CAJ | 학술논문

物联网中，无线传感器网络由于环境、资源等因素的变化和限制，会导致部分数据异常或丢失，使数据传输的可靠性降低。因此常用的 BP 神经网络方法在根据最终获取数据进一步处理时的准确性不高。提出 K 近邻算法和 BP 神经网络相结合的二阶段预测模型，先使用 K 近邻算法对 BP 神经网络输入数据中异常或缺失数据进行估值和替换预处理，同时进行初步预测，然后将预处理后的数据输入 BP 神经网络，综合 BP 神经网络和 KNN 的预测结果给出最终结论。实际环境中实验表明，所提出的模型能够有效地提高物联网环境中预测的准确度和稳定性。
물련망중，무선전감기망락유우배경、자원등인소적변화화한제，회도치부분수거이상혹주실，사수거전수적가고성강저。인차상용적 BP 신경망락방법재근거최종획취수거진일보처리시적준학성불고。제출 K 근린산법화 BP 신경망락상결합적이계단예측모형，선사용 K 근린산법대 BP 신경망락수입수거중이상혹결실수거진행고치화체환예처리，동시진행초보예측，연후장예처리후적수거수입 BP 신경망락，종합 BP 신경망락화 KNN 적예측결과급출최종결론。실제배경중실험표명，소제출적모형능구유효지제고물련망배경중예측적준학도화은정성。
In Internet of Things (IoT),wireless sensor network (WSN)may have part of data abnormal or lost caused by the change and limit in environment,resources and other factors,thus the reliability of data transmission is reduced.So the BP neural network method commonly used is not high in accuracy when further processing according to the final acquisition data.We propose a two-step prediction model which combines k-nearest neighbour (KNN)algorithm and BP neural network.First,it uses KNN algorithm on BP neural network input data to estimate and replace the abnormal and missing data in it as preprocessing,and gives a preliminary prediction at the same time.Then it uses the preprocessed data to input in BP neural network,and provides the final conclusion by integrating the prediction results of BP neural network and KNN.The experiment in actual environment shows that the proposed model can effectively improve the readiness and stability of prediction in IoT environment.