Arindam Chaudhuri

A Comparative Study of Kernels for the Multi-class Support Vector Machine

Support Vector Machine (SVM) is a powerful classification technique based on the idea of structural risk minimization. Use of a kernel function enables the curse of dimensionality to be addressed. However, a proper kernel function for a certain problem is dependent on the specific dataset and as such there is no good method on how to choose a kernel function. In this paper, the choice of the kernel function is studied empirically and optimal results are achieved for multiclass SVMs combining several binary classifiers. The performance of the Multi-class SVM is illustrated by extensive experimental results which indicate that with suitable Kernel and parameters better classification accuracy can be achieved as compared to other methods. The experimental results of the four datasets show that Gaussian Kernel is not always the best choice to achieve high generalization of classifier although it is often the default choice.

A Study of the Traveling Salesman Problem Using Fuzzy Self Organizing Map

Kohonen self organizing map is an important artificial neural network technique that uses competitive, unsupervised learning to produce a low-dimensional discretized representation of the input space of the training samples which preserves the topological properties of the input space. The fuzzy set theory introduces the concept of membership function to the learning process of Self Organizing Map which helps to handle the inherent vagueness involved in most of the real life problems. In this paper, fuzzy self organizing map with one dimensional neighborhood is used to find an optimal solution for the symmetrical Traveling Salesperson Problem. The solution generated by the Fuzzy Self Organizing Map algorithm is improved by the 2opt algorithm. Finally, the Fuzzy Self Organizing Map algorithm is compared with Lin-Kerninghan Algorithm and Evolutionary Algorithm with Enhanced Edge Recombination operator and self-adapting mutation rate.

Time Series Forecasting Using Hybrid Neuro-Fuzzy Regression Model

During the past few decades various time-series forecasting methods have been developed for financial market forecasting leading to improved decisions and investments. But accuracy remains a matter of concern in these forecasts. The quest is thus on improving the effectiveness of time-series models. Artificial neural networks (ANN) are flexible computing paradigms and universal approximations that have been applied to a wide range of forecasting problems with high degree of accuracy. However, they need large amount of historical data to yield accurate results. The real world situation experiences uncertain and quick changes, as a result of which future situations should be forecasted using small amount of data from a short span of time. Therefore, forecasting in these situations requires techniques that work efficiently with incomplete data for which Fuzzy sets are ideally suitable. In this work, a hybrid Neuro-Fuzzy model combining the advantages of ANN and Fuzzy regression is developed to forecast the exchange rate of US Dollar to Indian Rupee. The model yields more accurate results with fewer observations and incomplete data sets for both point and interval forecasts. The empirical results indicate that performance of the model is comparatively better than other models which make it an ideal candidate for forecasting and decision making.