World-of-AI
World-of-AI copied to clipboard
adithya-s-k
[UPDATE] Increasing accuracy by feature engineering
Machine learning, heart.csv
| Field | Description |
|---|---|
| About | Applying pca and keeping important features only |
| Github | Aaditikapre |
| [email protected] | |
| Label | Gssoc'23 |
Define You
- [ *] GSSOC Participant
- [*] Contributor
Is your feature request related to a problem? Please describe. Some features in the dataset create noise and are redundant hence should be eliminated. Some features like trestbps and restecg explain very less variance in the data and are not correlated with the target variable. Features like oldpeak and slope are highly correlated and can be combined with pca.
Describe the solution you'd like...
I checked the correlation between different features and the target as well as explained variance. I can do some data processing to lower the dimension of the dataset and make it better at predicting the target. I can increase accuracy of Decision Tree Classifier and Random forest classifier to 98% using this. It also increases accuracy of KNN and SVM more than the current accuracy
Describe alternatives you've considered?
I have considered clustering and ICA as well, they did not work.
Approach to be followed (optional):
Additional context
