20.500.12556/DKUM-61333 Ugotavljanje uspešnosti inteligentnih sistemov pri ločevanju rakov prostate in debelega črevesa Assessment of intelligent systems used for distinguishing between prostate and colon cancer Rak na debelem črevesu in rak prostate sta dve izmed najpogostejših rakavih obolenj moške populacije tako v Sloveniji kot tudi po svetu. Pri ženskah se prav tako pojavlja rak na črevesu kot ena izmed najpogostejših oblik raka. Rak je zelo kompleksna vrsta bolezni, za katero točnih vzrokov ne poznamo. Številne raziskave potrjujejo, da je rak posledica tako kombinacije medsebojnega delovanja dednih dejavnikov kot tudi najrazličnejših vplivov okolja. Cilj magistrske naloge je bil teoretično predstaviti področje raziskovanja ter v nadaljevanju pridobljeno znanje uporabiti na praktičnem primeru. Teoretični del opisuje tako metode podatkovnega rudarjenja, s poudarkom na metodi podpornih vektorjev, kot tudi odločitvena drevesa, s poudarkom na odločitvenem drevesu J48. Predstavili smo tudi proteine, ki regulirajo gensko ekspresijo, tako imenovane transkripcijske faktorje, ki pri nastanku rakavih obolenj igrajo pomembno vlogo. Ker se naš praktičen primer nanaša na obolenje raka na črevesu in raka prostate, smo na kratko predstavili tudi statistične podatke, dejavnike tveganja, simptome ter možnosti zdravljenja obeh omenjenih vrst raka. V empiričnem delu predstavljamo praktičen primer, ki smo ga izvedli s programskim orodjem Weka in z uporabo podatkovne baze podatkov AP_Colon_Prostate – za to smo se odločili, ker obolenje raka na črevesu in raka prostate spadata med najpogostejša rakava obolenja današnje populacije. V programskem okolju Weka smo ugotavljali, za kako natančna in zanesljiva se izkažeta klasifikator metode podpornih vektorjev (angl. Support vector machine – SVM ali angl. Sequntial minimal optimization – SMO) in klasifikator odločitvenega drevesa J48 pri ugotavljanju genskih ekspresij. S pomočjo programskega okolja Weka smo predstavili tudi najpogostejše gene iz naše podatkovne baze, ki se pojavljajo pri obolenju raka črevesja in raka prostate. Colorectal cancer and prostate cancer are the two most common cancers of the male population in Slovenia as well as around the world. Colorectal cancer is also one of the most common cancers with women. Cancer is a very complex type of disease the exact cause of which is unknown. Numerous researches confirm that cancer is a result of a combination of interaction among genetic factors, as well as various environmental influences. The objective of the master´s thesis was to theoretically present the field of research and the further use the obtained knowledge in a practical case. The theoretical part describes the methods of the data mining with the emphasis on the support vector machine as well as the decision trees with an emphasis on the decision tree J48. We also presented the proteins regulating gene expression, so called transcription factors, which play an important role in cancer formation. Since our practical case is related to the colorectal and prostate cancer, we also shortly presented the statistical data, risk factors, symptoms, and possibilities for treatment of the both mentioned types of cancer. In the empirical part, we present a practical case, which was performed with Weka software by using the AP_Colon_Prostate database. The latter was chosen because colorectal and prostate cancer are the two most common cancers of the present-day population. In the Weka software we were determining how accurate and reliable the classificator of support vector machine (SVM) or Sequntial minimal optimization (SMO), and classificator of the decision tree J48 in establishing gene expressions are. By using Weka software, we also presented the most common genes from our database, which occur in colorectal and prostate cancer. metoda podpornih vektorjev strojno učenje klasifikator J48 Weka rak na črevesu rak prostate support vector machine machine learning classificator J48 Weka colorectal cancer prostate cancer. true true false [S. Pucko] Slovenski jezik Angleški jezik Magistrsko delo/naloga 2016-08-01 11:04:57 2016-12-02 08:59:23 2022-06-26 03:06:07 2019-07-11 12:17:12 2016 0 Maribor 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 616.3:006:004.8(043.2) 2289316 URN:SI:UM:DK:HPVK7G4F MAG_Pucko_Simon_2016.pdf MAG_Pucko_Simon_2016.pdf 1 DC1D029F831AC628389000BB8FFFC151 2b2a099b9b09e88752cb025713899b255c4c2ff2163976f2fa9088597e90dec1 dec46a0a-7c0d-11eb-bb7a-00155d0001ca https://dk.um.si/Dokument.php?lang=slv&id=108417 Fakulteta za zdravstvene vede 0 0 0