Advances in Data Mining. Applications and Theoretical by Rainer Schmidt, Heike Weiss, Georg Fuellen (auth.), Petra

By Rainer Schmidt, Heike Weiss, Georg Fuellen (auth.), Petra Perner (eds.)

This booklet constitutes the refereed complaints of the twelfth commercial convention on information Mining, ICDM 2012, held in Berlin, Germany in July 2012. The 22 revised complete papers provided have been rigorously reviewed and chosen from ninety seven submissions. The papers are prepared in topical sections on facts mining in drugs and biology; facts mining for power undefined; information mining in site visitors and logistic; information mining in telecommunication; information mining in engineering; conception in info mining; thought in information mining: clustering; idea in facts mining: organization rule mining and choice rule mining.

Show description

Read or Download Advances in Data Mining. Applications and Theoretical Aspects: 12th Industrial Conference, ICDM 2012, Berlin, Germany, July 13-20, 2012. Proceedings PDF

Best industrial books

Advances in Blood Substitutes: Industrial Opportunities and Medical Challenges

Every one bankruptcy of this quantity is a contribution from a professional within the box, selected through the editors to give a contribution to the 1997 "Current matters in Blood alternative examine and improvement" direction given in San Diego, March 17-19. The members have been chosen as a result of their services in parts which the editors think to be severe to the development of the sector, and which mirror job in "hot" parts of proper examine.

Industrial parsing of software manuals

The duty of language engineering is to boost the expertise for development desktops that may practice helpful linguistic projects resembling computer assisted translation, textual content retrieval, message class and record summarisation. Such structures usually require using a parser which could extract particular different types of grammatical facts from pre-defined sessions of enter textual content.

Industrial Ethernet - How to Plan, Install, and Maintain TCP/IP Ethernet Networks: The Basic Reference Guide for Automation and Process Control Engineers

This booklet is a handy set up, troubleshooting, and reference instrument on one of many preferred themes in automation and strategy regulate. it is going to assist you comprehend Ethernet and TCP/IP terminology and supply very important information regarding commercial protocols and criteria. you'll speedy achieve an outstanding seize of Ethernet fundamentals, the restrictions of the commercial setting, and the really expert necessities of desktop regulate.

Additional resources for Advances in Data Mining. Applications and Theoretical Aspects: 12th Industrial Conference, ICDM 2012, Berlin, Germany, July 13-20, 2012. Proceedings

Example text

The average specificity of each classifier remains high, as also the average sensitivity is greater than 50%, but still highly varies from set to set. Table 6 provides evaluation results using the Third feature set with six attributes. The average results decreased, comparing to data in Table 5, showing that attributes 9 and 10 - the number of cigarettes smoked per day and the flatulence, should not be excluded from feature set. Figure 2 depicts the average values of classification accuracy, sensitivity and specificity for all classifiers separately for each feature set.

The training records of the second subset were removed from the initial dataset, leaving 769 of 819 records for testing - 762 negative examples and 7 positive examples. Table 10 shows the obtained results. Table 10. 1 (see Table 5), however the specificity and classification accuracy decreased by 10% in general. Analysing obtained results it can be stated that classifiers show high true positive rate - sensitivity, resulting in correct diagnosis for the most of patients with gastric cancer, which is good.

Constipations ≤ 12 Cig. per day Yes No Neg. Age ≤ 57 Neg. Yes No >57 Pos. >12 Pos. ≤5 >5 Neg. Constipations Yes No Age Neg. ≤ 62 >62 Pos. Pos. Neg. Fig. 4. Classification tree 36 A. Kirshners, S. Parshutin, and M. Leja not have other tumours AND Have no heartburn AND 1st-degree relatives have gastric cancer THEN result is Negative. This points out that classifiers are data specific and may contain rules that are normal for a training set, but do not concur with an opinion of a medical expert. That is one of the reasons for system to be a decision support not a decision system.

Download PDF sample

Rated 4.84 of 5 – based on 33 votes