Authors analyze financial data from a set of almost 200 major US companies from both manufacturing and service. The selection of the firms considered the spectrum of sectors of Dow Jones Industrial composite including manufacturing companies in aerospace & defence, automotive, beverages, footware & apparel, health technology, oil & gas and service-oriented companies like consumer services, discount stores, telecom services, insurance. Short overview of recent development in business health modelling is given. Based on the data and known expert ratings, the recently published Stagewise regression algorithm was employed to identify the most relevant predictors out of all possible financial ratios as based on the data. Support Vector Machine was used subsequently to construct a quantitative probabilistic prediction model for business investment risk evaluation. With this approach, it is possible to build less rigid, more specific models suitable for smaller sectors an shorter periods, thus allowing investors and management to better react to dynamic changes in business environment. Quantitative prediction based on Support Vector Machine (SVM) models provides more information and better decission support than traditional binary prediction (good/bad). The paper provides directly applicable parametric decission models predicting both numerical rating and good/bad classification probability  for manufacturing and service sectors. Short-term and more specific models can be used to characterize not only the business subject themselves, but also to characterize, parametrize and compare business environments.