Final project for the Data and Text Mining course, from my master in Computer Science. KNIME workflow, Report PDF.
Cervical Cancer is one of the most treatable cancers when diagnosed at early stages. Yet, it killed 311000 women in 2018 and it’s still the fourth cause of death from cancers in women, the second most common in developing areas, mainly because of the economic cost and the difficulties in implementing effectives screening programmes. Data Mining provides robust tools to verify the known causal relations and assess risk factors from medical datasets. Classification models and clustering can then exploit this scenario to help identifying groups of population at higer risk to improve planning of screening programmes.
This year in the US, there will be an estimated 13800 new cases of Cervical cancer. 4290 will be deadly. Worldwide, this type of cancer causes makes up 8% of the total cancer cases and deaths. It’s the fourth (second in developing countries) cause of death from cancer in women. A lot of progress has been made, and nowadays it is one of the most curable: when diagnosed and treated at the earliest stages, the five-year survival rate can reach 95%.
In this work, we investigate the possibility of classifying high-risk patients from demographic and other medical data, but excluding the results of the other related exams.
Known causal relations: HPV causal relation with the cervical cancer has been documented be- yond reasonable doubt. Additionally, the Human papilloma virus infection is necessary for the development of CIN, the abnormal growth of cells on the surface of the cervix, indicating a potentially precancerous transformation of cells of the cervix. Risk factors include smoking, early age at the first sexual intercourse and early pregnancies.