---
title: "Discussion on \"Episode 16:\nPhase 5: LightGBM Didn't Win the Way I Expected. It Won Differently.\" | Hashnode"
description: "Discussion on \"Episode 16:\nPhase 5: LightGBM Didn't Win the Way I Expected. It Won Differently.\". Act 1 — The Setup\nPhase 4 ended with LR + SMOTE as the model to beat. Churn recall of 0.80. ROC-AUC of 0.839. A logistic regression — the simplest model in the toolkit — holding the top spot after fou"
url: "https://hashnode.com/posts/episode-16-phase-5-lightgbm-didn-t-win-the-way-i-expected-it-won-differently/69ee04a63d6a492cddf9d04f"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Episode 16: Phase 5: LightGBM Didn't Win the Way I Expected. It Won Differently.](https://adityand.hashnode.dev/episode-16-phase-5-lightgbm-didn-t-win-the-way-i-expected-it-won-differently?utm_source=hashnode&utm_medium=feed)

Act 1 — The Setup Phase 4 ended with LR + SMOTE as the model to beat. Churn recall of 0.80. ROC-AUC of 0.839. A logistic regression — the simplest model in the toolkit — holding the top spot after fou

Apr 26 · 3 min read · adityand.hashnode.dev  
Read the article

## Written by

[Aditya Dandia](https://hashnode.com/@adityand)

@adityand · Bengaluru, India

  
[View profile](https://hashnode.com/@adityand)

ML engineer in progress. Building from scratch.

A learner building ML algorithms from scratch to actually understand what's happening under the hood. Writing about the math, the mistakes, and the process and not just the results.

## Responses

Markdown supported · ⌘↵ to post

No responses yet.