For many years, the relevant analyses have indicated problems in the functioning of the small-sized enterprises sector. It has been marked by recession, changes in legislation, financing problems, poor cash flow, dependence on large and state-owned companies and the low rate of survival. At the same time, the small-sized enterprises sector has been referred to as a generator of economic development, because of its size, innovation and adaptability. The uncertainties of the business future are forcing small enterprises to adopt ad hoc decisions, which are based on unstructured information. The most common aims are to reduce the current tax base, increase current liquidity and find financing sources. The aims are not associated with future obligations or the core business development. Decisions are rarely based on financial analysis, which requires a high-quality accounting information system, financial information about other enterprises, and methods that will generate usable and transparent information about several aspects of the business. Entrepreneurs, macro analysts, bankers and investors need analytical systems that generate simple, understandable and usable information, based on a number of relevant financial indicators. The construction of such analytical systems is the subject of this thesis. The thesis identifies the financial condition of the small-sized enterprises sector on the basis of data derived from the correlations between positions of the fundamental financial statements, that is, the financial indicators of liquidity, debt, activity, efficiency and profitability. Forensic indicators and the interdependency of economic and accounting aspects of profitability are also analyzed. Successful exploitation of the collected data has required the use of data mining methods which have advantage over traditional statistical methods for their classification and prediction capabilities, as well as the ability to work with non-linear relationships between features. Self-organizing neural networks have been used for clustering, the decision tree for classification and prediction of the new data set, while the association rules algorithm has been used to determine the relation between the created cluster models. Current global and domestic researches use these methods because they provide aggregated information in predicting financial distress and bankruptcy, fraud detection, credit risk assessment, measurement and comparison of financial performance. Based on a sample of 2,200 financial statements, which entities submitted in 2011 and 2012, models were built that successfully identify the current state of the small-sized enterprise sector and classify a new, unknown set of business entities. Each of the six cluster model has identified, described and proved the existence of 5 to 12 patterns that occur within the sector. In total, 48 clusters have been identified, whose sustainability is proven by clustering quality indexes. In addition, the research has found the rules which connect some clusters of different models. This doctoral thesis is a valuable collection of knowledge and skills, necessary for building analytical cluster systems. Through the development of management performance systems, revitalization of practical usage of financial indicators, and laying the foundations of development of financial forensic systems, the thesis contributes to a better understanding and development of the small-sized enterprises sector. The most important research limitations are related to the stage of data preprocessing. Future research may apply alternative methods of data preprocessing, include other structured and unstructured financial data, or apply the accumulated knowledge to other business sectors.