Missing data handling by a multi-step ahead predictive filter

A multi-step ahead predictive filter for missing data handling is presented in this paper. This is a simple FIR filter, useful for time and memory critical applications. Furthermore, our proposed algorithm is formulated in such a way that only one set of FIR weights are tuned (by steepest gradient descent method), memorised and used by the system for different numbers of steps ahead in prediction. Hence, this filter is also suitable for memory critical applications. Our proposed filter, the first order hold method and a non-linear neural predictive filter were applied to missing data handling by prediction of up to four missing samples of sensor data in the HMI section of a real time and memory critical application. Experimental results show that prediction error of our proposed filter is decreased by 80% with respect to the first order hold method and it is comparable to the neural filter. Moreover, execution time of data missing handling by our proposed method is 30 times shorter than execution time of the neural filter. Similarly, the number of weights required to be memorised in our method, is 30 times lower than the number of weights in the neural filter. Hence, a high prediction performance is achieved with low computational overhead, which makes our method suitable for real time predictive filtering in time and memory critical applications.