In this paper we apply a neural approach to develop classification models in order to assess the ... more In this paper we apply a neural approach to develop classification models in order to assess the performance of non-banking financial institutions (NFIs) in Romania. Our objective is twofold: to empirically validate our methodology and understand how different financial factors can and do contribute to the NFIs' movements from one performance class to another.
Abstract: Nowadays, the Internet comprises of huge amount of electronic information concerning di... more Abstract: Nowadays, the Internet comprises of huge amount of electronic information concerning different companies ’ financial performance. This amount greatly exceeds our capacity to analyze it, the problem being that we often lack tools to quickly and accurately process these data. DM techniques are interesting mechanisms that can be applied to rapidly changing industries, in order to get an overview of the situation. One such market is the international telecommunications industry. In this paper we construct a framework using DM techniques that enables us to make class predictions about telecommunication companies’ financial performance. Our methodology allows us to analyze the movements of the largest telecommunications companies, to see how companies perform financially compared to their competitors, what they are good at, who are the major competitors in this industry, etc. The dataset contains 88 companies from five different regions: Asia, Canada, Continental Europe, Norther...
We construct a benchmarking model in the form of a two- dimensional self-organising map (SOM) to ... more We construct a benchmarking model in the form of a two- dimensional self-organising map (SOM) to compare the performance of non- banking financial institutions (NFIs) in Romania. The NFIs are characterized by a number of performance dimensions such as capital adequacy, assets' quality and profitability. First, we apply Kohonen' SOM algorithm (an unsupervised neural network algorithm) to group the NFIs in clusters with similar characteristics. Then, we use the U-matrix method to build maps that facilitate the visualization of SOM results and select the best map in terms of quantisation error and ease of readability. The best map is used to analyze the companies over time by studying the cluster where each company was positioned for each period. We conclude that there are benefits in using SOM for interpreting large and complex financial data by identifying and visualizing clusters.
In this article we propose a modified version of Fuzzy CMeans (FCM) clustering algorithm in order... more In this article we propose a modified version of Fuzzy CMeans (FCM) clustering algorithm in order to better allocate the uncertain observations in the clusters. We change the objective function of the classic FCM by attaching different weights to the distances between observations and the clusters’ centers. We apply the modified FCM (Weighting FCM) to model the performance of non-banking financial institutions (NFIs) in Romania. We extend the experiment from our previous work by improving NFIs’ performance dataset from 3 to 8 performance ratios and from 44 to 769 observations. The results show a significant improvement in pattern allocation with the new proposed algorithm.
In this paper we analyze comparatively the macroeconomic performance of different Central and Eas... more In this paper we analyze comparatively the macroeconomic performance of different Central and Eastern European countries by the means of Data Mining (DM) techniques. We analyze the economic situations of three EU countries (Poland, Slovenia and Latvia), a newlyaccepted one (Romania), and other two non-EU countries (Russia and Ukraine). We have depicted economic performance of countries using a number of macroeconomic variables for the time period from 1993 till 2000. The economic variables that we used were: Currency Value (CV) (the inverse of the Exchange Rate - ER), Domestic Prime Rate (Refinancing Rate – RR), Industrial Output (IO) compared to previous periods in percentages, Unemployment Rate (UR), Foreign Trade (FT) in millions of USA dollars. The dataset consists of monthly/annual data during the period 1993-2000, in total 225 cases with five variables each. As DM techniques we firstly applied Self-Organizing Map (SOM) to group the countries according to their economic perform...
We use C-Means clustering algorithm to assess comparatively the performance of non-banking financ... more We use C-Means clustering algorithm to assess comparatively the performance of non-banking financial institutions (NFIs) in Romania. Firstly, we consider this real world application as a knowledge discovery problem (Data Mining) and we engage in a thorough literature review regarding the application of Data Mining methods in assessing comparatively companies' financial performance. Then, we apply C-Means on our NFIs' performance dataset. Next, we compare our results with those obtained when we apply a neural network-based clustering algorithm, called Self-organising Map (SOM) algorithm. The results show that C-Means can offer a good alternative to more sophisticated algorithms in terms of accuracy of the obtained clusters. However, the visualization capability of SOM should be taken into account when we are interested in how companies evolve over time.
In this paper we apply a new methodology presented in our previous work to classify telecom compa... more In this paper we apply a new methodology presented in our previous work to classify telecom companies in respect to their financial performance. We have two goals: to validate our methodology and, using it, to gain insights in a turbulent period in the telecommunications sector. We have obtained higher accuracy rates for the classification models than in our previous studies, and smaller differences between training and test dataset accuracy rates. The two classification techniques have performed similarly in terms of accuracy rates (decision tree, slightly better) and class predictions (multinomial logistic regression, slightly more optimistic). We have analyzed the movements of Scandinavian telecommunications companies. The results are similar to previous findings and show a strong connectivity with what had really happened to Scandinavian telecommunication companies during the second part of the last decade. Abstract . Within classical econometric modelling, because of the complexity of "data - generating process" we have chosen it to be presented with the help of a set of assumptions, where the random is as controlled as possible, so that the activity should be monitored and even dosed. This paper has chosen a systemic and cybernetic approach of the display of the "data - generating process". The first stage of the "data - generating process" decoding, the authors study a system made of three random variables defined on different probability space between which there is a special a priori dependence, called β dependence. The research includes setting a probabilistic model of such a system and making a representative example. The research ends with an analysis of the systemic and cybernetic repercussions corresponding to the β dependence.
Nowadays, the Internet comprises of huge amount of electronic information concerning different co... more Nowadays, the Internet comprises of huge amount of electronic information concerning different companies' financial performance. This amount greatly exceeds our capacity to analyze it, the problem being that we often lack tools to quickly and accurately process these data. DM techniques are interesting mechanisms that can be applied to rapidly changing industries, in order to get an overview of the
Financial decision makers are challenged by the access to massive amounts of both numeric and tex... more Financial decision makers are challenged by the access to massive amounts of both numeric and textual financial information made achievable by the Internet. They are in need of a tool that makes possible rapid and accurate analysis of both quantitative and qualitative information, in order to extract knowledge for decision making. In this paper we propose a conceptual model of a knowledge-building system for decision support based on a society of software agents, and data and text mining methods.
Proceedings of the International Conference on Applied Statistics
In this paper we evaluate comparatively the performance of non-banking financial institutions in ... more In this paper we evaluate comparatively the performance of non-banking financial institutions in Romania by the means of unsupervised neural networks in terms of Kohonen’ Self-Organizing Maps algorithm. We create a benchmarking model in the form of a two-dimensional map (a self-organizing map) that can be used to assess visually the performance of non-banking financial institutions based on different performance dimensions, such as capital adequacy, assets’ quality and profitability. We use the following indicators: Equity ratio (Leverage) for the capital adequacy dimension, Loans granted to clients (net value) / total assets (net value) for the assets’ quality dimension and Return on assets (ROA) for the profitability dimension. We have excluded from our analysis the other three dimensions used in evaluating the performance of banks, due to lack of data (for the two qualitative dimensions: quality of ownership and management) and irrelevance with the NFIs’ sector (liquidity). The p...
Proceedings of the International Conference on Applied Statistics
This paper assesses the financial performance of Romania’s non-banking financial institutions (NF... more This paper assesses the financial performance of Romania’s non-banking financial institutions (NFIs) using a neural network training algorithm proposed by Kohonen, namely the Self-Organizing Maps algorithm. The algorithm takes the financial dataset and positiones each observation into a self-organizing map (a two-dimensional map) which can be latter used to visualize the trajectories of an individual NFI and explain it based on different performance dimensions, such as capital adequacy, assets’ quality and profitability. Further, we use the map as an early-warning system that would accurately forecast the NFIs future performance (whether they would stay or be eliminated from the NFI’s Special Register three quarters into the future). The results are promising: the model is able to correctly predict NFIs’ performance movements. Finally, we compared the results of our SOM-based model with those obtained by applying a multivariate logit-based model. The SOM model performed worse in dis...
Technological and Economic Development of Economy, 2017
In this paper we propose a two-stage methodology to classify the non-banking financial institutio... more In this paper we propose a two-stage methodology to classify the non-banking financial institutions (NFIs) based on their financial performance. The first stage of the methodology consists of grouping the companies in similar financial performance classes (e.g.: “good”, “average”, “poor” performance classes). We optimise the allocation of the observations within the performance clusters by applying an enhanced version of an observation re-allocation procedure proposed in our previous work. Next, based on the result of the grouping phase, we construct a performance class variable by attaching a performance label to each data row. Then, in the second phase of our methodology, we propose a feed-forward neural-network classification model that maps the input space to the newly-constructed performance class variable. This model allows us to forecast the performance of new companies as data become available.
... Tomas Eklund Turku Centre for Computer Science and IAMSR / Åbo Akademi University Lemminkäise... more ... Tomas Eklund Turku Centre for Computer Science and IAMSR / Åbo Akademi University Lemminkäisenkatu 14 B, FIN-20520 Turku, Finland Jonas Karlsson IAMSR / Åbo Akademi University Lemminkäisenkatu 14 B, FIN-20520 Turku, Finland ...
In this paper we apply a neural approach to develop classification models in order to assess the ... more In this paper we apply a neural approach to develop classification models in order to assess the performance of non-banking financial institutions (NFIs) in Romania. Our objective is twofold: to empirically validate our methodology and understand how different financial factors can and do contribute to the NFIs' movements from one performance class to another.
Abstract: Nowadays, the Internet comprises of huge amount of electronic information concerning di... more Abstract: Nowadays, the Internet comprises of huge amount of electronic information concerning different companies ’ financial performance. This amount greatly exceeds our capacity to analyze it, the problem being that we often lack tools to quickly and accurately process these data. DM techniques are interesting mechanisms that can be applied to rapidly changing industries, in order to get an overview of the situation. One such market is the international telecommunications industry. In this paper we construct a framework using DM techniques that enables us to make class predictions about telecommunication companies’ financial performance. Our methodology allows us to analyze the movements of the largest telecommunications companies, to see how companies perform financially compared to their competitors, what they are good at, who are the major competitors in this industry, etc. The dataset contains 88 companies from five different regions: Asia, Canada, Continental Europe, Norther...
We construct a benchmarking model in the form of a two- dimensional self-organising map (SOM) to ... more We construct a benchmarking model in the form of a two- dimensional self-organising map (SOM) to compare the performance of non- banking financial institutions (NFIs) in Romania. The NFIs are characterized by a number of performance dimensions such as capital adequacy, assets' quality and profitability. First, we apply Kohonen' SOM algorithm (an unsupervised neural network algorithm) to group the NFIs in clusters with similar characteristics. Then, we use the U-matrix method to build maps that facilitate the visualization of SOM results and select the best map in terms of quantisation error and ease of readability. The best map is used to analyze the companies over time by studying the cluster where each company was positioned for each period. We conclude that there are benefits in using SOM for interpreting large and complex financial data by identifying and visualizing clusters.
In this article we propose a modified version of Fuzzy CMeans (FCM) clustering algorithm in order... more In this article we propose a modified version of Fuzzy CMeans (FCM) clustering algorithm in order to better allocate the uncertain observations in the clusters. We change the objective function of the classic FCM by attaching different weights to the distances between observations and the clusters’ centers. We apply the modified FCM (Weighting FCM) to model the performance of non-banking financial institutions (NFIs) in Romania. We extend the experiment from our previous work by improving NFIs’ performance dataset from 3 to 8 performance ratios and from 44 to 769 observations. The results show a significant improvement in pattern allocation with the new proposed algorithm.
In this paper we analyze comparatively the macroeconomic performance of different Central and Eas... more In this paper we analyze comparatively the macroeconomic performance of different Central and Eastern European countries by the means of Data Mining (DM) techniques. We analyze the economic situations of three EU countries (Poland, Slovenia and Latvia), a newlyaccepted one (Romania), and other two non-EU countries (Russia and Ukraine). We have depicted economic performance of countries using a number of macroeconomic variables for the time period from 1993 till 2000. The economic variables that we used were: Currency Value (CV) (the inverse of the Exchange Rate - ER), Domestic Prime Rate (Refinancing Rate – RR), Industrial Output (IO) compared to previous periods in percentages, Unemployment Rate (UR), Foreign Trade (FT) in millions of USA dollars. The dataset consists of monthly/annual data during the period 1993-2000, in total 225 cases with five variables each. As DM techniques we firstly applied Self-Organizing Map (SOM) to group the countries according to their economic perform...
We use C-Means clustering algorithm to assess comparatively the performance of non-banking financ... more We use C-Means clustering algorithm to assess comparatively the performance of non-banking financial institutions (NFIs) in Romania. Firstly, we consider this real world application as a knowledge discovery problem (Data Mining) and we engage in a thorough literature review regarding the application of Data Mining methods in assessing comparatively companies' financial performance. Then, we apply C-Means on our NFIs' performance dataset. Next, we compare our results with those obtained when we apply a neural network-based clustering algorithm, called Self-organising Map (SOM) algorithm. The results show that C-Means can offer a good alternative to more sophisticated algorithms in terms of accuracy of the obtained clusters. However, the visualization capability of SOM should be taken into account when we are interested in how companies evolve over time.
In this paper we apply a new methodology presented in our previous work to classify telecom compa... more In this paper we apply a new methodology presented in our previous work to classify telecom companies in respect to their financial performance. We have two goals: to validate our methodology and, using it, to gain insights in a turbulent period in the telecommunications sector. We have obtained higher accuracy rates for the classification models than in our previous studies, and smaller differences between training and test dataset accuracy rates. The two classification techniques have performed similarly in terms of accuracy rates (decision tree, slightly better) and class predictions (multinomial logistic regression, slightly more optimistic). We have analyzed the movements of Scandinavian telecommunications companies. The results are similar to previous findings and show a strong connectivity with what had really happened to Scandinavian telecommunication companies during the second part of the last decade. Abstract . Within classical econometric modelling, because of the complexity of "data - generating process" we have chosen it to be presented with the help of a set of assumptions, where the random is as controlled as possible, so that the activity should be monitored and even dosed. This paper has chosen a systemic and cybernetic approach of the display of the "data - generating process". The first stage of the "data - generating process" decoding, the authors study a system made of three random variables defined on different probability space between which there is a special a priori dependence, called β dependence. The research includes setting a probabilistic model of such a system and making a representative example. The research ends with an analysis of the systemic and cybernetic repercussions corresponding to the β dependence.
Nowadays, the Internet comprises of huge amount of electronic information concerning different co... more Nowadays, the Internet comprises of huge amount of electronic information concerning different companies' financial performance. This amount greatly exceeds our capacity to analyze it, the problem being that we often lack tools to quickly and accurately process these data. DM techniques are interesting mechanisms that can be applied to rapidly changing industries, in order to get an overview of the
Financial decision makers are challenged by the access to massive amounts of both numeric and tex... more Financial decision makers are challenged by the access to massive amounts of both numeric and textual financial information made achievable by the Internet. They are in need of a tool that makes possible rapid and accurate analysis of both quantitative and qualitative information, in order to extract knowledge for decision making. In this paper we propose a conceptual model of a knowledge-building system for decision support based on a society of software agents, and data and text mining methods.
Proceedings of the International Conference on Applied Statistics
In this paper we evaluate comparatively the performance of non-banking financial institutions in ... more In this paper we evaluate comparatively the performance of non-banking financial institutions in Romania by the means of unsupervised neural networks in terms of Kohonen’ Self-Organizing Maps algorithm. We create a benchmarking model in the form of a two-dimensional map (a self-organizing map) that can be used to assess visually the performance of non-banking financial institutions based on different performance dimensions, such as capital adequacy, assets’ quality and profitability. We use the following indicators: Equity ratio (Leverage) for the capital adequacy dimension, Loans granted to clients (net value) / total assets (net value) for the assets’ quality dimension and Return on assets (ROA) for the profitability dimension. We have excluded from our analysis the other three dimensions used in evaluating the performance of banks, due to lack of data (for the two qualitative dimensions: quality of ownership and management) and irrelevance with the NFIs’ sector (liquidity). The p...
Proceedings of the International Conference on Applied Statistics
This paper assesses the financial performance of Romania’s non-banking financial institutions (NF... more This paper assesses the financial performance of Romania’s non-banking financial institutions (NFIs) using a neural network training algorithm proposed by Kohonen, namely the Self-Organizing Maps algorithm. The algorithm takes the financial dataset and positiones each observation into a self-organizing map (a two-dimensional map) which can be latter used to visualize the trajectories of an individual NFI and explain it based on different performance dimensions, such as capital adequacy, assets’ quality and profitability. Further, we use the map as an early-warning system that would accurately forecast the NFIs future performance (whether they would stay or be eliminated from the NFI’s Special Register three quarters into the future). The results are promising: the model is able to correctly predict NFIs’ performance movements. Finally, we compared the results of our SOM-based model with those obtained by applying a multivariate logit-based model. The SOM model performed worse in dis...
Technological and Economic Development of Economy, 2017
In this paper we propose a two-stage methodology to classify the non-banking financial institutio... more In this paper we propose a two-stage methodology to classify the non-banking financial institutions (NFIs) based on their financial performance. The first stage of the methodology consists of grouping the companies in similar financial performance classes (e.g.: “good”, “average”, “poor” performance classes). We optimise the allocation of the observations within the performance clusters by applying an enhanced version of an observation re-allocation procedure proposed in our previous work. Next, based on the result of the grouping phase, we construct a performance class variable by attaching a performance label to each data row. Then, in the second phase of our methodology, we propose a feed-forward neural-network classification model that maps the input space to the newly-constructed performance class variable. This model allows us to forecast the performance of new companies as data become available.
... Tomas Eklund Turku Centre for Computer Science and IAMSR / Åbo Akademi University Lemminkäise... more ... Tomas Eklund Turku Centre for Computer Science and IAMSR / Åbo Akademi University Lemminkäisenkatu 14 B, FIN-20520 Turku, Finland Jonas Karlsson IAMSR / Åbo Akademi University Lemminkäisenkatu 14 B, FIN-20520 Turku, Finland ...
Uploads
Papers by Adrian Costea