Virgilijus Sakalauskas

Evaluation framework of hierarchical clustering methods for binary data

The article aims to evaluate hierarchical clustering methods according to their performance for binary data type. We explore the accuracy of ten hierarchical clustering methods by experimenting with ten different distance measures. The three types of well, poorly and very poorly separated clusters of binary data sets are generated by selecting the appropriate parameters for binomial distribution and Monte Carlo method. In order to evaluate the precision of clustering methods the binary data sets are transformed to distance matrices. The error level each method is explored in relationship to distance measures, cluster types and data distributions. The Complete linkage, Flexible-beta and Wards methods have best clustering performance for the case of two well separated clusters of binary data.

Intelligent Systems for Retail Banking Optimization - Optimization and Management of ATM Network System.

The article analyzes the problems of optimization and management of ATM (Automated Teller Machine) network system, related to minimization of operating expenses, such as cash replenishment, costs of funds, logistics and back office processes. The suggested solution is based on merging up two different artificial intelligence methodologies – neural networks and multi-agent technologies. The practical implementation of this approach enabled to achieve better effectiveness of the researched ATMs network. During the first stage, the system performs analysis, based on the artificial neural networks (ANN). The second stage is aimed to produce the alternatives for the ATM cash management decisions. The performed simulation and experimental tests of method in the distributed ATM networks reveal good forecasting capacities of ANN.

Analysis of the day-of-the-week anomaly for the case of emerging stock market

The aim of the article is to explore the day-of-the-week effect in emerging stock markets. This effect relates to the attempts to find statistically significant dependences of stock trading anomalies, which occur in particular days of the week (usually the first or the last trading day), and which could be important for creating profitable investment strategies. The main question of the research is to define, if this anomalies affects the entire market, or it is applicable only for the specific groups of stocks, which could be recognized by identifying particular features. The investigation of the day-of-the-week effect is performed by applying two methods: traditional statistical analysis and artificial neural networks. The experimental analysis is based on financial data of the Vilnius Stock Exchange, as of the case of emerging stock market with relatively low turnover and small number of players. Application of numerous tests and methods reveals better effectiveness of the artificial neural networks for indicating significance of day-of-the-week effect.

Short-Term investment risk measurement using var and CVaR

The article studies the short-term investment risk in currency market. We present the econometric model for measuring the market risk using Value at Risk (VaR) and conditional VaR (CVaR). Our main goals are to examine the risk of hourly time intervals and propose to use seasonal decomposition for calculation of the corresponding VaR and CVaR values. The suggested method is tested using empirical data with long position EUR/USD exchange hourly rate.

Dynamic Simulation of Pension Funds’ Portfolio

The article investigates the dynamics of pension fund portfolio by using adaptive Powersim simulation models. Many countries use systems of investment to pension funds for ensuring older peoples’ financial stability and encourage their participation by partial tax relief. There is a need to create advanced tools that could help the unprofessional investors, who make up the majority of pension fund customers, to make well informed decisions and reduce the potential risk. Financial companies use a variety of spreadsheets based on parameters describing investors and markets, which remain constant during all calculation process. Due to high volatility of the financial markets, personal investment power and change in tax relief system the actual portfolio dynamics highly deviates from the initial forecast in the initial stage of the investment. Description of the tool, which was created with the Powersim software package, has been presented in the article. The model, proposed by the authors, introduces adaptive and dynamic variables used in the portfolio simulation. Volatile values of return rate, contributions and fees are used in calculations. The results of research show the advantages of such method over traditional spreadsheets.

Entropy-Based Indicator for Predicting Stock Price Trend Reversal

Predicting changes of stock price long term trend is an important problem for validating strategies of investment to the financial instruments. In this article we applied the approach of analysis of information efficiency and long term correlation memory in order to distinguish short term changes in trend, which can be evaluated as informational ‘nervousness’, from the reversal point of long term trend of the financial time series. By integrating two econometrical measures of information efficiency - Shannon’s entropy (SH) and local Hurst exponent (HE) - we designed aggregated entropy-based (EB) indicator and explored its ability to forecast the turning point of trend of the financial time series and to calibrate the stock market trading strategy.

The Impact of Taxes on Intra-week Stock Return Seasonality

In this paper we explore the impact of trading taxes (commissions) on day-of-the-week effect in the Lithuanian Stock market. We applied the computational model for processing trading activities only on the particular days of the week. The suggested algorithm of trading shares not only reveals presence of the day-of-the-week anomaly, but allows comparing it to the influence of the trading taxes by estimating the final return of the selected shares. As the taxes of each transaction depend on the investment sum, therefore the suggested algorithm had to optimize the number of operations for ensuring the biggest gain. The research revealed significance of intra-week stock return seasonality for majority of shares (17 out of total 24). The advantages of the suggested method include its ability to better specify the shares for performing intra-week seasonality-based transactions, even though embracing of the trading commissions reduces visibility of the effect.

Research of Conventional Data Mining Tools for Big Data Handling in Finance Institutions

The article investigates the usability of conventional data mining tools for handling data sets generated in financial institutions. It discloses the characteristics of “big data” which reveal limitations and new requirements for analytical software to deal with huge data flows related to financial transactions. The performance characteristics of four different conventional data mining tools, their visualization and clustering models are tested for experimental set of big data extracted from bank local data warehouse. The ranking of critical characteristics is provided for each stage of analysis of big data set.

Evolution of Information Efficiency in Emerging Markets

In the article we investigate the informational efficiency of Nasdaq OMX Baltic stock exchange by applying Shannon’s entropy measure for symbolized time series. The complexity of the problem of market efficiency evaluation has lead to application of various soft computing methods and to even contradictory outcomes confirming or denying the efficient market hypothesis. The goal of the article is to explore the possibilities of quantitative evaluation of market effectiveness, by presenting the computational method and its experimental research for the financial data of the emerging Baltic market. The computations were performed for different time spans and symbolic word lengths. The research results allowed to conclude that the efficiency of Baltic market strongly falls behind the developed countries, and it raises expectations for modelling profitable trading strategies. Application of the entropy measure allows to explore the evolution of the market efficiency and to apply the algorithm for predicting the forthcoming crises of financial markets.

Neural Networks Approach to the Detection of Weekly Seasonality in Stock Trading

In this article we investigate the problem of detection the statistically significant dependences of stock trading return, which occur in particular days of the week (usually the first or the last trading day), and which could be important for creating profitable investment strategies. The identifying such days of the week (day-of-the-week effect) is performed by using artificial neural networks. The research results helped to conclude the effectiveness of application of neural networks, as compared to the traditional linear statistical methods for finding stock trading anomalies. The effectiveness of the method was confirmed by exploring impact of different variables to the day-of-the-week effect, evaluation of their influence and sensitivity analysis, and by selecting best performing neural network type. The experimental verification was implemented by using Vilnius Stock Exchange trading data.