No abstract

No abstract

The Sigma-Point Kalman Filters (SPKF) is a family of filters that achieve very good performance when applied to time series. Currently most researches involving time series forecasting use the Sigma-Point Kalman Filters, however they do not use an ensemble of them, which could achieve a better performance. The REC analysis is a powerful technique for visualization and comparison of regression models. The objective of this work is to advocate the use of REC curves in order to compare the SPKF and ensembles of them and select the best model to be used.

Abstract. RISE is a well-known multi-strategy learning algorithm that combines rule induction and instance-based learning. It achieves higher accuracy than some state-of-the-art learning algorithms, but for large data sets it has a very high average running time. This work presents the analysis and experimental evaluation of SUNRISE, a new multi-strategy learning algorithm based on RISE, developed to be faster than RISE with similar accuracy. The SUNRISE AlgorithmRISE (Rule Induction from a Set of Exemplars) [2] induces all the rules together. If a generalization of a rule has positive or null effect on the global accuracy, the change is kept. The RISE algorithm is presented in Table 1. SUNRISE tries to generalize the rules more than once before including them in the rule set and only accepts the changes if the effect on the global accuracy is strictly positive; i.e., a new rule is only added to the rule set if the set achieves a higher accuracy than before its inclusion. The SUNRISE algorithm is presented in Table 2. The fact that SUNRISE does not use Occam's Razor increases the algorithm's speed because it increases the probability of no modification in the rule set after an iteration of the outermost loop Repeat, thus causing the algorithm's stop. Only after k generalizations a rule is evaluated to determine if it must or not belong to the rule set. It makes SUNRISE faster than RISE, since the latter evaluates each generalization made in each rule. The value k is a parameter of the SUNRISE algorithm whose value has to be experimentally determined. Experimental EvaluationIn the experiments, 22 data sets [1] were used to compare the performance of the new algorithm, SUNRISE, to that of the RISE algorithm. The test method used in this research was the paired t test with n-fold cross-validation [5]. To adjust the parameter k, an internal cross-validation was made [5]. The value for k that achieved better performance in most of the data sets was k ≤ 3. All tests were carried through in a Pentium III 450MHz computer with 64MBytes RAM. Table 3 presents the running time (training and testing) of each algorithm for each one of the data sets. The two last columns show the results obtained by the SUNRISE

No abstract