ECSE PSM 2011
+
Advanced Data Mining and Pattern Recognition Workshop

February 23-25, 2011
Mekrijärvi Research Station, Ilomantsi, Finland

pictures


Program

Wednesday, February 23 - Workshop

Thursday, February 24 - Workshop

Friday, February 25 - ECSE Progress Reports


Organizers:

Chair: Pasi Fränti, ECSE graduate school
Local arrangements: Amit Roy and Qinpei Zhao

contact information:
email: first.last@uef.fi
phone:

+358 (0)44 3582021 (Qinpei Zhao )
+358 (0)46 5843162 (Amit Roy)

Transportation

23.2. Science Park (11:30) -> Joensuu Railway station (11:45) -> Mekrijärvi
25.2. Mekrijävri -> Joensuu Railway station (IC12 at 18:17) -> Science Park


Abstracts

Mining Road Traffic Accidents, Prof. Tommi Kärkkäinen
Absract: We present results from a research study on applying data mining methods into analysis of traffic accidents in Finnish roads. The data set, collected by the Finnish Road Administration between 2004-2008 consists of more than 83000 accidents of which 1203 are fatal. The results show that understandable and (partly) unexpected patterns from data were obtained.

Variable Selection and Sparse Linear Models, Dr. Tapio Pahikkala
Absract: Feature selection is one of the fundamental tools in machine learning and data mining. The purpose of this talk is to give a overview of the concepts of feature selection in machine learning as well as to present some of our recent contributions in the field. The presentation focuses especially on greedy selection algorithms and their efficient implementations.

Web Mining for Mobile Location-aware Applications, Prof. Pasi Fränti
Absract: Relevance of information depends on the content of the data, time, location, and user behavior. We study how these aspects are used in a media-sharing service called MOPSI that combines traditional search engine with location-based service directory with media sharing. We show how the location is currently utilized and discuss what kind of data and web mining problems there are unsolved.

Document image segmentation: A new knowledge-based approach, Mohammad Rezaei
Absract: New segmentation method is proposed for document images of both simple and complex layouts, different qualities and resolutions. Specialized segmentation-based binarization is applied for layout analysis, after which connected components are extracted. The method is tested for documents consisting of Persian and English language, and mixture of both.

Integration of Learning Supportive Applications to Development of e-Portfolio Construction Process, Harri Hämäläinen
Absract:While students are planning their study path and selecting which modules and courses they will include in their studies, they do not necessarily know what are their strengths and what are the skills needed. We present a few of tools that can be used to support this process by providing additional information about the skills.

Can low vision user benefit from eye tracking?, Tersia Gowases
Absract:While students are planning their study path and selecting which modules and courses they will include in their studies, they do not necessarily know what are their strengths and what are the skills needed. We present a few of tools that can be used to support this process by providing additional information about the skills.

Solving Learning-to-Rank Problems, Antti Airola
Absract: The problem of learning-to-rank has attracted considerable attention in the machine learning community over the past decade. The research has been driven by applications in areas such as information retrieval, recommender systems and decision making. The presentation will discuss how ranking can be modelled as a pairwise preference learning task, and how the resulting problem can be solved with kernel methods such as RankSVM and RankRLS.

Developments and Industrial Applications of MLP Neural Networks, Paavo Nieminen
Absract: Multilayer Perceptron (MLP), although dating back to the last century, is still a viable alternative for some data mining tasks. Some recent developments and applications of MLP training methodology (especially cost function formulations), related to the author's doctoral studies will be presented in the talk.

Spectral Imaging and Analysis, Paras Pant
Absract: Spectral imaging and its analysis is getting popularity in developing the industrial application. Spectral image is a function of the wavelength. The principal component analysis can be applied for the analysis of the spectral image and other method too. Moreover, spectral imaging in the pulp and paper industry for dirt particle counting and classification are discussed.

Text Similarity and Clustering in MOPSI, Qinpei Zhao
Absract: Computing the similarity between short segments of text (words) is increasingly important in applications such as text mining, web information retrieval and question answering. Without context information, short texts similarity is more difficult than document analysis. Clustering analysis can be applied on the short texts with the similarity measures. Other clusterings in MOPSI project are also discussed.

Mining Human Activities by GPS Trajectories, Minjie Chen
Absract: Recently, mobile phone has been used more than just communication. A number of trajectories/routes are collected of users' position information uses a mobile phone with built-in GPS receiver. The focus of this work is to analyze the human behavior based on the collected GPS data. The collected routes are divided into several segments with different properties (transportation modes), such as stationary, walking, biking, running, or car driving.

Introduction, applicability and challenges of Sentiment Analysis from text, Myriam Munezero
Absract: The analysis of sentiments is generating a vast amount of research, as it is now more freely available in online social communication environments. However, with the large amount of texts, it becomes impossible for human beings to read them tentatively. In particular sentimental content can provide valuable information to various interest groups and unfortunately also poses various information extraction challenges.