Applying an Unsupervised Machine Learning Approach to Detect Dietary Habits of Breast Cancer Patients in Bangladesh

Abstract

Purpose: The study aims to conduct a cross-sectional study to know the association between food habits/lifestyle and breast cancer using an unsupervised machine approach.  Methods: The dataset was collected from the hospitals of eight divisional cities in Bangladesh using a semi-structured questionnaire during January 2019. Descriptive statistical tools and unsupervised machine learning approach- Principal Component Analysis (PCA) were used to analyze the data. Results: Among 73 breast cancer patients out of 384 cancer patients from eight divisions in Bangladesh, 87.67% were housewife, 78.08% and 79.45% of breast cancer patients had no family history of cancer and no other disease before cancer respectively. The highest breast cancer patients were observed in Sylhet division followed by Dhaka and Khulna divisions. It is noted that overall left breast cancer patients are more than right breast. We found that betel-nut, beverages and beef/mutton etc. are high communities, indicates that these food habits are highly associated with breast cancer. Moreover, around 71.23% of the patients can’t bear the cost of treatment. Conclusion: Taken together, as prevention is better than cure, we should avoid unhealthy and junk foods.  In addition, many health education programs in different areas are urgently needed to improve the dietary habits for sustainable livelihood. Government and non-government organizations should take necessary steps to circulate the healthy lifestyle program to improve public health sector.