This repo shows simple proof of concept of simple auto machine learning app. It can do regression and classification tasks so far (no object detection etc.).
I have found it useful for quick data exploration. Linearity of data, most important variables, ...
Features included:
- Filling missing values - not with ML approach yet.
- Transforming categorical features.
- Selecting features for model. Based on F test and Chi-squared test.
- Model selection. Choosing among linear model, random forest model and SVM model.
- Visualising the most important features.
- Visualising regression line (decision boundary). If number of features > 1 (2) then PCA ist used.
- Testing
Features not included:
- Outlier detection. (Planned).
- Cann't export model. (Planned).
- Creating synthetic features.
- No neural networks for object detection etc.