head img  
NTU Taiwan University img


The NTU Research Team of Department of Computer Science and Integrated Engineering Won the World Champion from the 2012 ACM KDD Cup under the Title of Predicting the Click-through Rate

This August in Beijing, the NTU research team led by Professor Shou- De Lin and Professor Hsuan- Tien Lin of Department of Computer Science and Integrated Engineering won the world champion under the title of predicting the click-through rate from the most prestigious competition, ACM KDD Cup, on the field of knowledge discovery and data mining.

This medal of champion is the fifth prize in a row that National Taiwan has already won from this prestigious research competition over the past five years, including the world champions back in 2008, 2010 and 2011 as well. Such a great performance set an outstanding record for ACM KDD CUP ever. The NTU research team just did their presentation with posters in the ACM SIGKDD Conference on Knowledge Discovering and Data Mining and the received the honor of the first prize with USD$5000 in front of thousands of scholars this August in Beijing.

ACM KDD Cup was first held in 1997, hosting by the Special Interest Group on Knowledge Discovery and Data Mining of ACM (Association for Computing Machinery) along with the world top ACM Conference on Knowledge Discovery and Data Mining every year. ACM KDD Cup has been the most important and the well- known competition on the research field of knowledge discovery and data mining.

Every year, the current issues are included in the ACM KDD Cup with a great value of academic challenge and the countless practical value in its business use. Competitors around the world are required to be able to integrate the knowledge of theory development and actual programming code to successfully develop the intellectual techniques and system of data mining within only three to four months so that they can predict the specific CTR (click- through rate) provided by the hosting organization. ACM KDD Cup attracts many elite research teams from both the universities and the corporations around the world, such as Stanford University and IBM Research, making it quite competitive every year.

The NTU research team of the Department of Computer Science and Integrated Engineering has been signing up for the ADM KDD Cup ever since 2008 to compete for the honor of the prize, and they have been the first place for the past five years: In 2008, the NTU research team led by Professor Shou- De Lin successfully designed the intellectual system of breast imaging reporting and data, making them the champion along with IBM Research, the team from the corporation. In 2009, the research team held by Professor Shou- De Lin, Professor Hsuan- Tien and Professor Chih- Jen Lin won the third place in the group of the long- term analysis with their precise business data from the selling of mobile phones, which could predict the consumers’ behavior pattern. Furthermore in 2010, the NTU team led by the same professors made a progress by winning two champions in both groups of the general and the students with their “Feature Engineering and Classifier Ensemble” to efficiently analyze the specific data, provided by the hosting organization, to get to predict students’ learning in their academic performances. Once again in 2011, Professor Shou- De Lin, Professor Hsuan- Tien and Professor Chih- Jen Lin led their NTU research team and won two world champions of the titles, “predicting users’ comments to the genres of music” and “automatically distinguishing users’ favorites and the unread”, in the recommendation system for music.

As for the 2012 ADM KDD Cup, the titles of the competition this year was provided by Tencent, the Chinese corporation. The first title of the competition is “predicting the favorite recommendations in the service of social networking, and the second is “the CTR (click- through rate) of the ads in search engines”. To be specifically noted, the second title covers the most data ever, the number of 1.5 billion data, in the history of the ACM KDD Cup in hopes that it can be achieved to conclude the precise prediction out of the database to tell whether users click on a certain ads, showing on the page of the search engine. There is potential business value lies beneath CTR (click- through rate), and thus it has always been the top issue on the fields of knowledge discovery and data mining. Based on the successful experiences and history, this school year Professor Shou- De Lin and Professor Hsuan- Tien Lin lectured in their course, the machinery theory and practice, with the full support from College of Electrical Engineering and Computer Science, Department of Computer Science and Integrated Engineering and Graduate Institute of Networking and Multimedia in order to provide the NTU students more related training so that they will be able to have a research team to compete in the ACM KDD Cup. During the course, the students were assigned to assemble to different groups to analyze the data and to constitute the pattern figures by using their creativity in diversity. With the weekly presentation, the students exchange ideas with their experiment results that they inspired more creative thoughts to make the presumption more complete. When it was approaching the end of ACM KDD Cup competition, each team integrated the pattern figures in order to make their prediction of the click- through rate more precise with a better standard. ACM KDD Cup brought a fierce competition among each research team, and it was not until the last day of the competition that the NTU research team could make sure that they did leave the other competitors behind. By doing the great efforts, the team members finally won the world champion and made it an unbreakable record: the fact that the NTU research team has been the prize winner over the past five years in the history of ACM KDD Cup competition history.

Chinese version