Asymmetric Missing-data Problems: Overcoming the Lack of Negative Data in Preference Ranking

In certain classification problems there is a strong a asymmetry between the number of labeled examples available for each of the classes involved. In an extreme case, there may be a complete lack of labeled data for one of the classes while, at the same time, there are adequate labeled examples for the others, accompanied by a large body of unlabeled data. Since most classification algorithms require some information about all classes involved, label estimation for the un-represented class is desired. An important representative of this group of problems is that of user interest/preference modeling where there may be a large number of examples of what the user likes with essentially no counterexamples.Recently, there has been much interest in applying the EM algorithm to incomplete data problems in the area of text retrieval and categorization. We adapt this approach to the asymmetric case of modeling user interests in news articles, where only labeled positive training data are available, with access to a large corpus of unlabeled documents. User modeling is here equivalent to that of user-specific document ranking. EM is used in conjunction with the Naive Bayes model while its output is also utilized by a Support Vector Machine and Rocchio's technique.Our findings demonstrate that the EM algorithm can be quite effective in modeling the negative class under a number of different initialization schemes. Although primarily just the negative training examples are needed, a natural question is whether using all of the estimated labels (i.e., positive and negative) would be more (or less) beneficial. This is important considering that, in this context, the initialization of the negative class for EM is likely not to be very accurate. Experimental results suggest that EM output should be limited to negative label estimates only.     

Over-education and its opportunity cost in Japan

In this paper, we investigate the determinants of over-education in Japan and evaluate its opportunity costs for university graduates. To this end, we use the REFLEX data. Results reveal that over-education level in Japan is high and it brings an important wage penalty for Japanese workers. Large firm and high occupations point toward a significant reduction in the likelihood of over-education. Results for wage regressions for over-education indicate that Japanese workers who achieve jobs in large organizations will experience 20 % increase in their wage due to the firm size and, if over-educated, their increase in wage will remain positive. Results for women are puzzling as there is no significant effect associated with gender.     

Why do tertiary education graduates regret their study program? A comparison between Spain and the Netherlands

In this paper we investigate the determinants of regret of study program for tertiary education graduates in Spain and the Netherlands. These two countries differ in their educational system in terms of the tracking structure in their secondary education and the strength of their education-labor market linkages in tertiary education. Therefore, by comparing Spain and the Netherlands, we aim at learning about the consequences that the two educational systems might have on the regret of study program in tertiary education. Basing on the psychological literature on regret, we derive some expectations on the determinants of regret of study program. Results reveal that both, education track and education-labor mismatch of tertiary education, are important determinants of the likelihood of program regret. Results allow us to derive some policy recommendations on the tertiary education system.     

A Conceptual View of Web-Based E-Learning Systems

Starting from a general framework for web-based e-learning systems that is based on an abstraction layer model, this paper presents a conceptual modelling approach, which captures the modelling of learners, the modelling of courses, the personalisation of courses, and the management of data in e-learning systems. Courses are modelled by outline graphs, which are further refined by some form of process algebra. The linguistic analysis of word fields referring to an application domain helps to set up these course outlines. Learners are modelled by classifying value combinations for their characteristic properties. Each learner type gives rise to intentions as well as rights and obligations in using a learning system. Intentions can be formalised as postconditions, while rights and obligations lead to deontic constraints. The intentions can be used for the personalisation of the learning system to a learner type. Finally, the management of data in an e-learning system is approached on two different levels dealing with the content of individual learning units and the integrated content of the whole system, respectively. This leads to supporting databases and views defined on them.     

Exploitation characteristics of the inspec data base in the SDI system at the Technical University of Wroc aw

The investigation results as well as the conclusions which can be drawn from them, on the subject of exploitation the INSPEC data base are presented in this paper. The investigations were concerned with the length of information queries and the number of answers as well as the relationship between them, depending on the character of the user. The results obtained suggest the possibility of lowering the costs of exploitation of the INSPEC data base in the SDI system.