Finding the Truth from Uncertain Time Series Based on Entropy
Bo Jiang, Jizhou Sun · 2022
Truth discovery on time series data aims to estimate true data at each timestamp from measurements of multiple uncertain data sources such as sensors. It has received much attention in recent years and has been applied practical Internet of Things applications. Many truth discovery methods have been proposed for scalar categorical and numerical data. However, very few methods focus on time series data, and require different prior knowledge for different datasets. In this paper, we propose a novel Entropy based Truth Discovery method (ETD). The main idea behind is that dirty data usually has high entropy in frequency domain. The main benefit of our method is that it is adaptive, and there is no need for user to specify parameters manually. Experimental results on real-world datasets demonstrate the superior performance of our approach.