बायोमेडिकल डेटा माइनिंग का अंतर्राष्ट्रीय जर्नल
खुला एक्सेस

अमूर्त

Potential Risk Factors of Blood Cancer and Association of Significant Factors: A Survey in Bangladesh Using Data Mining Techniques

Farzana Tasnim*, Tasniya Ahmed, Kawsar Ahmed, Fazlul Karim Patwary, Karam Newaz

In recent years, cancer is one of the main causes of death worldwide and in Bangladesh. Therefore, the research interest is associated with blood cancer patients as only a few blood cancer studies are available in Bangladesh. Initially, we collect 340 patient data (blood cancer and non-blood cancer) from BSMMU hospital. Then we implement data mining techniques to rank 30 factors (conducted with questionnaires) that are frequently related to blood cancer. In this research, we use Data mining approaches such as classification, chi-square (χ 2 ) test, P-value, and association rule mining. This study finds the predictive role of 15 factors among 30 input factors, among them muscle pull, inability to control the bladder, unusual bleeding, fever/raised temperature, weakness usually in legs identified as most potent predictors (p-value<0.001). Subsequently, an association among these significant elements is anticipated using the most popular rule mining algorithm such as Apriori, Predictive Apriori, and Tertius. The experimental result shows that weakness usually in legs=yes, fever/raised temperature=yes, and not being able to control bladder=no is a common rule for blood cancer=yes. Again, unusual bleeding=yes, rapidly becoming more ill=yes and, muscle pull=yes rule is likely to have a significant association. Besides, a frequent pattern of having a fair skin tone with fast breathing and weakness in the legs might be a threat.