Control loops are the most critical components in many production processes. In this process, the economic yield is strongly linked to the performance of the control loops since aspects such as safety conditions, process quality, and energy and raw material consumption depend on this. However, experience has shown that most of the control loops can be improved by identifying and correcting the causes of the poor perfor-mance. The indices to evaluate the performance of the control loops can be divided into two groups, stochastic and deterministic. The most known of the former is the minimum variance index. Stochastic indices only require data collected under normal operating conditions and minimum knowledge of the process, making it possible to evaluate performance online. However, some disadvantages, such as scale and span problems, make performance analysis difficult. The deterministic indices (rise time, settling time, overshoot, phase and gain margins, etc.) are easy to interpret, facilitating the analysis; however, invasive plant tests are necessary to estimate them, making them impractical. Is it possible to link these two approaches? With that question in mind, in this work, it is proposed to build a model to estimate deterministic indices (to evaluate robustness and performance of control loops), considering stochastic indices and some process information as model inputs. This paper shows the procedure to build the inferential model by using machine learning techniques.