mlearn.py

classic machine learning algorithm implementation with clean and readable code and sklearn-like api

Install

cd mlearn.py && pip install -e . # [dev]
# or
cd mlearn.py && pip install .
# or
pip install git+https://github.com/gseismic/mlearn.py.git

Feature 特性

• sklearn-like api & clean and readable code
• Both Chinese(中文) and English version comments
• linear_model.LinearRegression 线性回归
• tree.DecisionTreeClassifier 决策树分类
• tree.DecisionTreeRegressor 决策树回归
• metrics.r2_score/accurency_score/.. 评价指标
• ensemble.random_forest.RandomForestClassifier-naive-version 随机森林分类
• ensemble.random_forest.RandomForestRegressor-naive-version 随机森林回归
• ensemble.gbdt.GBDTClassifier-naive-version GBDT分类
• ensemble.gbdt.GBDTRegressor-naive-version GBDT回归
• K nearest neighbors 最近邻
• naive_bayes 朴素贝叶斯算法
• max_entropy 最大熵算法
• svm 支持向量机
• mlp 多层感知机

Examples 示例

# examples/demo_linear_regression.py
import numpy as np
from mlearn import linear_model
import matplotlib.pyplot as plt

# get dataset | 获取数据集
np.random.seed(0)
X = np.random.rand(100, 1)  # shape: (100, 1)
y = 2 + 3 * X + np.random.randn(100, 1) * 0.1  # shape: (100, 1)

# create and train model | 创建并训练模型
model = linear_model.LinearRegression()
model.fit(X, y)  # X shape: (100, 1), y shape: (100, 1)
# print | 打印结果
print("Coefficients:", model.coef_)  # shape: (1,)
print("Intercept:", model.intercept_)  # shape: ()
print("**R² Score:", model.score(X, y))  # X shape: (100, 1), y shape: (100, 1)

# predict | 预测
X_test = np.array([[0.5]])  # shape: (1, 1)
print("Prediction for X=0.5:", model.predict(X_test))  # 输出 shape: (1,)

# plot | 绘图
plt.scatter(X, y, color='blue', label='Data points')
plt.plot(X, model.predict(X), color='red', label='Fitted line')
plt.xlabel('X')
plt.ylabel('y')
plt.legend()
plt.show()

output:

Reference 参考资料

• sklearn
• [机器学习-周志华]
• [统计学习方法-李航]

ChangeLog

• [@2024-08-30] project created
• [@2024-09-09] add ensemble.random_forest & test tree_regressor & tree_classifier
• [@2024-09-10] add ensemble.gbdt & test gbdt_classifier & gbdt_regressor & metric.log_loss