In order to improve the prediction accuracy and reduce the noise influence of link quality for wireless sensor network (WSNs)

a link quality prediction algorithm based on improved kernel FCM and intelligent SVR (IKFCM-ISVR) is proposed. Firstly

the validity index based on compactness and dispersion is introduced into the kernel FCM (KFCM) method

which realizes the automatic division of cluster number for samples. Then the improved kernel FCM method is used to process the data of link quality

and the membership degree of sample clustering is obtained. On this basis

the SVR prediction model based on social spider optimization (SSO) algorithm is constructed

and the SSO based on dynamic refraction learning mechanism is used to optimize the parameters

getting the best combination of SVR parameters for different clustering. Finally the IKFCM-ISVR algorithm is used to predict the WSNs link data in different experimental scenarios. The simulation results show that

compared with other prediction algorithms

the prediction accuracy of the algorithm is improved by 36.8~68.4%.