Item Response Theory Based Ensemble in Machine Learning

In this article,we propose a novel probabilistic framework to improve the accuracy of a weighted majority voting algorithm.In order to assign higher weights to the classifiers which can correctly classify hard-to-classify instances,we introduce the item response theory(IRT)framework to evaluate the samples′difficulty and classifiers′ability simultaneously.We assigned the weights to classifiers based on their abilities.Three models are created with different assumptions suitable for different cases.When making an inference,we keep a balance between the accuracy and complexity.In our experiment,all the base models are constructed by single trees via bootstrap.To explain the models,we illustrate how the IRT ensemble model constructs the classifying boundary.We also compare their performance with other widely used methods and show that our model performs well on 19 datasets.

A Probabilistic Framework for Temporal Cognitive Diagnosis in Online Learning Systems

Cognitive diagnosis is an important issue of intelligent education systems,which aims to estimate students'proficiency on specific knowledge concepts.Most existing studies rely on the assumption of static student states and ig-nore the dynamics of proficiency in the learning process,which makes them unsuitable for online learning scenarios.In this paper,we propose a unified temporal item response theory(UTIRT)framework,incorporating temporality and random-ness of proficiency evolving to get both accurate and interpretable diagnosis results.Specifically,we hypothesize that stu-dents'proficiency varies as a Wiener process and describe a probabilistic graphical model in UTIRT to consider temporali-ty and randomness factors.Furthermore,based on the relationship between student states and exercising answers,we hy-pothesize that the answering result at time k contributes most to inferring a student's proficiency at time k,which also re-flects the temporality aspect and enables us to get analytical maximization(M-step)in the expectation maximization(EM)algorithm when estimating model parameters.Our UTIRT is a framework containing unified training and inferenc-ing methods,and is general to cover several typical traditional models such as Item Response Theory(IRT),multidimen-sional IRT(MIRT),and temporal IRT(TIRT).Extensive experimental results on real-world datasets show the effective-ness of UTIRT and prove its superiority in leveraging temporality theoretically and practically over TIRT.

Evaluation of A Scale of Patient-Reported Outcomes for the Assessment of Myasthenia Gravis Patients in China

Objective： To evaluate a scale of patient-reported outcomes for the assessment of myasthenia gravis patients （MG-PRO） in China. Methods： A total of 100 MG patients were interviewed for the field testing. Another 56 MG patients were selected and assessed with the MG-PRO scale before treatment and at 1, 2 and 4 weeks after treatment. The classical test theory and item response theory （IRT） were used to assess the psychometric characteristics of the MG-PRO scale, Results： The MG-PRO scale included 4 dimensions： physical, psychological, social environment, and treatment. Confirmatory factor analysis showed that each dimension was consistent with the theoretical construct. The scores of the physical and psychological dimensions increased significantly at 1 week after treatment （P〈0.05）. All the dimension scores and the MG-PRO score increased significantly at 2 and 4 weeks after treatment （P〈0.05）. IRT showed that person separation indices were greater than 0.8, most of the item fit residual statistics were within ＋ 2.5, and no item had uniform or non-uniform differential item functioning （DIF） between gender and age （〈40, 〉140）. Conclusions： The MG-PRO scale is valid for measuring the quality of life （QOL） of MG patients, with good reliability, validity, responsiveness, and good psychometric characteristics from IRT. It can be applied to evaluate the QOL of MG patients and to assess treatment effects in clinical trials.

Protective behaviors regarding coastal flooding risk in a context of climate change

The physical vulnerability of coastal areas due to rising sea level and the flooding risk consequent,does not guarantee the implementation of protective behaviors by these risk zones’inhabitants.This study aims to establish the link between the willingness to carry out protective behaviors and physical and perceived indicators of vulnerability.A typology of coastal flooding vulnerability,uses various physical indicators and their perceived counterparts which have been collected from 490 inhabitants of Cartagena(Colombia,declared world heritage of humanity by UNESCO in 1984),resident in areas of coastal flooding risks.The item-response theory(IRT)approach has been used.The results reveal that the implementation of protective behaviors is more related to perceived indicators,such as distance to the sea,than to actual physical vulnerability.We observe that physical vulnerability is linked to the intention to carry out protective behaviors.The presence of a defensive structure against coastal flooding could be considered as a visual cue and be a good predictor of the willingness to carry out protective behaviors.On the contrary,people in the most vulnerable situation(single-storey house)do not demonstrate a higher level of willingness to carry out protective behavior,as well of participants who lived in residential buildings which have demonstrated lower level of willingness to carry out such behaviors.Therefore,vulnerability of the house is not seen as a criterion that encourages participants to better protect themselves.