Expanding the portfolio of instruments for finance management of an enterprise in order in order to increase the return on investments is a task of current interest. In the article, we discuss a model of the trading decision support system in financial markets based on probability analysis and machine learning, which can be used to solve the above problem. The aim of this work is to develop and test a model of the decision support system for trading operations with stock financial instruments as a part of the enterprise financial management process. The model is based on machine learning technologies that provide acquisition of large amounts of input data, its primary processing, the formation of a multi-dimensional space of feature vectors, and its transformation. The forecasting method is based on the Bayesian rule. The obtained Bayesian probabilities are stored in a hypercube which is used to determine the rules for trading decisions making. The developed model was tested on historical data of the futures market of the Moscow Exchange in the case of the RTS Index Futures as the main instrument for transactions and the USD-RUB Futures as an auxiliary instrument used for analysis. To evaluate the results of testing, quantitative metrics have been developed, which include the number and volume of profitable and unprofitable transactions, the average profit/loss per transaction. These metrics were used for analysis of effectiveness and limits of applicability for the developed model. The model can be implemented as a software HFT robot that can provide the probability to get profit greater than the probability of losses. As a further step in the development of this topic, research can be undertaken on the mechanisms for the formation of feature vectors using data mining methods.