Predicting an individual’s risk of getting a stroke has been a research subject for many authors worldwide since it is a frequent illness and there is strong evidence that early awareness of having that risk can be beneficial for prevention and treatment. Many Governments have been collecting medical data about their own population with the purpose of using artificial intelligence methods for making those predictions. The most accurate ones are based on so called black-box methods which give little or no information about why they make a certain prediction. However, in the medical field the explanations are sometimes more important than the accuracy since they allow specialists to gain insight about the factors that influence the risk level. It is also frequent to find medical information records with some missing data. In this work, we present the development of a prediction method which not only outperforms some other existing ones but it also gives information about the most probable causes of a high stroke risk and can deal with incomplete data records. It is based on the Dempster-Shafer theory of plausibility. For the testing we used data provided by the regional hospital in Okayama, Japan, a country in which people are compelled to undergo annual health checkups by law. This article presents experiments comparing the results of the Dempster-Shafer method with the ones obtained using other well-known machine learning methods like Multilayer perceptron, Support Vector Machines and Naive Bayes. Our approach performed the best in these experiments with some missing data. It also presents an analysis of the interpretation of rules produced by the method for doing the classification. The rules were validated by both medical literature and human specialists.
People’s health care cost prediction is nowadays a valuable tool to improve accountability in health care. In this work, we study if an interpretable method can reach the performance of black-box methods for the problem of predicting health care costs. We present an interpretable regression method based on the Dempster-Shafer theory, using the Evidence Regression model and a discount function based on the contribution of each dimension. Optimal parameters are learned using gradient descent. The k-nearest neighbors’ algorithm was also used to speed up computations. With the transparency of the evidence regression model, it is possible to create a set of rules based on a patient’s vicinity. When making a prediction, the model gives a set of rules for such a result. We used Japanese health records from Tsuyama Chuo Hospital to test our method, which includes medical checkups, exam results, and billing information from 2016 to 2017. We compared our model to an Artificial Neural Network and Gradient Boosting method. Our results showed that our transparent model outperforms the Artificial Neural Network and Gradient Boosting with an of .
Although many authors have highlighted the importance of predicting people’s health costs to improve healthcare budget management, most of them do not address the frequent need to know the reasons behind this prediction, i.e., knowing the factors that influence this prediction. This knowledge allows avoiding arbitrariness or people’s discrimination. However, many times the black box methods (that is, those that do not allow this analysis, e.g., methods based on deep learning techniques) are more accurate than those that allow an interpretation of the results. For this reason, in this work, we intend to develop a method that can achieve similar returns as those obtained with black box methods for the problem of predicting health costs, but at the same time it allows the interpretation of the results. This interpretable regression method is based on the Dempster-Shafer theory using Evidential Regression (EVREG) and a discount function based on the contribution of each dimension. The method “learns” the optimal weights for each feature using a gradient descent technique. The method also uses the nearest k-neighbor algorithm to accelerate calculations. It is possible to select the most relevant features for predicting a patient’s health care costs using this approach and the transparency of the Evidential Regression model. We can obtain a reason for a prediction with a k-NN approach. We used the Japanese health records at Tsuyama Chuo Hospital to test our method, which included medical examinations, test results, and billing information from 2013 to 2018. We compared our model to methods based on an Artificial Neural Network, Gradient Boosting, Regression Tree and Weighted k-Nearest Neighbors. Our results showed that our transparent model performed like the Artificial Neural Network and Gradient Boosting with an of .
Foot traffic, conversion rate, and total sales during a period of time may be considered to be important indicators of store performance. Forecasting them may allow for business managers plan stores operation in the near future in an efficient way. This work presents a regression method that is able to predict these three indicators based on previous data. The previous data includes values for the indicators in the recent past; therefore, it is a requirement to have gathered them in a suitable manner. The previous data also considers other values that are easily obtained, such as the day of the week and hour of the day of the indicators. The novelty of the approach that is presented here is that it provides a confidence interval for the predicted information and the importance of each parameter for the predicted output values, without additional processing or analysis. Real data gathered by Follow Up, a customer experience company, was used to test the proposed method. The method was tried for making predictions for up to one month in the future. The results of the experiments show that the proposed method has a comparable performance to the best methods proposed in the past that do not provide confidence intervals or parameter rankings. The method obtains RMSE of for foot traffic prediction, for conversion rate forecasting, and for sales prediction.
The pedagogical usability is an important characteristic of applications that support learning as it relates to the added value students perceive while using it for learning. A good pedagogical usability means that an application has more chances to be accepted and used by the students thus raising the possibilities that students actually will learn with it. However important, this concept tends to be neglected by many authors. In this work we show how this concept can be applied to evaluate an application by presenting a real example of an application that has been re-designed to improve its usability, thus showing how pedagogical usability can be operationalized to be applied in general. The application shown in this work is called RedCoMulApp (Reading Collaborative Multiple-option Application) and its goal is to raise the reading comprehension ability of 12th grade high school students. In order to capture the pedagogical usability, we used 12 metrics to design a questionnaire with 26 questions to be answered with a 5-point Likert-scale, plus two open-ended questions to capture aspects that were positively valued, and those that should be improved. The results for the pedagogical usability allow us to validate that the design of the application was perceived by the students as pedagogically useful to learning about reading comprehension.
Preparing a plan for reaction to a grave emergency is a significant first stage in disaster management. A group of experts can do such preparation. Best results are obtained with group members having diverse backgrounds and access to different relevant data. The output of this stage should be a plan as comprehensive as possible, taking into account various perspectives. The group can organize itself as a collaborative decision-making team with a process cycle involving modeling the process, defining the objectives of the decision outcome, gathering data, generating options and evaluating them according to the defined objectives. The meeting participants may have their own evidences concerning people’s location at the beginning of the emergency and assumptions about people’s reactions once it occurs. Geographical information is typically crucial for the plan, because the plan must be based on the location of the safe areas, the distances to move people, the connecting roads or other evacuation links, the ease of movement of the rescue personnel, and other geography-based considerations. The paper deals with this scenario and it introduces a computer tool intended to support the experts to prepare the plan by incorporating the various viewpoints and data. The group participants should be able to generate, visualize and compare the outcomes of their contributions. The proposal is complemented with an example of use: it is a real case simulation in the event of a tsunami following an earthquake at a certain urban location.