EDA and Data Pre-Processing

Owned by Isaac Godfried
Last updated: Jan 20, 2021

We all know that pre-processing data is often a crucial step to having models perform well. Flow-Forecast therefore has built in capabilities to:

  • Scale the data

  • Create new features from existing columns.

  • Make features from the specific date-time stamp.

  • Impute missing values and check for improperly entered values.

  • Augment the data (TS-AUG transformations)

The preprocessed data is now saved automatically in a temp_df.csv file so you can view it and make sure it is correct.

Preliminary Data Visualizations

Simple plots

Lag plots

Seasonal plots

 

Statistical Tests

  • Granger Causality

  • Stationarity tests