Browsing by Author "Roshan, WDS"

Now showing 1 - 3 of 3

item: Conference-Abstract
Financial forecasting based on artificial neural networks : promising directions for modeling
(2016-09-23) Roshan, WDS; Gopura, RARC; Jayasekara, AGBP
Financial forecasting plays a critical role in present economic context where neural networks have become a good alternative technique over traditional methods. Vast ranges of neural models are developed to achieve better accuracy in forecasting. In addition, the ways to find out a good neural architecture is being explored by the research community. In the literature, main problems are figured out within the area of data preparing and neural network design. In this paper, the reasons that affect the performance of the models are discussed based on empirical and mathematical evidence. Finally, this paper presents the directions towards a more suitable neural model for financial forecasting by combining data preprocessing techniques, clustering techniques and support vector machine.
item: Conference-Abstract
Financial market forecasting by integrating wavelet transform and K-means clustering with support vector machine
(2016-09-23) Roshan, WDS; Gopura, RARC; Jayasekara, AGBP; Bandara, DSV;
Financial market forecasting is a challenging problem and researchers are still exploring the ways to improve the performance of the existing models. This paper presents a forecasting model by integrating wavelet transform, K-means clustering with support vector machine. At the first stage, noise of the input prices is removed by using wavelet denoising. Wavelet multi resolution analysis is used to decompose the original time series in to multiple details and approximated decompositions. Individual support vector models are trained for each detail part. Approximated part is further analyzed by clustering and training support vector models for each cluster. Finally the forecast is made for the wavelet denoised time series by summing up the forecasts of each support vector model. Results have shown that the proposed model has given the accurate forecast and has the capability to support decisions in real world trading.
item: Thesis-Full-text
Hybrid approach for financial forecasting with support vector machines
(2015-08-28) Roshan, WDS; Gopura, RARC; Jayasekara, AGBP
Financial markets are the biggest business platforms in the world. Therefore, financial forecasting is getting a lot of attention in today’s economic context. Accurate forecast is beneficial to broker firms, governments, individuals etc. Vast range of forecasting methods, models have introduced by the research community. However, the risk involved with trading on those markets are very high. Such complexity makes a difficulty of making consistent profit. Building an accurate forecasting model is still an active and interesting research area for the academic community. Recently, nonlinear statistical models such as neural network, support vector machine have shown greater capability to forecast financial markets over conventional methods. This dissertation pro-posed a hybrid support vector machine model which consists of wavelet transform and k-means clustering for foreign exchange market forecasting. The proposed model analyzes the trends and makes a forecast by entirely depending on the past exchange data. Wavelet transform is used to remove the noise of the time series. K-means clustering cluster the input space according to the similarities of the input vectors and finally support vector models make a forecast for the relevant cluster. The proposed hybrid forecasting system was tested on real market environment to check the fore-casting capability. Auto trading algorithm developed on ‘metatrader4’ platform used the forecast of the model to trade on the real conditions. Results confirmed that the proposed model can fore-cast price movements with greater accuracy that leads to profitable trades on foreign exchange market