Takayuki Sekiya

Analysis of computer science related curriculum on LDA and Isomap

A good curriculum is crucial for a successful university education. When developing a curriculum, topics, such as natural science, informatics, and so on are set first, course syllabi are written accordingly. However, the topics actually by the courses are not guaranteed to be identical to the initially set topics. To find out if the actual topics are covered by the developed course syllabi, we developed a method of systematically analyzing syllabi that uses latent Dirichlet allocation (LDA) and Isomap. We applied this method to the syllabi of MIT and those of the Open University, and verified that the method is effective.

Case studies on ontology for the knowledge intensive engineering framework

Knowledge intensive engineering is a new style of engineering in which engineering knowledge is used in a flexible and integrated manner and aims at generating more added value. A Knowledge Intensive Engineering Framework (KIEF) is a computational framework for knowledge intensive engineering. To develop KIEF, we have to build a Very Large-scale Knowledge Base (VLKB) that contains a wide variety of engineering design knowledge, from commonsense knowledge about the physical world to domain specific knowledge systematized as physics theories. VLKB for KIEF has a system of fundamental concepts called “ontology.” This paper describes the role of the ontology in KIEF, and validates the effectiveness of our approach through case studies.

Mining Interesting Rules by Association and Classification Algorithms

The important goal in data mining is to reveal hidden knowledge from data and various algorithms have been proposed so far. But the problem is that typically not all rules are interesting - only small fractions of the generated rules would be of interest to any given user. Hence, numerous measures such as confidence, support, lift, information gain, and so on, have been proposed to determine the best or most interesting rules. However, some algorithms are good at generating rules high in one interestingness measure but bad in other interestingness measures. The relationship between the algorithms and interestingness measures of the generated rules is not clear yet. In this paper, we studied the relationship between the algorithms and interesting measures. We used synthetic data so that the obtained result is not limited to specific cases. We report our experimental results and present the best combination between algorithms and parameters in order to generate interesting rules.

Classification of Knowledge for Generating Engineering Models

During design, a designer uses various computational tools, such as a geometric modeling system, analysis tools, and databases. To support these design processes, a system that can integrate such computational tools and support him/her to use the tools and to build, operate and modify models is required. For this purpose, we have been developing the Knowledge Intensive Engineering Framework (KIEF) system. This paper describes the structure of knowledge model used in a modeling process of engineering models. We analyzed engineering knowledge used for Finite Element Analysis (FEA), categorized it, and finally built an environment to support FEA on KIEF.

Development of a curriculum analysis tool

A good curriculum is crucial for a successful university education. When developing a curriculum, topics, such as economics, natural science, informatics, etc. are set first, and course syllabi are written accordingly. However, the topics actually covered by the course syllabi are not guaranteed to be identical to the initially set topics. To find out if the actual topics covered by the developed course syllabi, we developed a method of systematically analyzing course syllabi that uses latent Dirichlet allocation (LDA) and Isomap. In this paper, we propose the web-based curriculum analysis tool with this method, and demonstrate an example of the way the tool is used for analyzing computer science curricula.

Skill hierarchy revised by SEM and additional skills

In this paper, we reexamine the Lopezs skill hierarchy by using structural equation modeling (SEM), by which we can explicitly see the relationship between variables. We also introduce two more skills, a skill to modify a given piece of code to satisfy a given specification and a skill to write code from scratch to satisfy a given specification. The analysis shows that these additional skills are useful. The path diagram (model) with the highest fitness measures suggests an effective order of training and pivotal roles of the Tracing2 and Explain skills in training.

Design Knowledge Collection by Modeling

Since design is a process in which designers use various kinds of knowledge and it is an open-ended problem, it is difficult to prepare all necessary knowledge before design. Also, due to advances of technology, design knowledge is largely fragmental and compartmentized. This makes the communication among design experts or design agents difficult. Due to these, a knowledge sharing mechanism among agents based on a common ontology is needed even for collecting design knowledge. In this paper, first we analyze the nature and structure of design knowledge. Then, we propose a knowledge collection methodology through design object modeling and verify the method with an example of the suspension design for a motorcycle. Finally, we identify knowledge that should be prepared before the design process and that can be collected during modeling processes.

Mapping analysis of CS2013 by supervised LDA and isomap

The ACM Education Board and the IEEE Computer Societys Education released Computing Science Curricula CS2013 in December 2013. In this paper, we applied our curriculum analysis method to analyze the Computing Science Curricula CS2013. By comparing the maps and the topic words of previous curricula CS2008 and CS2013, we found some significantly changed Knowledge Units (KUs) and Knowledge Areas (KAs). We also found other significant changes of CS2013 by using the actual courses of the three universities. Moreover, we discovered that the Body of Knowledge (BOK) of CS2013 had higher explainability that those of other previous computing curricula.

Relationship between programming concepts underlying programming skills

We investigated the relationship between programming skills and that of programming concepts in the introductory programming course in this paper. We introduced Modification and Programming skills from other existing skills. From the results of the midterm and final exams of our course, we got the scores in the skills, and from them we estimated the ability in concepts. By correlation, contingency, and covariance of the result of examinations, we found out that Modification and Programming are different from Tracing. We found things about concepts too, that some concepts are local while others are global, that their thinkings sometimes conflict, that the conditional statement is difficult than the loop, that the nested one is significantly difficult than the not nested one, and that understanding the concept of a structure is crucial for understanding the conditional structure and nested loop. We also found that the skills get more separated according to the development of abilities of students.

Discovery of Interconnection Among Knowledge Areas of Standard Computer Science Curricula by a Data Science Approach

Computer Science Curricula 2013 (CS2013) is a widely-used standard curricula of computer science, which has been developed jointly by the ACM and the IEEE Computer Society. CS2013 consists of 18 Knowledge Areas (KAs) such as Programming Languages and Software Engineering. Though it is obvious that there are strong interconnections among the KAs, it was hard to investigate the interconnections objectively and quantitatively. In this paper, the interconnections among the KAs of CS2013 are investigated by a data science approach. For this purpose, a collection of actual syllabi from the world’s top-ranked universities was constructed. Then, every actual syllabus is projected to the KA space by a probabilistic model-based method named simplified, supervised Latent Dirichlet Allocation (denoted by ssLDA). Consequently, the following interesting properties of the interconnections among the KAs were discovered: (1) There are the high interconnections among the KAs in each syllabi; (2) A plausible hierarchical structure of the KAs is found by utilizing the interconnections; (3) The structure shows that the KAs are classified into the three principal independent factors (HUMAN, THEORY, and IMPLEMENTATION). The factor of IMPLEMENTATION can be divided into PROGRAMMING and SYSTEM. The factor of SYSTEM can be divided further into DEVICES and NETWORK.