Monitoring fatigue in sport is critical to achieve elite performance and may benefit from machine learning techniques that are liable to predict changes in fatigue state. In this paper we present and compare different machine learning models to predict the Rate of Perceived Exertion (RPE) of training or game sessions for professional football (soccer) players. We compare different approaches to train predictive models in a supervised setting (regression) with a focus on individualized and group-based approaches, i.e. training a specific model for each player or predefined groups of players (full team or clusters defined using unsupervised learning). Both player-informed and player-agnostic models are compared in the group-based approach, i.e. providing or not player id as feature during training and inference. Compared models have been trained on real data collected during a full season of professional football players, and using among others, anthropometric, running activity, heart rate and weather data. The best results are obtained using a player-informed team-based approach with a Random Forest regressor (0.793 MAE, 1.033 RMSE). Results obtained are competitive with the best reported in the literature for this predictive task in elite Football players.

When solving complex and critical decision problems involving several aspects, collecting decision-maker (DM) preferences remains a delicate issue. Indirect parameter identification procedures are more adequate in such situations than procedures consisting to ask directly the DM to provide preferences. In the case of methods based on additive models such as MACBETH, disaggregation procedures are implemented to handle this problem. In case of sophisticated non-additive model, e.g. Choquet integral based fuzzy measure, most of the existing indirect procedures assume a 2-additive fuzzy measure and the DM is asked to change her/his preferences if they do not match this model. This paper proposes an approach to determine preferences model that achieves a trade-off between model sophistication and the fastidious MACBETH questioning procedure when considering interactions between more than two attributes. First, it proposes a practical approach to solve a real case of multiple criteria decision problem while considering sophistical model of preferences. On one hand, particular binary alternatives are selected for a first questioning procedure to determine the fuzzy measure parameters. In order to make the later procedure as simple as possible to handle, some simplifications are introduced to reduce both the amount of information and the cognitive effort requested from the DM. Our proposition consists in fixing several non-linear optimization problems until reaching the first k for which the k-additive fuzzy measure match the DM preferences. On the other hand, starting from a decision matrix and the answers to a second questioning procedure, value functions are determined. The second contribution is a theoretical one and consists in proving the existence of a k-additive fuzzy measure matching the DM preferences. Finally, the approach is applied to a real world case that concerns the comparison of some electrical and thermal vehicles regarding their environmental impacts.

In this paper we propose methods that can help the decision-makers to find a compromise between willingness to do and ability to do by introducing the difficulty considerations in the multiple criteria decision analysis problems. Two problems are considered: ranking alternatives and improving existing solution. Usually, in the classical approaches of multiple criteria decision analysis, only the degree of satisfaction is considered to compare alternatives. However, sometimes a good alternative is difficult to implement by a decision-maker even if he spends necessary cost for it. First we give the definition of the concept of difficulty function, then we show how to introduce it in decision problems using operators based on fuzzy measures. This allows us to consider interactions between criteria under two aspects: 1) the overall satisfaction resulting from the simultaneous satisfaction or not of certain criteria; 2) the overall difficulty resulting from the difficulty or not of satisfying certain criteria simultaneously. After that, we present two examples of the difficulty function assessment in the case of a non-linear model. Finally, we propose an illustration concerning the problem of managing the students effort when improving their scores on a set of subjects. This illustration focus on the extension of the concept of worth index which quantifies the gain of improvement related to a subset of objectives when it is difficult to improve all the objectives simultaneously.

Belief functions are quite generic models when it comes to represent uncertain data, as it extends a wide range of uncertainty models (possiblity and probability distributions, among others). Usually, belief functions are defined over finite spaces, however many real word problems require to deal with beliefs over a continuous space while maintaining computational efficiency. This paper discusses the case of focal sets on the unit simplex, and proposes efficient inference tools to manipulate them. Such sets can be used to represent unknown proportions that one may face in various fields like soil contamination managing, plastic sorting or image reconstruction. In this paper, we illustrate their use on an industrial problem of plastic sorting, where the proportion of material impurities must not go over a limit while minimizing the rejection of sorted materials, whose nature is uncertain.

Ontologies are core elements of numerous applications that are based on computer-processable expert knowledge. They can be used to estimate the Information Content (IC) of the key concepts of a domain: a central notion on which depend various ontology-driven analyses, e.g. semantic measures. This paper proposes new IC models based on the belief functions theoretical framework. These models overcome limitations of existing ICs that do not consider the inductive inference assumption intuitively assumed by human operators, i.e. that occurrences of a concept (e.g. Maths) not only impact the IC of more general concepts (e.g. Sciences), as considered by traditional IC models, but also the one of more specific concepts (e.g. Algebra). Interestingly, empirical evaluations show that, in addition to modelling the aforementioned assumption, proposed IC models compete with best state-of-the-art models in several evaluation settings.

Within an emergency unit, the head manager is required to make difficult decisions based on experts’ assessments of many criteria, including personal injuries, environmental impacts, and economic and media consequences. Uncertainty in this collective assessment is related to the multiplicity of experts’ points of view and imprecise assessments. We are proposing a decision support system derived from a situation-awareness model, generalized herein to the case of multiple actors. It is able of representing, merging, and aggregating expert assessments. Imprecise criteria assessments are first represented by intervals and then merged in the form of a possibility distribution that keeps track of all the information provided, that is, without any loss of information. Next, a Choquet integral based aggregation is carried out to consider the relative importance of criteria and interactions between criteria in the overall assessment of the foreseeable alternatives to get out of the crisis. Finally, a determination of the contributions of each criterion assessment uncertainty to the overall assessment uncertainty provides useful information to the head manager in controlling the decision deliberation by reducing the inconsistent points in the experts’ assessments. The proposals are applied to the emergency issues resulting from a traffic accident occurring at a grade crossing.