(a) Normalized mean square error. The error increases slightly with the percentage of missing data when RIDA uses DWT. However, the error remains constant when RIDA uses DCT. (b) Compression performance of temperature data. The compression ratios degrade slowly with the percentage of missing data. (c) Classification recall. When half the network data is missing, RIDA can still have a recall of more than 95% using either DCT or DWT.