A working likelihood approach to support vector regression with a data-driven insensitivity parameter

Abstract

The insensitivity parameter in support vector regression determines the set of support vectors that greatly impacts the prediction. A data-driven approach is proposed to determine an approximate value for this insensitivity parameter by minimizing a generalized loss function originating from the likelihood principle. This data-driven support vector regression also statistically standardizes samples using the scale of noises different from conventional response scaling method. Statistical standardization together with probabilistic regularization based on a working likelihood function produces data-dependent values for the hyperparameters including the insensitivity parameter. The exact asymptotical solutions are provided when the noises are normally distributed. Nonlinear and linear numerical simulations with three types of noises ($$\epsilon$$ ϵ -Laplacian distribution, normal distribution, and uniform distribution), and in addition, five real benchmark data sets, are used to test the capacity of the proposed method. Based on all the simulations and the five case studies, the proposed support vector regression using a working likelihood, data-driven insensitivity parameter is superior and has lower computational costs.

An adaptive trimming approach to Bayesian additive regression trees

“…To demonstrate the significance of our proposed BART(Atr) method in forecasting, we conducted a Wilcoxon signedrank test using MAE and MAPE indexes. These indexes were obtained from 100 repeated experiments on the test set, specifically under rand-attack 1, as mentioned in Wu and Wang [33]. The results of the statistical tests are documented in Table 8.…”

“…Furthermore, we specifically concentrated on the weight function ψ(•) mentioned in (8), following the recommendation in VandenHeuvel et al [18]. We acknowledge that there exist various alternative weight functions, as extensively discussed in papers such as Wang et al [21,32], Wu and Wang [33], and Pratola et al [35]. Among these alternatives, the two most commonly utilized functions are Huber's weight function and the bisquare weight function shown in Wang et al [21], Jiao et al [36].…”

Modeling Energy Gap of Doped Tin (II) Sulfide Metal Semiconductor Nanocatalyst Using Genetic Algorithm‐Based Support Vector Regression

“…The number of support vectors in insensitive zones is determined by the epsilon parameter, while the kernel function parameter controls the transfer of the input data to a feature space of higher dimension. [93,94].…”