Feature selection for breast cancer diagnosis via visualization / Izzah Khairina Muhadi
Early diagnosis of breast cancer is important as it is one of the reason that causes death among women and men. Most diagnostic systems suffer the feature multiplicity problem. Some of these features are redundant and irrelevant to be used for breast cancer classification. Feature selection techni...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2019
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/110719/1/110719.pdf https://ir.uitm.edu.my/id/eprint/110719/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Early diagnosis of breast cancer is important as it is one of the reason that causes death among women and men. Most diagnostic systems suffer the feature multiplicity problem. Some of these features are redundant and irrelevant to be used for breast cancer classification. Feature selection techniques have been used to find the most important features that are suitable to classify the type of breast cancer either malignant or benign. From the previous research work, it shows that visualization can contribute to feature selection. This project explores the feature selection through visualization as opposed to chi square filter feature selection technique. The visualization technique used for this project is a radial chart using d3.js library. Each feature of the data is the axis in radial chart and it will plot based on the value of the features. The radial chart used two different colours to differentiate between malignant and benign and the features are selected based on the features that are least overlap when the data being plotted. The new features that have been chosen from the visualization technique are used to classify the breast cancer type using K-Nearest Neighbour (KNN) classifier. To evaluate effectiveness of the proposed feature selection technique, the results are compared with the chi square filter feature selection technique using accuracy, specificity and sensitivity measurement. The results show that feature selection via visualization produce higher accuracy, specificity and sensitivity after being compared with chi square technique. |
|---|