Browsing by Subject "Support Vector Machines"

Now showing 1 - 2 of 2

Open Access
Classification of agricultural kernels using impact acoustic signal processing
(2006) Onaran, İbrahim
The quality is the main factor that directly affects the price for many agricultural produces. The quality depends on different properties of the produce. Most important property is associated with health of consumers. Other properties mostly depend on the type of concerned vegetable. For instance, emptiness is important for hazelnuts while openness is crucial for the pistachio nuts. Therefore, the agricultural produces should be separated according to their quality to maintain the consumers health and increase the price of the produce in international trades. Current approaches are mostly based on invasive chemical analysis of some selected food items or sorting food items according to their color. Although chemical analysis gives the most accurate results, it is impossible to analyze large quantities of food items. The impact sound signal processing can be used to classify these produces according to their quality. These methods are inexpensive, noninvasive and most of all they can be applied in real-time to process large amount of food. Several signal processing methods for extracting impact sound features are proposed to classify the produces according to their quality. These methods are including time and frequency domain methods. Several time and frequency domain methods including Weibull parameters, maximum points and variances in time windows, DFT (Discrete Fourier Transform) coefficients around the maximum spectral points etc. are used to extract the features from the impact sound. In this study, we used hazelnut and wheat kernel impact sounds. The success rate over 90% is achieved for all types produces.
Open Access
Risk-averse multi-class support vector machines
(2018-12) Karagöz, Ayşenur
A classification problem aims to identify the class of new observations based on the previous observations whose classes are known. It has many applications in a variety of disciplines such as medicine, finance and artificial intelligence. However, presence of outliers and noise in previous observations may have significant impact on the classification performance. Support vector machine (SVM) is a classifier introduced to solve binary classification problems under the presence of noise and outliers. In the literature, risk-averse SVM is shown to be more stable to noise and outliers compared to the original SVM formulations. However, we often observe more than two classes in real-life datasets. In this study, we aim to develop riskaverse multi-class SVMs following the idea of risk-averse binary SVM. We use risk measures, VaR and CVaR, to implement risk-aversion to multi-class SVMs. Since VaR constraints are nonconvex in general, SVMs with VaR constraints are more complex than SVMs with CVaR. Therefore, we propose a strong big-M formulation to solve multi-class SVM problems with VaR constraints efficiently. We also provide a computational study on the classification performance of the original multi-class SVM formulations and the proposed risk-averse formulations using artificial and real-life datasets. The results show that multi-class SVMs with VaR are more stable to outliers and noise compared to multi-class SVMs with CVaR, and both of them are more stable than the original formulations.