决策树分类与回归模型的实现和可视化

Overview

DecisionTree

决策树分类与回归模型,以及可视化

ID3

ID3决策树是最朴素的决策树分类器:

  • 无剪枝
  • 只支持离散属性
  • 采用信息增益准则

data.py中,我们记录了一个小的西瓜数据集,用于离散属性的二分类任务。我们可以像下面这样训练一个ID3决策树分类器:

from ID3 import ID3Classifier
from data import load_watermelon2
import numpy as np

X, y = load_watermelon2(return_X_y=True) # 函数参数仿照sklearn.datasets
model = ID3Classifier()
model.fit(X, y)
pred = model.predict(X)
print(np.mean(pred == y))

输出1.0,说明我们生成的决策树是正确的。

C4.5

C4.5决策树分类器对ID3进行了改进:

  • 用信息增益率的启发式方法来选择划分特征;
  • 能够处理离散型和连续型的属性类型,即将连续型的属性进行离散化处理;
  • 剪枝;
  • 能够处理具有缺失属性值的训练数据;

我们实现了前两点,以及第三点中的预剪枝功能(超参数)

data.py中还有一个连续离散特征混合的西瓜数据集,我们用它来测试C4.5决策树的效果:

from C4_5 import C4_5Classifier
from data import load_watermelon3
import numpy as np

X, y = load_watermelon3(return_X_y=True) # 函数参数仿照sklearn.datasets
model = C4_5Classifier()
model.fit(X, y)
pred = model.predict(X)
print(np.mean(pred == y))

输出1.0,说明我们生成的决策树正确.

CART

分类

CART(Classification and Regression Tree)是C4.5决策树的扩展,支持分类和回归。CART分类树算法使用基尼系数选择特征,此外对于离散特征,CART决策树在每个节点二分划分,缓解了过拟合。

这里我们用sklearn中的鸢尾花数据集测试:

from CART import CARTClassifier
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score

X, y = load_iris(return_X_y=True)
train_X, test_X, train_y, test_y = train_test_split(X, y, train_size=0.7)
model = CARTClassifier()
model.fit(train_X, train_y)
pred = model.predict(test_X)
print(accuracy_score(test_y, pred))

准确率95.55%。

回归

CARTRegressor类实现了决策树回归,以sklearn的波士顿数据集为例:

from CART import CARTRegressor
from sklearn.datasets import load_boston
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error

X, y = load_boston(return_X_y=True)
train_X, test_X, train_y, test_y = train_test_split(X, y, train_size=0.7)
model = CARTRegressor()
model.fit(train_X, train_y)
pred = model.predict(test_X)
print(mean_squared_error(test_y, pred))

输出26.352171052631576,sklearn决策树回归的Baseline是22.46,性能近似,说明我们的实现正确。

决策树绘制

分类树

利用python3的graphviz第三方库和Graphviz(需要安装),我们可以将决策树可视化:

from plot import tree_plot
from CART import CARTClassifier
from sklearn.datasets import load_iris

X, y = load_iris(return_X_y=True)
model = CARTClassifier()
model.fit(X, y)
tree_plot(model)

运行,文件夹中生成tree.png

iris_tree

如果提供了特征的名词和标签的名称,决策树会更明显:

from plot import tree_plot
from CART import CARTClassifier
from sklearn.datasets import load_iris

iris = load_iris()
model = CARTClassifier()
model.fit(iris.data, iris.target)
tree_plot(model,
          filename="tree2",
          feature_names=iris.feature_names,
          target_names=iris.target_names)

iris_tree2

绘制西瓜数据集2对应的ID3决策树:

from plot import tree_plot
from ID3 import ID3Classifier
from data import load_watermelon2

watermelon = load_watermelon2()
model = ID3Classifier()
model.fit(watermelon.data, watermelon.target)
tree_plot(
    model,
    filename="tree",
    font="SimHei",
    feature_names=watermelon.feature_names,
    target_names=watermelon.target_names,
)

这里要自定义字体,否则无法显示中文:

watermelon

回归树

用同样的方法,我们可以进行回归树的绘制:

from plot import tree_plot
from ID3 import ID3Classifier
from sklearn.datasets import load_boston

boston = load_boston()
model = ID3Classifier(max_depth=5)
model.fit(boston.data, boston.target)
tree_plot(
    model,
    feature_names=boston.feature_names,
)

由于生成的回归树很大,我们限制最大深度再绘制:

regression

调参

CART和C4.5都是有超参数的,我们让它们作为sklearn.base.BaseEstimator的派生类,借助sklearn的GridSearchCV,就可以实现调参:

from plot import tree_plot
from CART import CARTClassifier
from sklearn.datasets import load_wine
from sklearn.model_selection import train_test_split, GridSearchCV

wine = load_wine()
train_X, test_X, train_y, test_y = train_test_split(
    wine.data,
    wine.target,
    train_size=0.7,
)
model = CARTClassifier()
grid_param = {
    'max_depth': [2, 4, 6, 8, 10],
    'min_samples_leaf': [1, 3, 5, 7],
}

search = GridSearchCV(model, grid_param, n_jobs=4, verbose=5)
search.fit(train_X, train_y)
best_model = search.best_estimator_
print(search.best_params_, search.best_estimator_.score(test_X, test_y))
tree_plot(
    best_model,
    feature_names=wine.feature_names,
    target_names=wine.target_names,
)

输出最优参数和最优模型在测试集上的表现:

{'max_depth': 4, 'min_samples_leaf': 3} 0.8518518518518519

绘制对应的决策树:

wine

剪枝

在ID3和CART回归中加入了REP剪枝,C4.5则支持了PEP剪枝。

对IRIS数据集训练后的决策树进行PEP剪枝:

iris = load_iris()
model = C4_5Classifier()
X, y = iris.data, iris.target
train_X, test_X, train_y, test_y = train_test_split(X, y, train_size=0.7)
model.fit(train_X, train_y)
print(model.score(test_X, test_y))
tree_plot(model,
          filename="src/pre_prune",
          feature_names=iris.feature_names,
          target_names=iris.target_names)
model.pep_pruning()
print(model.score(test_X, test_y))
tree_plot(model,
          filename="src/post_prune",
          feature_names=iris.feature_names,
          target_names=iris.target_names,
)

剪枝前后的准确率分别为97.78%,100%,即泛化性能的提升:

prepre

Owner
Welt Xing
Undergraduate in AI school, Nanjing University. Main interest(for now): Machine learning and deep learning.
Welt Xing
Steganography is the art of hiding the fact that communication is taking place, by hiding information in other information.

Steganography is the art of hiding the fact that communication is taking place, by hiding information in other information.

Priyansh Sharma 7 Nov 09, 2022
Forecast dynamically at scale with this unique package. pip install scalecast

🌄 Scalecast: Dynamic Forecasting at Scale About This package uses a scaleable forecasting approach in Python with common scikit-learn and statsmodels

Michael Keith 158 Jan 03, 2023
dirty_cat is a Python module for machine-learning on dirty categorical variables.

dirty_cat dirty_cat is a Python module for machine-learning on dirty categorical variables.

637 Dec 29, 2022
A single Python file with some tools for visualizing machine learning in the terminal.

Machine Learning Visualization Tools A single Python file with some tools for visualizing machine learning in the terminal. This demo is composed of t

Bram Wasti 35 Dec 29, 2022
Bayesian Additive Regression Trees For Python

BartPy Introduction BartPy is a pure python implementation of the Bayesian additive regressions trees model of Chipman et al [1]. Reasons to use BART

187 Dec 16, 2022
Dragonfly is an open source python library for scalable Bayesian optimisation.

Dragonfly is an open source python library for scalable Bayesian optimisation. Bayesian optimisation is used for optimising black-box functions whose

744 Jan 02, 2023
Dual Adaptive Sampling for Machine Learning Interatomic potential.

DAS Dual Adaptive Sampling for Machine Learning Interatomic potential. How to cite If you use this code in your research, please cite this using: Hong

6 Jul 06, 2022
A simple application that calculates the probability distribution of a normal distribution

probability-density-function General info An application that calculates the probability density and cumulative distribution of a normal distribution

1 Oct 25, 2022
A collection of video resources for machine learning

Machine Learning Videos This is a collection of recorded talks at machine learning conferences, workshops, seminars, summer schools, and miscellaneous

Dustin Tran 1.5k Dec 29, 2022
Decentralized deep learning in PyTorch. Built to train models on thousands of volunteers across the world.

Hivemind: decentralized deep learning in PyTorch Hivemind is a PyTorch library to train large neural networks across the Internet. Its intended usage

1.3k Jan 08, 2023
All-in-one web-based development environment for machine learning

All-in-one web-based development environment for machine learning Getting Started • Features & Screenshots • Support • Report a Bug • FAQ • Known Issu

3 Feb 03, 2021
A Lightweight Hyperparameter Optimization Tool 🚀

The mle-hyperopt package provides a simple and intuitive API for hyperparameter optimization of your Machine Learning Experiment (MLE) pipeline.

Robert Lange 137 Dec 02, 2022
A fast, scalable, high performance Gradient Boosting on Decision Trees library, used for ranking, classification, regression and other machine learning tasks for Python, R, Java, C++. Supports computation on CPU and GPU.

Website | Documentation | Tutorials | Installation | Release Notes CatBoost is a machine learning method based on gradient boosting over decision tree

CatBoost 6.9k Jan 05, 2023
Laporan Proyek Machine Learning - Azhar Rizki Zulma

Laporan Proyek Machine Learning - Azhar Rizki Zulma Project Overview Domain proyek yang dipilih dalam proyek machine learning ini adalah mengenai hibu

Azhar Rizki Zulma 6 Mar 12, 2022
The code from the Machine Learning Bookcamp book and a free course based on the book

The code from the Machine Learning Bookcamp book and a free course based on the book

Alexey Grigorev 5.5k Jan 09, 2023
A collection of neat and practical data science and machine learning projects

Data Science A collection of neat and practical data science and machine learning projects Explore the docs » Report Bug · Request Feature Table of Co

Will Fong 2 Dec 10, 2021
A statistical library designed to fill the void in Python's time series analysis capabilities, including the equivalent of R's auto.arima function.

pmdarima Pmdarima (originally pyramid-arima, for the anagram of 'py' + 'arima') is a statistical library designed to fill the void in Python's time se

alkaline-ml 1.3k Dec 22, 2022
2021 Machine Learning Security Evasion Competition

2021 Machine Learning Security Evasion Competition This repository contains code samples for the 2021 Machine Learning Security Evasion Competition. P

Fabrício Ceschin 8 May 01, 2022
BigDL: Distributed Deep Learning Framework for Apache Spark

BigDL: Distributed Deep Learning on Apache Spark What is BigDL? BigDL is a distributed deep learning library for Apache Spark; with BigDL, users can w

4.1k Jan 09, 2023
A complete guide to start and improve in machine learning (ML)

A complete guide to start and improve in machine learning (ML), artificial intelligence (AI) in 2021 without ANY background in the field and stay up-to-date with the latest news and state-of-the-art

Louis-François Bouchard 3.3k Jan 04, 2023