Goran Andjelic

Fuzzy prediction based on regression models: Evidence from the Belgrade Stock Exchange - prime market and graphic industry

In the conventional regression model, deviations between the observed values and the estimated values are supposed to be due to measurement errors. Here, taking a different perspective, these deviations are regarded as the fuzziness of the systems parameters. Thus, these deviations are reflected in a linear function with fuzzy parameters. Using linear programming algorithm, this fuzzy linear regression model might be very convenient and useful for finding a fuzzy structure in an evaluation system. In this paper, the details of the fuzzy linear regression concept and its applications in an uncertain environment are shown and discussed on data of the Belgrade Stock Exchange. In addition, the prediction of stock market prices is performed using fuzzy linear trend. Having in mind the characteristics of trading methods, fuzzy linear trend is used for prediction of stock market prices based on historical data, which are not precisely given within a trading day. Results of the research indicate the significance of fuzzy prediction based on regression models, i.e. fuzzy linear trend.

Fuzzy uncertainty analysis in the investment processes

The analyses of the investment processes, especially the possibility of monitoring and forecasting the intense return fluctuations, demands usage of an appropriate framework, which adequately evaluates the dynamic nature of those processes. In the recessive business conditions, volatility and extreme stock market movements presents a challenge while evaluation of the investment processes. Thus, the stock market prices volatility and the possibility of stock market forecasting in an uncertain environment are in the focus of the research. Fuzzy sets allow usage of approximated values of the ambiguous data, that is, the intraday stock market prices data. The sample comprises stocks of Alfa plam a.d. Vranje (ALFA), Komercijalna banka a.d. Beograd (KMBN) and Metalac a.d. Gornji Milanovac (MTLC), that are traded at the regulated market of the Belgrade Stock Exchange. The main goal of the research is to present an overview and performance test of the fuzzy sets as a part of soft computing (SC) technology in the investment processes. The research results indicate the necessity of the fuzzy sets application in the forecasting of stock market prices, i.e. its approximate values.

The performance of the investment return prediction models: Theory and evidence

The market structure has been adjusted in order to be as simple as possible in sense of its economic components. The aim of the investment return prediction is constructing as good models of the market movement as possible. As for as the stock market is concerned, the price rise of some stocks indicate the good results of the management of that company, while the price fall shows the inadequate management. Prompt and accurate information of the market movement enable the managers to take some measures which lead to optimal investment decision. The Autoregressive Moving Average (ARIMA) model is one of the most frequently linear models of the time series used for the investment return prediction. The prediction researches in the last years from the areas of Artificial Neural Networks (ANNs) indicate that ANNs with a combination of other prediction models give better prediction results. This research aim is to introduce a hybrid model ARIMA fuzzy-neural network for the prediction of the stock market index BELEX15 values. The research results indicate that the linear model ARIMA and fuzzy ANNs exhibit more superior investment return prediction performances.

Application of neural networks in investments: A case of BELEX15 stock index

The time series represent a line of observations of unintentional variable in different time periods. The aim of the time series is, on the basis of those observations, to forecast the values of the variable. Further reactions should be predicted on the basis of the data from the past. The research covers the sample representing the Serbian (BELEX15) stock index. The aim of this study is the survey of the algorithm for predicting the reaction of the problem of this type by using the neural networks in function of prediction of the daily stock index values in the emerging market of the Republic of Serbia. The proposal of the algorithm is implemented by software and the results are significant both to academics and professionals in the subject area, especially having in mind the possibility of prediction in investments.

Fuzzy sets and decision analysis in emerging markets

This paper investigates the performance of fuzzy sets theory in investment processes with the daily stock index returns of three different emerging markets. The research covers the sample representing the Serbian (BELEXline), Croatian (CROBEX) and Slovenian (LJSEX) stock indexes. The main goal is to determine whether the application of fuzzy numbers in investment processes adequately estimates the daily stock index returns in the emerging markets of Serbia, Croatia and Slovenia, especially regarding decision analysis. Research results indicate that the fuzzy sets theory is significant in the framework of decision analysis in investment processes in selected emerging markets.

Hybrid system prediction for the stock market: The case of transitional markets

The subject of this paper is the creation and testing of an enhanced fuzzy neural network backpropagation model for the prediction of stock market indexes, including the comparison with the traditional neural network backpropagation model. The objective of the research is to gather information concerning the possibilities of using the enhanced fuzzy neural network backpropagation model for the prediction of stock market indexes focusing on transitional markets. The methodology used involves the integration of fuzzified weights into the neural network. The research results will be beneficial both for the broader investment community and the academia, in terms of the application of the enhanced model in the investment decisionmaking, as well as in improving the knowledge in this subject matter.

The comparative analyses of the nonparametric methods for investment return prediction

The financial market is complex, evolving and dynamic system, which has an extremely non-linear movement. Thus, investment return prediction represents a significant challenge, especially because of its great diversity, unsteadiness and unstructured data with a high degree of instability and pronounced hidden connections. It is known that accurate prediction of the stock market indexes is very important for the development of effective trading strategies in investments. The main objective of the research is to perform the comparative analyses of different nonparametric methods, that is, fuzzy artificial neural networks (fuzzyANN) and genetic algorithm artificial neural networks (GAANN) for predicting the movements of the stock market indexes. The survey is conducted on the BELEX15, SBITOP, BUX and CROBEX stock market indexes. Model estimates were carried out through the prediction error MAE, MAPE and RMSE. The research results point to the adequacy of the nonparametric methods application in investments.

A comparative study of the extreme value theory model in investments

The subject of the research is the implementation of the comparative study of the Extreme Value Theory model (EVT) in investments. In that sense, the scope of the research is to analyze performances of the Extreme Value Theory model (EVT), Delta normal VaR (D VaR) and Historical Simulation (HS VaR) models on the transitional market of the Republic of Serbia with confidence level of 95% for 100 and 300 days. The aim of the research is to provide quality and reliable information about the implementation possibilities of different VaR calculation models in investments optimization on the domestic market. The methodology of the research comprises the MANOVA analyses, discriminant analysis and Roys test. The main results of the research show that there is a statistically significant difference in the performance of the tested models; that is, in the accuracy of prediction of the risk from investment activities. The results of the research will be useful to both academic and practical communities regarding significant growth of the cognitive base of implementation possibilities of different VaR calculation models in the investment activities on the transitional market of the Republic of Serbia.

Application of fuzzy sets in emerging markets: An empirical treatment

The possibility of application of fuzzy sets in global recessive business conditions is significant in investment processes on emerging markets. Thus, in this paper, we investigate the application of fuzzy numbers as functions of handling the imprecise conditions, volatility peculiarities and attributes of emerging markets. The research covers the sample representing stock indexes on emerging markets in selected Central and Eastern European countries.