On the other hand, using a fixed length with Markov-chain algorithm resulted in different accuracy values.
P-DPA_M1 and P-DPA_MN achieved a Recall value of 66.94 % with client-M5, and Markov-chain using a fixed length of 7 achieved 32.61 %.
Even though client-S2 has an extremely regular behavior, Markov-chain algorithm achieved only a Recall of 12.43 % using a fixed length of 3, while both P-DPA_M1 and P-DPA_MN achieved a recall value of 100%%.
Using a fixed length of 7 helped the Markov-chain algorithm predict the number of destinations with an accuracy of 43.32 % with client-M4 and 81.04 % with client-M5.
On the other hand, the Markov-chain algorithm with a fixed length of 3 was able to achieve a length prediction accuracy of 63.16 % with SUB-2 and 61.76 % with SUB-9, while using a fixed length of 7 resulted in an accuracy of 60.56 % with SUB-8 and 66.08 % with SUB-12.
With client-F8, P-DPA_MN achieved a Precision of 50.03 %, P-DPA_M1 achieved 48.92 % and Markov-chain achieved 44.20 % using a fixed length of 3. The extreme regularity of client-S2 helped both algorithms, P-DPA and Markov-chain, to achieve a Precision value of 100%%.
Epicenters of past earthquakes are supposed to contribute to earthquake density estimates, after those epicenters have been smoothed using a fixed length scale; this scale is optimized so that it minimizes the average area skill score misfit function in a retrospective experiment (Zechar and Jordan, 2010b).
The P-DPA algorithm was able to achieve the highest average F-measure of 60.77 %, while with Markov-chain algorithm, the best average F-measure of 21.71 % was achievable when using a fixed length of 7. Figure 18 shows the average F-measure with all subjects in the GPS dataset of [6].
CNV-seq [ 8] uses a fixed length sliding window and normalization of the analyzed (test) genome using a control genome.
