techno-ballu / data_exploration

A powerful data science console in R / Shiny

Data Exploration prototype

A powerful data science console in R / Shiny.

Demo @ https://bolaka.shinyapps.io/data_exploration

=========================================================================================

Roadmap

1. Loading data (implemented).

2. Generate meta data for all the features/variables and identify the target/s (implemented).

3. Visualize statistics about rows and columns and take actions if required i.e. drop sparse rows and columns (implemented).

4. Univariate analysis - histograms for continuous and bar plots for categorical features. Also interactively remove outliers from charts just by clicking on them! (implemented).

5. Important features - This step is just conjecture and needs more thought but the idea is to be able to identify useful / important features by fitting a simple model and looking at say the top ten influencers impacting the target (if continuous). Not sure how this can be done for a categorical target (roadmap).

6. Bi-variate/Multivariate analysis - With knowlegde about the important influencers from the previous step, plot scatter charts, eat maps, box plots etc... to better understand relationships between features / variables (in progress...).

7. Collinearity and Multicollinearity - Eliminate highly correlated pairs / groups of features (not yet implemented).

8. Missing value treatment (not yet implemented).

9. Outlier treatment (not yet implemented).

10. Feature transformations (not yet implemented).

11. Feature engineering (not yet implemented).

12. Predictive modeling (roadmap).