A time-dependent enhanced support vector machine for time series regression

Ristanoski, Goce; Liu, Wei; Bailey, James E.

doi:10.1145/2487575.2487655

Cited by 14 publications

(22 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The z-normalization which transforms the data to ensure zero average and unit standard deviation, has been strongly advocated in tasks that need to search for TS subsequences [5]. Figure 2 illustrates a subsequence S (1) that is the exact match to the query Q. The subsequence S (1) has the exact same values as Q, with just one difference, a small offset increase.…”

Section: The Knn-tspi Algorithmmentioning

confidence: 99%

“…Figure 2 illustrates a subsequence S (1) that is the exact match to the query Q. The subsequence S (1) has the exact same values as Q, with just one difference, a small offset increase. We also included a second subsequence S (2) that is completely different from Q.…”

Section: The Knn-tspi Algorithmmentioning

confidence: 99%

“…In our example, S (2) is a straight line in order to make our argument stronger, but also for reasons that will become clearer when we discuss the complexity invariance. According to ED, the most similar subsequence to Q is S (2) (ED(Q, S (2) ) = 84.26 e ED(Q, S (1) ) = 114.54), even though the offset difference between Q and S (1) is just 10 units. The reason is very simple: Small offset differences quickly accumulate making the final ED grow very fast.…”

Section: The Knn-tspi Algorithmmentioning

confidence: 99%

“…The reason is very simple: Small offset differences quickly accumulate making the final ED grow very fast. Small offset difference makes ED consider Q more similar to S (2) than to S (1) Although for some applications, the amplitude and the offset may constitute a relevant feature to characterize the subsequence, in the majority of the application domains searching without invariance to offset and amplitude does not lead to meaningful results. That happens because the similar subsequences hardly occur in the exact same offset.…”

Section: The Knn-tspi Algorithmmentioning

confidence: 99%

“…In the last two decades, with the rise of the Data Mining (DM) process, there is an increasing interest in the adaptation of Machine Learning (ML) methods, especially those for regression tasks, to support Time Series (TS) non-parametric modeling and prediction [1]. Due to their simplicity and comprehensibility, the ML methods have established themselves as serious candidates to the classical parametric models, which are based on autoregression and moving averages [2].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A Study of the Use of Complexity Measures in the Similarity Search Process Adopted by kNN Algorithm for Time Series Prediction

Parmezan

Batista

2015

2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA)

View full text Add to dashboard Cite

Abstract-In the last two decades, with the rise of the Data Mining process, there is an increasing interest in the adaptation of Machine Learning methods to support Time Series non-parametric modeling and prediction. The non-parametric temporal data modeling can be performed according to local and global approaches. The most of the local prediction data strategies are based on the k-Nearest Neighbor (kNN) learning method. In this paper we propose a modification of the kNN algorithm for Time Series prediction. Our proposal differs from the literature by incorporating three techniques for obtaining amplitude and offset invariance, complexity invariance, and treatment of trivial matches. We evaluate the proposed method with six complexity measures, in order to verify the impact of these measures in the projection of the future values. Besides, we face our method with two Machine Learning regression algorithms. The experimental comparisons were performed using 55 data sets, which are available at the ICMC-USP Time Series Prediction Repository. Our results indicate that the developed method is competitive and the use of a complexity-invariant distance measure generally improves the predictive performance.

show abstract

Section: The Knn-tspi Algorithmmentioning

confidence: 99%