A Framework to Predict Failures for Ground Tests on Aircrafts Table 1. (Continued ) Parameter Meaning INCIDENCE_TYPE The type of abortive incidence (if there is one registered in the corresponding test). DESCRIPTION The description of the abortive incidence (if there is one registered in the corresponding test) Figure 1. Distribution of test executions with respect to abortive incidence (simulated data). Figure 2. Distribution of test executions with respect to the nonabortive incidence (simulated data). Figure 3. Distribution of test executions with respect to any type of incidence (simulated data). Figure 4. Importance of the STATION parameter on the incidences (simulated data). types of algorithms (as opposed to, for example, artificial neural networks) for our objective: - An explanatory component was added to the data of the probability. Since the model is translated into a set of rules, and these rules carry implicit an additional explanation. - Allocated a probability to each of the records in the database. The set of rules allowed us to have a probability value for each pattern of the test execution found in the database. This probability 34 is obtained from the proportion of target occurrences in each leaf of the tree. In addition, the use of these types of algorithms as predictors is well extended and validated in the research literature [13]-[16]. Specifically, among the different decision tree algorithms, such as C4.5, CHAID, and C&RT, we decided to use C&RT (classification and regression trees) which obtained the best results in the tests carried out. Some advantages of the C&RT algorithm are as follows: IEEE A&E SYSTEMS MAGAZINE MAY 2019