CODEWITHSUNDEEP
statisticsrandom-forestqgisoptuna
Mcpaws
Mcpaws

Reputation: 1

Whitebox Workflow Random Forest Regression Fit Hyperparameter Tuning

I've been using Whitebox Workflow Random Forest Regression fit for my undergraduate thesis, a plugin from QGIS. The plugin creates a model from the input data such as Raster files, number of trees, Leafs, and splits then trains itself to achieve the target data. However, I can't find any documentation on how to perform Hyperparameter tuning, a process to achieve the optimal number of trees, Leafs, and splits. I was wondering if anyone have encountered this kind of situation with this plugin/library?

I tried using Optuna Library, which I found through different tutorials on YouTube. However, it produces a different output. For example, the Optuna Library produces R-Coefficient of 0.78872 using 300 Trees, 4 Min sample leaf, and 14 split. Meanwhile, Whitebox Workflow Random Forest Regression fit (the plugin) produces R-Coefficient of 0.773 even though the number of Trees, Min sample leaf and split are the same. The difference between the two outputs may be small but I need more than 400 trials, and this discrepancy can significantly impact my study.

Upvotes: 0

Views: 20

Answers (0)

Related Questions