Özet:

Purpose: This article provides a) how the Global Adult Tobacco Research (KYTA) data can be used to reveal valuable information about the use of tobacco behaviors and b)the performance of the classification algorithms applied to the KYTA data.Method: The three well-known classification methods of K -the closest neighboring algorithm, the C4.5 algorithm and the multi-layer detector of the KYTA participants, based on the socio-demographic characteristics (age group, gender, settlement, education level and work status), the proper classification of the state of smoking (pre-defined classes: smoking and non-smoking) was assessed. The first analysis was carried out on the data of KYTA Turkey. The model that produced the best performance for Turkey was also applied to the data from six different European countries: Greece, Kazakhstan, Poland, Romania, Russia and Ukraine.The findings: All tree algorithms give more accurate results in identifying those who do not smoke. The correct classification rate of the C4.5 algorithm is the highest for Turkey. Comparative analysis for countries shows that the C4.5 algorithm can accurately classify the status of participants smoking in Ukraine above 80%, but for Greece this rate remains below 70%.Result: This article shows that the information provided by KYTA, such as demographic data, can help calculate the possibility of a individual to smoke in the future.

Anahtar Kelimeler: