Skip to main content

Sweep Analysis

Created on August 29|Last edited on September 3

Analysis of all the sweeps that include data with end station id features


024681012141618202224Index12.81313.213.413.6validation-rmse.min

Parameter importance with respect to
validation-rmse.min

Config parameter
Importance
Correlation
Loading...
6e-27e-28e-21e-12e-13e-14e-15e-16e-17e-18e-1learning_rate4681012141618202224262830max_depth9e-31e-22e-23e-24e-25e-26e-27e-28e-29e-21e-1reg_alpha4e-35e-31e-22e-23e-24e-25e-21e-12e-13e-1reg_lambda12.712.812.913.013.113.213.313.413.513.613.7validation-rmse.min
Sweep: XGBoost Sweep 2
25


Predictions vs true for retrained model with best params



Final results for the model with best hyperparameters

This model shows slight improvement in validation RMSE comparing to the baseline model, and it was only trained for 500 steps (vs 1000 for the baseline model)

Run: scarlet-microwave-144
1

File<(table)>