# Super Massive LGBM Grid Search

**URL:** <https://forum.numer.ai/t/super-massive-lgbm-grid-search/6463>\
**Category:** Tournament\
**Created:** [June 15, 2023, 1:48pm UTC](https://forum.numer.ai/t/super-massive-lgbm-grid-search/6463 "2023-06-15T13:48:33Z")\
**Posts on this page:** 1\
**Showing post:** 1

<div class="post-metadata">

**Author:** ![maul](https://avatars.discourse-cdn.com/v4/letter/m/9de0a6/32.png) [@maul](https://forum.numer.ai/u/maul)\
**Post date:** [June 15, 2023, 1:48pm UTC](https://forum.numer.ai/t/super-massive-lgbm-grid-search/6463/1 "2023-06-15T13:48:33Z")

</div>

# Introduction

We at Numerai have spent the last couple of weeks conducting a new grid search on the Sunshine V4.1 data set. We have built hundreds of models with different hyperparameters on the V4.1 dataset focussed on target\_cyrus\_v4\_20 which we believe is the best single target at present for our hedge fund strategy.

We are sharing the best of these grid searched results in terms of correlation and correlation Sharpe to enable users to benefit from the grid search and either use these results directly in their models or allow them to do more targeted searches of their own around the sweet spots.

# Experimental Setup

We did the grid search using the following parameters:-

Features = all the features in the V4.1 dataset

Target = target\_cyrus\_v4\_20

Algorithm = Scikit-learn API LGBMRegressor

Hyperparameter ranges for the better results are shared below:-

n\_estimators = 30k - 60k

learning\_rate = 0.001

max\_depth = 5, 6, 7

num\_leaves = 2\*\*max\_depth - 1

colsample\_bytree = 0.1

# Results

The first plot below contains the correlation and the second the correlation Sharpe computed using the out of sample predictions from era 578 to 1059 inclusive. The training period was from era1 to era 574 inclusive.

 ![](https://canada1.discourse-cdn.com/flex009/uploads/numerai/original/2X/7/7701075d801ab6492986c72657e4a71e3afd5173.png)

Table 1 below contains the 20 best correlation results.

Table1

 ![](https://canada1.discourse-cdn.com/flex009/uploads/numerai/original/2X/f/fc6c8fb1de192da64ecee82aa10be10e7e51098f.png)

Table 2 below contains the 20 best correlation Sharpe results.

Table 2

 ![](https://canada1.discourse-cdn.com/flex009/uploads/numerai/original/2X/9/95a483fbdd394079f467f806999f9d78e87cb743.png)

\*\*The plot below shows the cumulative correlation of

1. The sunshine recommended param model with learning\_rate = 0.001, n\_estimators = 20K and  
max\_depth = 6.

2. The best correlation model from the above table.

3. The 2 best correlation Sharpe models from the above table.

 ![](https://canada1.discourse-cdn.com/flex009/uploads/numerai/original/2X/3/355455ad2e9c5e408dd4d6e9a9061d122e50e81f.png)

# Alternate hyperparameters for less compute

The above parameters require 6 hours to compute for a tree of max\_depth = 6 , a learning rate of 0.001 and n\_estimators = 100k on a 24 core processor. This may be a heavy compute burden for users.

We show below parameters that work with lower compute using a learning rate of 0.01, n\_estimators = 20k and columnsample\_bytree = 0.1. This reduces the compute time to less than 2 hours.

 ![](https://canada1.discourse-cdn.com/flex009/uploads/numerai/original/2X/e/e44d09ab251e00fd880f56b6b3cebbe33391afeb.png)

# Conclusion

The above results are very competitive in correlation Sharpe space while slightly worse in terms of pure correlation but with significant compute saving.

---

_[View the full topic](https://forum.numer.ai/t/super-massive-lgbm-grid-search/6463)._