python sklearn 分类算法简单调用（借鉴）

 scikit-learn已经包含在Anaconda中。也可以在官方下载源码包进行安装。本文代码里封装了如下机器学习算法，我们修改数据加载函数，即可一键测试：

数据为近红外测试猕猴桃软硬和时间差异的数据，可以作为分类软硬以及前后时间差的分类。

[python] view
plain copy

 





# coding=gbk  

''''' 

Created on 2016年6月4日 

 

@author: bryan 

'''  

   

import time    

from sklearn import metrics    

import pickle as pickle    

import pandas as pd  

  

    

# Multinomial Naive Bayes Classifier    

def naive_bayes_classifier(train_x, train_y):    

    from sklearn.naive_bayes import MultinomialNB    

    model = MultinomialNB(alpha=0.01)    

    model.fit(train_x, train_y)    

    return model    

    

    

# KNN Classifier    

def knn_classifier(train_x, train_y):    

    from sklearn.neighbors import KNeighborsClassifier    

    model = KNeighborsClassifier()    

    model.fit(train_x, train_y)    

    return model    

    

    

# Logistic Regression Classifier    

def logistic_regression_classifier(train_x, train_y):    

    from sklearn.linear_model import LogisticRegression    

    model = LogisticRegression(penalty='l2')    

    model.fit(train_x, train_y)    

    return model    

    

    

# Random Forest Classifier    

def random_forest_classifier(train_x, train_y):    

    from sklearn.ensemble import RandomForestClassifier    

    model = RandomForestClassifier(n_estimators=8)    

    model.fit(train_x, train_y)    

    return model    

    

    

# Decision Tree Classifier    

def decision_tree_classifier(train_x, train_y):    

    from sklearn import tree    

    model = tree.DecisionTreeClassifier()    

    model.fit(train_x, train_y)    

    return model    

    

    

# GBDT(Gradient Boosting Decision Tree) Classifier    

def gradient_boosting_classifier(train_x, train_y):    

    from sklearn.ensemble import GradientBoostingClassifier    

    model = GradientBoostingClassifier(n_estimators=200)    

    model.fit(train_x, train_y)    

    return model    

    

    

# SVM Classifier    

def svm_classifier(train_x, train_y):    

    from sklearn.svm import SVC    

    model = SVC(kernel='rbf', probability=True)    

    model.fit(train_x, train_y)    

    return model    

    

# SVM Classifier using cross validation    

def svm_cross_validation(train_x, train_y):    

    from sklearn.grid_search import GridSearchCV    

    from sklearn.svm import SVC    

    model = SVC(kernel='rbf', probability=True)    

    param_grid = {'C': [1e-3, 1e-2, 1e-1, 1, 10, 100, 1000], 'gamma': [0.001, 0.0001]}    

    grid_search = GridSearchCV(model, param_grid, n_jobs = 1, verbose=1)    

    grid_search.fit(train_x, train_y)    

    best_parameters = grid_search.best_estimator_.get_params()    

    for para, val in list(best_parameters.items()):    

        print(para, val)    

    model = SVC(kernel='rbf', C=best_parameters['C'], gamma=best_parameters['gamma'], probability=True)    

    model.fit(train_x, train_y)    

    return model    

    

def read_data(data_file):    

    data = pd.read_csv(data_file)  

    train = data[:int(len(data)*0.9)]  

    test = data[int(len(data)*0.9):]  

    train_y = train.label  

    train_x = train.drop('label', axis=1)  

    test_y = test.label  

    test_x = test.drop('label', axis=1)  

    return train_x, train_y, test_x, test_y  

        

if __name__ == '__main__':
   

        datafilename = 'softunion20_21.csv'

    

    data_file = "L:\\Python\\output\\"+datafilename    

    thresh = 0.5    

    model_save_file = 1    

    model_save = {}    

     

    test_classifiers = ['NB', 'KNN', 'LR', 'RF', 'DT', 'SVM','SVMCV', 'GBDT']    

    classifiers = {'NB':naive_bayes_classifier,     

                  'KNN':knn_classifier,    

                   'LR':logistic_regression_classifier,    

                   'RF':random_forest_classifier,    

                   'DT':decision_tree_classifier,    

                  'SVM':svm_classifier,    

                'SVMCV':svm_cross_validation,    

                 'GBDT':gradient_boosting_classifier    

    }    

        

    print('reading training and testing data...')    

    train_x, train_y, test_x, test_y = read_data(data_file)    

        

    for classifier in test_classifiers:    

        print('******************* %s ********************' % classifier)    

        start_time = time.time()    

        model = classifiers[classifier](train_x, train_y)    

        print('training took %fs!' % (time.time() - start_time))    

        predict = model.predict(test_x)

        if model_save_file != None:    

            model_save[classifier] = model    

        precision = metrics.precision_score(test_y, predict)    

        recall = metrics.recall_score(test_y, predict)    

        print('precision: %.2f%%, recall: %.2f%%' % (100 * precision, 100 * recall))    

        accuracy = metrics.accuracy_score(test_y, predict)    

        print('accuracy: %.2f%%' % (100 * accuracy))

    import numpy as np

    model = classifiers['LR'](train_x, train_y)

    predict = model.predict(test_x)

    print "LR :"

    print "Predict：",test_x,predict.T

     

    

    if model_save_file != None:    

        pickle.dump(model_save, open(model_save_file, 'wb'))    

测试结果如下：

reading training and testing data...

******************* NB ********************

training took 0.004986s!

precision: 78.08%, recall: 71.25%

accuracy: 74.17%

******************* KNN ********************

training took 0.017545s!

precision: 97.56%, recall: 100.00%

accuracy: 98.68%

******************* LR ********************

training took 0.061161s!

precision: 89.16%, recall: 92.50%

accuracy: 90.07%

******************* RF ********************

training took 0.040111s!

precision: 96.39%, recall: 100.00%

accuracy: 98.01%

******************* DT ********************

training took 0.004513s!

precision: 96.20%, recall: 95.00%

accuracy: 95.36%

******************* SVM ********************

training took 0.242145s!

precision: 97.53%, recall: 98.75%

accuracy: 98.01%

******************* SVMCV ********************

Fitting 3 folds for each of 14 candidates, totalling 42 fits

[Parallel(n_jobs=1)]: Done  42 out of  42 | elapsed:    6.8s finished

probability True

verbose False

coef0 0.0

degree 3

tol 0.001

shrinking True

cache_size 200

gamma 0.001

max_iter -1

C 1000

decision_function_shape None

random_state None

class_weight None

kernel rbf

training took 7.434668s!

precision: 98.75%, recall: 98.75%

accuracy: 98.68%

******************* GBDT ********************

training took 0.521916s!

precision: 97.56%, recall: 100.00%

accuracy: 98.68%

附上近红外测试数据集