Upgrading the Fusion of Imprecise Classifiers.

Imprecise classification is a relatively new task within Machine Learning. The difference with standard classification is that not only is one state of the variable under study determined, a set of states that do not have enough information against them and cannot be ruled out is determined as well. For imprecise classification, a mode called an Imprecise Credal Decision Tree (ICDT) that uses imprecise probabilities and maximum of entropy as the information measure has been presented. A difficult and interesting task is to show how to combine this type of imprecise classifiers. A procedure based on the minimum level of dominance has been presented; though it represents a very strong method of combining, it has the drawback of an important risk of possible erroneous prediction. In this research, we use the second-best theory to argue that the aforementioned type of combination can be improved through a new procedure built by relaxing the constraints. The new procedure is compared with the original one in an experimental study on a large set of datasets, and shows improvement.

A Variation of the Algorithm to Achieve the Maximum Entropy for Belief Functions.

Evidence theory (TE), based on imprecise probabilities, is often more appropriate than the classical theory of probability (PT) to apply in situations with inaccurate or incomplete information. The quantification of the information that a piece of evidence involves is a key issue in TE. Shannon's entropy is an excellent measure in the PT for such purposes, being easy to calculate and fulfilling a wide set of properties that make it axiomatically the best one in PT. In TE, a similar role is played by the maximum of entropy (ME), verifying a similar set of properties. The ME is the unique measure in TE that has such axiomatic behavior. The problem of the ME in TE is its complex computational calculus, which makes its use problematic in some situations. There exists only one algorithm for the calculus of the ME in TE with a high computational cost, and this problem has been the principal drawback found with this measure. In this work, a variation of the original algorithm is presented. It is shown that with this modification, a reduction in the necessary steps to attain the ME can be obtained because, in each step, the power set of possibilities is reduced with respect to the original algorithm, which is the key point of the complexity found. This solution can provide greater applicability of this measure.

Decision Tree Ensemble Method for Analyzing Traffic Accidents of Novice Drivers in Urban Areas.

Presently, there is a critical need to analyze traffic accidents in order to mitigate their terrible economic and human impact. Most accidents occur in urban areas. Furthermore, driving experience has an important effect on accident analysis, since inexperienced drivers are more likely to suffer fatal injuries. This work studies the injury severity produced by accidents that involve inexperienced drivers in urban areas. The analysis was based on data provided by the Spanish General Traffic Directorate. The information root node variation (IRNV) method (based on decision trees) was used to get a rule set that provides useful information about the most probable causes of fatalities in accidents involving inexperienced drivers in urban areas. This may prove useful knowledge in preventing this kind of accidents and/or mitigating their consequences.

Extracting decision rules from police accident reports through decision trees.

Given the current number of road accidents, the aim of many road safety analysts is to identify the main factors that contribute to crash severity. To pinpoint those factors, this paper shows an application that applies some of the methods most commonly used to build decision trees (DTs), which have not been applied to the road safety field before. An analysis of accidents on rural highways in the province of Granada (Spain) between 2003 and 2009 (both inclusive) showed that the methods used to build DTs serve our purpose and may even be complementary. Applying these methods has enabled potentially useful decision rules to be extracted that could be used by road safety analysts. For instance, some of the rules may indicate that women, contrary to men, increase their risk of severity under bad lighting conditions. The rules could be used in road safety campaigns to mitigate specific problems. This would enable managers to implement priority actions based on a classification of accidents by types (depending on their severity). However, the primary importance of this proposal is that other databases not used here (i.e. other infrastructure, roads and countries) could be used to identify unconventional problems in a manner easy for road safety managers to understand, as decision rules.