基于改进Kalman的传感器数据加权处理算法

物联网(IOT)和大数据的发展,对物联网数据质量和处理速度提出了新的要求,而物联网触感器原始数据由于噪音和虚假异常值的影响,如直接应用于大数据分析会严重影响分析结果的可靠性和有效性,大量传感器的部署也导致了虚假异常值数量的成倍增长;同时,由于物联网终端数量庞大且性能有限,数据价值密度低,使用机器学习方法进行处理性价比不...

A Sensor Data Weighting Algorithm Based on Improved Kalman

The development of the Internet of Things(IOT)and big data puts forward new requirements for the data quality and processing speed of the IOT,while the original data of the IOT will seriously affect the reliability and effectiveness of the analysis results due to the impact of noise and false outliers,such as the direct application to big data analysis,and the deployment of a large number of sensors also leads to the multiple growth of the number of false outliers.At the same time,due to the large number and limited performance of IOT terminals,low data value density,low cost performance and versatility of using machine learning methods to process sensor data,how to efficiently,reliably and universally process sensor data and carry out anomaly detection has become a hot issue.Based on the statistical method,combined with Kalman filtering and weighted fusion,we propose a weighted sensor processing and prediction algorithm.Compared with the moving average,MSE and MAE have been improved and the performance has been improved obviously.The performance of the prediction model before and after data processing has been verified by lightgbm algorithm,which proves that the data processed by this algorithm is easier for model training and prediction.