series clustering on optimal sensor arrangement

Many researches focus on clustering location or time series data in the sensor environment. In time series data, similar metric is often used to measure similar data, and many works use different algorithms to calculate similarity between two subsequences. Also, in location clustering, the well-known algorithm k-NN also presents excellent results, even though it is a classification approach. Many works focus on how to increase the efficiency and accuracy of these algorithms. However, in some cases, we need to consider both similarities between locations and time series together. Like greenhouse or wildfire detection, key points (sensors) in these cases occupy an important status. In this paper, we identify a KTL problem (Key point based on Time series and Location), which need to consider time series and location of sensors together. Goal of KTL is to find key point in the region and reduce cost of sensors in greenhouse building or wild fire detection. Given a sensor distribution, we find key points of the region and remove other useless sensors. That is, we reduce the cost of sensor distribution in green house building. This paper addresses how to solve KTL problem. We propose a new algorithm, namely LTS, to consider location and time series together. We further show the algorithm we proposed can solve the KTL in real cases well.