XGBoost-Tensorflow-Wide-and-deep-comparison

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

XGBoost-Tensorflow-Wide-and-deep-comparison

(★ 19)

Comparison how Tensorflow Wide and Deep model performs vs. XGBoost implementation

XGBoost-Tensorflow-Wide-and-deep-comparison 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Comparison of different models

1. **TFLearn based wide and deep model** (code copied from https://www.tensorflow.org/tutorials/wide_and_deep and adapted to jupyter notebook format)

2. **TFLearn wide and deep re-implemented in Keras**

3. **XGBoost** based implementation

## Results

* TFLearn wide and deep model - similarly to results from TF tutorial has accuracy is 84.5%

* XGBoost - best accuracy is 86.1%

* TFLearn wide and deep re-implemented in Keras - best accuracy is 85.1%

## Notes

* If you want to test it by yourself, download the data using TFLearn notebook

* Keras version skipped tf.contrib.layers.crossed_column features. Implementing them could further improve accuracy

## Summary

* This type of "tabular" based dataset is still easiest to implement using XGBoost

* Keras version was implemented using one-hot encoddings and separately embeddings. Surprisingly the one-hot encoding version achieved better accuracy

* Probably when given more data, with more options in categorical columns ("workclass", "education", "marital_status" etc.) both TFLearn wide and deep and Keras embedding versions would perform better than XGBoost version.

**Check notebooks for details**

XGBoost-Tensorflow-Wide-and-deep-comparison

XGBoost-Tensorflow-Wide-and-deep-comparison 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

XGBoost-Tensorflow-Wide-and-deep-comparison

XGBoost-Tensorflow-Wide-and-deep-comparison 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README