All Projects → DeqianBai → Your First Machine Learning Project End To End In Python

DeqianBai / Your First Machine Learning Project End To End In Python

这是一个完整的,端到端的机器学习项目,非常适合有一定基础后拿来练习,以提高对完整机器学习项目的认识

Projects that are alternatives of or similar to Your First Machine Learning Project End To End In Python

Siml
Machine Learning algorithms implemented from scratch
Stars: ✭ 319 (-1.24%)
Mutual labels:  jupyter-notebook
Notebooks Contrib
RAPIDS Community Notebooks
Stars: ✭ 321 (-0.62%)
Mutual labels:  jupyter-notebook
Probability
Probabilistic reasoning and statistical analysis in TensorFlow
Stars: ✭ 3,550 (+999.07%)
Mutual labels:  jupyter-notebook
Datos Covid19
En formato estándar
Stars: ✭ 316 (-2.17%)
Mutual labels:  jupyter-notebook
Bokeh Python Visualization
A Bokeh project developed for learning and teaching Bokeh interactive plotting!
Stars: ✭ 321 (-0.62%)
Mutual labels:  jupyter-notebook
Joypy
Joyplots in Python with matplotlib & pandas 📈
Stars: ✭ 322 (-0.31%)
Mutual labels:  jupyter-notebook
Self Correction Human Parsing
An out-of-box human parsing representation extractor.
Stars: ✭ 319 (-1.24%)
Mutual labels:  jupyter-notebook
Machine Learning For Trading
Code for Machine Learning for Algorithmic Trading, 2nd edition.
Stars: ✭ 4,979 (+1441.49%)
Mutual labels:  jupyter-notebook
Julia Dataframes Tutorial
A tutorial on Julia DataFrames package
Stars: ✭ 318 (-1.55%)
Mutual labels:  jupyter-notebook
Cs329s Ml Deployment Tutorial
Code and files to go along with CS329s machine learning model deployment tutorial.
Stars: ✭ 314 (-2.79%)
Mutual labels:  jupyter-notebook
Standalone Deeplearning
2019 KAIST 딥러닝 홀로서기 세미나용 저장소입니다.
Stars: ✭ 318 (-1.55%)
Mutual labels:  jupyter-notebook
Sars tutorial
Repository for the tutorial on Sequence-Aware Recommender Systems held at TheWebConf 2019 and ACM RecSys 2018
Stars: ✭ 320 (-0.93%)
Mutual labels:  jupyter-notebook
Carnd Traffic Sign Classifier Project
Classify Traffic Signs.
Stars: ✭ 323 (+0%)
Mutual labels:  jupyter-notebook
Bmc
Notes on Scientific Computing for Biomechanics and Motor Control
Stars: ✭ 319 (-1.24%)
Mutual labels:  jupyter-notebook
Adanet
Fast and flexible AutoML with learning guarantees.
Stars: ✭ 3,340 (+934.06%)
Mutual labels:  jupyter-notebook
Python
🐍 Python Programs
Stars: ✭ 320 (-0.93%)
Mutual labels:  jupyter-notebook
Ml prep
Machine Learning interview prep guide
Stars: ✭ 298 (-7.74%)
Mutual labels:  jupyter-notebook
Bdci2019 Sentiment Classification
CCF BDCI 2019 互联网新闻情感分析 复赛top1解决方案
Stars: ✭ 317 (-1.86%)
Mutual labels:  jupyter-notebook
Autoeq
Automatic headphone equalization from frequency responses
Stars: ✭ 5,989 (+1754.18%)
Mutual labels:  jupyter-notebook
Jupyter Themer
Apply custom CSS styling to your jupyter notebooks
Stars: ✭ 322 (-0.31%)
Mutual labels:  jupyter-notebook

Your-first-machine-learning-Project---End-to-End-in-Python

目标:

  • 使用提供的建筑能源数据开发一个模型,该模型可以预测建筑物的能源之星得分,
  • 然后解释结果以找到最能预测得分的变量。

介绍:

这是一个受监督的回归机器学习任务:给定一组包含目标(在本例中为分数)的数据,我们希望训练一个可以学习将特征(也称为解释变量)映射到目标的模型。

  • 受监督问题: 我们可以知道数据的特征和目标,我们的目标是训练可以学习两者之间映射关系的模型。
  • 回归问题: Energy Star Score是一个连续变量。

在训练中,我们希望模型能够学习特征和分数之间的关系,因此我们给出了特征和答案。然后,为了测试模型的学习效果,我们在一个从未见过答案的测试集上进行评估

收获

  • 数据清理,探索性数据分析,特征工程和选择等常见问题的解决办法
  • 随机搜索,网格搜索,交叉验证等方法寻找最优超参数
  • 可视化决策树
  • 对完整的机器学习项目流程建立一个宏观的了解

目录

1. 数据清理和格式化

1.1 加载并检查数据

1.2 数据类型和缺失值

1.2.1 将数据转换为正确的类型

1.3 处理缺失值

2. 探索性数据分析

2.1 单变量图

2.2 去除异常值

2.3 寻找关系

2.4 特征与目标之间的相关性

2.5 双变量图(Two-Variable Plots)

2.5.1 Pairs Plot

3. 特征工程和选择

3.1 特征工程

3.2 特征选择(去除共线特征)

3.3 划分训练集和测试集

3.4 建立Baseline

4. 基于性能指标比较几种机器学习模型

4.1 输入缺失值

4.2 特征缩放

4.3 - 需要评估的模型

5. 对最佳模型执行超参数调整,即优化模型

5.1 超参数

5.2 使用随机搜索和交叉验证进行超参数调整

6. 在测试集上评估最佳模型

7. 解释模型结果

7.1 特征重要性 - Feature Importances

7.2 使用特征重要性进行特征选择

7.3 本地可解释的与模型无关的解释- Locally Interpretable Model-agnostic Explanations (LIME)

7.4 检查单个决策树

8.得出结论&&记录发现

8.1 得出结论

8.2 记录发现

联系我

[email protected]

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].