笔记 |Macos安装pycaret

pycaret

pycaret是机器学习的懒人包。与其他开源机器学习库相比，pycaret是一个备用的低代码库，可用于仅用很少几个单词替换数百行代码。它本质上就是组装了多个机器学习库和框架，例如scikit-learn，XGBoost，Microsoft LightGBM，spaCy等。

比如几年前，为了这样对比sklearn的几个estimator，你需要以下的代码：

# Regression problem
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import numpy as np
from sklearn import model_selection
from sklearn.metrics import make_scorer, mean_squared_error 
from sklearn.svm import SVR, LinearSVR
from sklearn.tree import DecisionTreeRegressor
from sklearn.linear_model import LinearRegression,Ridge,Lasso,ElasticNet,BayesianRidge,SGDRegressor
from sklearn.gaussian_process import GaussianProcessRegressor       
from sklearn.neighbors import KNeighborsRegressor  
from sklearn.tree import DecisionTreeRegressor
from sklearn.neural_network import MLPRegressor
from sklearn.ensemble import GradientBoostingRegressor,RandomForestRegressor,ExtraTreesRegressor
from sklearn.kernel_ridge import KernelRidge


models=[] 
models.append(('DecisionTree', DecisionTreeRegressor()))
models.append(('Ridge', Ridge()))
models.append(('Lasso', Lasso()))
models.append(('EN', ElasticNet(alpha=0.001,max_iter=10000)))
models.append(('BayesianRidge',BayesianRidge()))
models.append(('SVM',SVR()))
models.append(('KNeighbors',KNeighborsRegressor()))
models.append(('NN',MLPRegressor()))
models.append(('GBoosting',GradientBoostingRegressor()))
models.append(('RF',RandomForestRegressor()))
models.append(('ExtraTrees',ExtraTreesRegressor()))       
models.append(('SGD',SGDRegressor(max_iter=1000,tol=1e-3)))
models.append(('Kernel_Ridge',KernelRidge(alpha=0.6, kernel='polynomial', degree=2, coef0=2.5)))
models.append(('LR_SVR',LinearSVR()))
models.append(('LR',LinearRegression()))






def compare_scores_mae(models, X, y):
    cv_means = []
    cv_std = []
    cv_resutls= []
    names=[]
    for name,model in models:
        kfold = model_selection.KFold(n_splits=10)
        cv_results = model_selection.cross_val_score(model, X, y, cv=kfold, scoring='neg_mean_absolute_error',n_jobs=10)
        cv_means.append(cv_results.mean())
        cv_std.append(cv_results.std())
        names.append(name)
        msg = "%s: %f (%f)" % (name, cv_results.mean(), cv_results.std())       
        print(msg)
    cv_res=pd.DataFrame({"CrossValMeans":cv_means,"CrossValerrors": cv_std,"Algorithm":names})
    g = sns.barplot("CrossValMeans","Algorithm",data = cv_res, palette="Set3",orient = "h",**{'xerr':cv_std})
    g.set_xlabel("negative MAE")
    g = g.set_title("Cross validation scores")
    return cv_res

对，我在认识pycaret之前就是这么干的。

pycaret就是这样把这些代码封装成了一个函数：compare_models()

安装

官方给出了通过pip

#installing for the first time
pip install pycaret
#if you have installed beta version in past, run the below code to upgrade
pip install --upgrade pycaret
#Run the below code in your notebook to check the installed version
from pycaret.utils import version
version()

或者conda安装的方法，

#create a conda environment
conda create --name yourenvname python=3.6
#activate environment
conda activate yourenvname
#install pycaret
pip install pycaret
#create notebook kernel connected with the conda environment
python -m ipykernel install --user --name yourenvname --display-name "display-name-here"

如果在colab或者kaggle的instance的话使用!pip就好。然鹅如果安装过python或者R包的你知道，事情可能并没有那么简单，Macos在安装llvmlite和LightGBM的时候各种error让人猝不及防，导致安装失败。大概花了一个小时在Macos上安装pycaret （悄悄告诉你kaggle的instance上安装没有任何毛病）。

llvmlite

pip安装总是出现python setup_tools的相关错误，过程中发现llvmite这个包需要cmake。用brew安装了cmake，结果还是不行。最后在github一个角落发现可以使用easy_install的命令轻松解决其不能在python3.8上安装的问题，果断试了试（自己用的conda环境python3.5），问题解决。原理不得而知，pip不行，easy_install就可以。

brew install cmake
easy_install llvmlite

LightGBM

LihgtGBM是树模型中模型能力最优异的模型之一，作为pycaret包含的模型之一，安装pycaret的过程中也需要安装LightGBM。LightGBM在window上的安装很简单（微软自家开发），直接使用python自带的pip安装工具安装即可。在Mac上用pip安装会遇到错误。因此需要安装C版本LightGBM。

pip uninstall lightgbm
git clone --recursive https://github.com/Microsoft/LightGBM ; cd LightGBM
export CXX=g++-8 CC=gcc-8
mkdir build ; cd build
cmake ..
make -j4

如果发现自己没有gcc-8的话，使用brew安装gcc-8，记忆中cmake也是需要用到到。

brew install gcc@8

最后的建议

conda和pip安装最好不要混搭。

不要升级pip，升级过后你会有一种需要重新装python的赶脚。

升级之后使用pip如下

File "F:\anaconda\envs\emotion\lib\site-packages\pkg_resources\__init__.py", line 2331, in resolve
    module = __import__(self.module_name, fromlist=['__name__'], level=0)
  File "F:\anaconda\envs\emotion\lib\site-packages\pip\_internal\__init__.py", line 42, in 
    from pip._internal import cmdoptions
  File "F:\anaconda\envs\emotion\lib\site-packages\pip\_internal\cmdoptions.py", line 16, in 
    from pip._internal.index import (
ImportError: cannot import name 'FormatControl'

附赠一份降级教程：

https://pypi.org/project/pip/19.1.1/#files

手动下载第二个文件并解压，在其目录下运行

python setup.py install

End