160 likes | 278 Views
STULONG Discovery Challenges Feedback. Marie Tomečková EuroMISE – Cardio This work is supported by the project LN00B107 of the Ministry of Education of the Czech Republic. STULONG Challenges – Medical feedback STULONG = acronym LONG itudinal STU dy. Main aims of the study:
E N D
STULONG Discovery Challenges Feedback Marie Tomečková EuroMISE – Cardio This work is supported by the project LN00B107 of the Ministry of Education of the Czech Republic
STULONG Challenges – Medical feedbackSTULONG = acronymLONGitudinal STUdy Main aims of the study: To determine prevalence of the risk factors of atherosclerosis in middle-age men • To follow up the development of the risk factors • To asses the possibilities and the influence of the complex intervention on the incidence and values of the risk factors and on the cardiovascular mortality
Atherosclerosis: • a total complicated disease of all over organism • a dynamic process, it begins in childhood and adolescence and continues for the whole life • opinions on the origin and progress of the disease are developing • interaction and influence of genetic predisposition and exterior environment • the influence of so-called risk factors is still regarded • some so-called protective factors exist
STULONG - analysis • Statistical - descriptive statistics - logistic regression - survival analysis • Data mining - different methods - resulting in different conclusions
Mortality caused by atherosclerotic CVD depending on the number of RFA
Kaplan-Meier analysisin RG (at age of 38-44years)depending on the number of RFA
Kaplan-Meier analysisin RG (at age of 45-53years)depending on the number of RFA
Different approaches to solve the analytic questions • univaried and bivaried data analysis • association rules • trend analysis • analysis so called time windows • ROC analysis • Miner tool SDS, WEKA tool, STATISTICA tool • genetic approach • standard attribute-value data mining techniques • inductive logic programming technique
Different approaches to solve the analytic questions – continue: • fuzzy approximate dependencies • explicit relations - functional dependencies • the inductive logic programming technique • Rough Set Exploration System to solve both classification and descriptive tasks • approach to generate a mathematical algebraic model – discriminate function - Werner, Kalganova • the selection of = interesting = emerging patterns (strong emerging patterns
Challenge 2003 Some approaches to solve the analytic questions • Genetic approach – function based on the Area Under the ROC curve - Conclusions very good understanable Azé,J. - Lucas,N. - Sebag,M.: A New Medical Test for Atherosclerosis Detection GeNo • Fuzzy Approximate Dependencies – Fuzzy logic – Interesting relations – for discussion Berzal,F. - Cubero,J.C.- Sanchez,D. - Serrano,J.M. - Vila,M.A.: Finding Fuzzy Approximate Dependencies within STULONG Data
Challenge 2003 – cont.Some approaches to solve the analytic questions • Association rules – see later (Prague) Burian,J. - Rauch,J.: Analysis of Death Causes in the STULONG Data Set • Strongest Emerging Patterns – very interesting approach, results to discuss Cremilleux,B.- Soulet,A. - Rioult,F.: Mining the Strongest Emerging Patterns Characterizing Patients Affected by Diseases Due to Atherosclerosis • Rough Set Exploration System (RSES) – experimental tool, not yet implemented, without explications Hoa,N.S. - Son,N.H.: Analysis of STULONG Data by Rough Set Exploration System (RSES)
Challenge 2003 – cont.Some approaches to solve the analytic questions • SDS rules (Set Differs of Set) – some very interesting results – diferences among the groups in more than two variables – good conclusions Karban,T.: SDS-Rules and Classification on PKDD2003 Discovery Challenge • Trend analysis, analysis so called time windows- interesting approache, some conlusiones to discuss from medical point of view Novakova,L. - Klema,J. - Jakob,M.- Rawles,S. - Stepankova,O.: Trend Analysis and Risk Identification
Challenge 2003 – cont.Some approaches to solve the analytic questions • WEKA tool, ACE data tool – very good presentation with ilustrative explanations Van Assche,A. - Verbaeten,S. - Krzywania,D. - Struyf,J. - Blockeel,H.: Attribute-Value and First Order Data Mining within the STULONG Project • Discriminate function Werner,J.C. - Kalganova,T.: Risk Evaluation using Evolvable Discriminate Function