第6.1.2章 机器学习之八字强弱分析

我对韦千里的《呱呱集》中108个命例，进行标注，尝试使用knn算法，计算命例的用神。结果先用最简单的“克泄耗”和“生助”两种，
计算规则使用李洪城老师的《具体断四柱导读》

1 knn算法

import numpy as np
import pandas as pd
from sklearn import metrics
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import MinMaxScaler

def get_datas():
    file_name = r'/Users/dzm/Documents/bazi/201903291933.csv'
    file = pd.read_csv(file_name,usecols=['生','助','克','泄','耗','用神'],encoding='GBK')
    ar = np.array(file)
    return ar[:,:5],ar[:,5]
    
def predict_1(x_data,y_data):
    scaler = MinMaxScaler()
    X = scaler.fit_transform(x_data)
    x_train,x_test,y_train,y_test = train_test_split(X,y_data,test_size=0.2)
    knn = KNeighborsClassifier(n_neighbors=3)
    knn.fit(x_train,y_train)
    print(knn)
    expected = y_test
    predicted = knn.predict(x_test)
    print(metrics.classification_report(expected, predicted))
    label = list(set(y_data))
    print(metrics.confusion_matrix(expected, predicted, labels=label))
   
if __name__ == '__main__':
    x_data, y_data = get_datas()
    predict_1(x_data,y_data)

上面的脚本，有三段输出，怎么解读呢？
第一段输出的，解释参考sklearn 翻译笔记：KNeighborsClassifier

KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',
           metric_params=None, n_jobs=None, n_neighbors=3, p=2,
           weights='uniform')

下面的输出，须要了解召回率、准确率、调和平均几个概念，参考
准确率与召回率、如何解释召回率与准确率？，召回率指的样本中想匹配的种类占该类总数的比率，召回率又叫查全率，这个就容易理解一些。而准确率时对抽取的样本而言。下图描述了1和2两种分类的查全率和查准率。

precision    recall  f1-score   support

           1       0.83      0.45      0.59        11
           2       0.62      0.91      0.74        11

   micro avg       0.68      0.68      0.68        22
   macro avg       0.73      0.68      0.66        22
weighted avg       0.73      0.68      0.66        22

下面的输出，须要了解基于混淆矩阵的评价指标，查看ML01 机器学习后利用混淆矩阵Confusion matrix 进行结果分析，就容易理解了。通过混淆矩阵，可以看出，我们的结果并不理想

[[ 5  6]
 [ 1 10]]

2 决策树
决策树需要理解信息熵，参考白话信息熵

import numpy as np
import pandas as pd
from sklearn import metrics
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import MinMaxScaler
import matplotlib.pylab as plt
from sklearn import tree

def predict_2(x_data,y_data):
    scaler = MinMaxScaler()
    X = scaler.fit_transform(x_data)
    x_train, x_test, y_train, y_test = train_test_split(X, y_data, test_size=0.1)
    clf = tree.DecisionTreeClassifier()
    clf.fit(x_train,y_train)
    print(clf)
    expected = y_test
    predicted = clf.predict(x_test)
    print(metrics.classification_report(expected, predicted))
    label = list(set(y_data))
    print(metrics.confusion_matrix(expected, predicted, labels=label))

输出的结果为

DecisionTreeClassifier(class_weight=None, criterion='gini', max_depth=None,
            max_features=None, max_leaf_nodes=None,
            min_impurity_decrease=0.0, min_impurity_split=None,
            min_samples_leaf=1, min_samples_split=2,
            min_weight_fraction_leaf=0.0, presort=False, random_state=None,
            splitter='best')
              precision    recall  f1-score   support

           1       0.38      1.00      0.55         3
           2       1.00      0.38      0.55         8

   micro avg       0.55      0.55      0.55        11
   macro avg       0.69      0.69      0.55        11
weighted avg       0.83      0.55      0.55        11

[[3 0]
 [5 3]]