机器学习——数据清洗，特征选择

数据清洗的方法：
设置阈值去掉异常值
随机森林预测去掉点的数值加进去

onehot编码（不适用于决策树和随机森林）：
先将一个属性分成几个类别
然后再将样本的数据变成矩阵01，1表示其所在类别
会导致特征数增多

数据清洗代码实现

import numpy as np
import pandas as pd
from fuzzywuzzy import fuzz
from fuzzywuzzy import process


def enum_row(row):
    print row['state']


def find_state_code(row):
    if row['state'] != 0:
        print process.extractOne(row['state'], states, score_cutoff=80)


def capital(str):
    return str.capitalize()


def correct_state(row):
    if row['state'] != 0:
        state = process.extractOne(row['state'], states, score_cutoff=80)
        if state:
            state_name = state[0]
            return ' '.join(map(capital, state_name.split(' ')))
    return row['state']


def fill_state_code(row):
    if row['state'] != 0:
        state = process.extractOne(row['state'], states, score_cutoff=80)
        if state:
            state_name = state[0]
            return state_to_code[state_name]
    return ''


if __name__ == "__main__":
    pd.set_option('display.width', 200)
    data = pd.read_excel('sales.xlsx', sheetname='sheet1', header=0)
    print 'data.head() = \n', data.head()
    print 'data.tail() = \n', data.tail()
    print 'data.dtypes = \n', data.dtypes
    print 'data.columns = \n', data.columns
    for c in data.columns:
        print c,
    print
    data['total'] = data['Jan'] + data['Feb'] + data['Mar']
    print data.head()
    print data['Jan'].sum()
    print data['Jan'].min()
    print data['Jan'].max()
    print data['Jan'].mean()

    print '============='
    # 添加一行
    s1 = data[['Jan', 'Feb', 'Mar', 'total']].sum()
    print s1
    s2 = pd.DataFrame(data=s1)
    print s2
    print s2.T
    print s2.T.reindex(columns=data.columns)
    # 即：
    s = pd.DataFrame(data=data[['Jan', 'Feb', 'Mar', 'total']].sum()).T
    s = s.reindex(columns=data.columns, fill_value=0)
    print s
    data = data.append(s, ignore_index=True)
    data = data.rename(index={15:'Total'})
    print data.tail()

    # apply的使用
    print '==============apply的使用=========='
    data.apply(enum_row, axis=1)

    state_to_code = {"VERMONT": "VT", "GEORGIA": "GA", "IOWA": "IA", "Armed Forces Pacific": "AP", "GUAM": "GU",
                     "KANSAS": "KS", "FLORIDA": "FL", "AMERICAN SAMOA": "AS", "NORTH CAROLINA": "NC", "HAWAII": "HI",
                     "NEW YORK": "NY", "CALIFORNIA": "CA", "ALABAMA": "AL", "IDAHO": "ID",
                     "FEDERATED STATES OF MICRONESIA": "FM",
                     "Armed Forces Americas": "AA", "DELAWARE": "DE", "ALASKA": "AK", "ILLINOIS": "IL",
                     "Armed Forces Africa": "AE", "SOUTH DAKOTA": "SD", "CONNECTICUT": "CT", "MONTANA": "MT",
                     "MASSACHUSETTS": "MA",
                     "PUERTO RICO": "PR", "Armed Forces Canada": "AE", "NEW HAMPSHIRE": "NH", "MARYLAND": "MD",
                     "NEW MEXICO": "NM",
                     "MISSISSIPPI": "MS", "TENNESSEE": "TN", "PALAU": "PW", "COLORADO": "CO",
                     "Armed Forces Middle East": "AE",
                     "NEW JERSEY": "NJ", "UTAH": "UT", "MICHIGAN": "MI", "WEST VIRGINIA": "WV", "WASHINGTON": "WA",
                     "MINNESOTA": "MN", "OREGON": "OR", "VIRGINIA": "VA", "VIRGIN ISLANDS": "VI",
                     "MARSHALL ISLANDS": "MH",
                     "WYOMING": "WY", "OHIO": "OH", "SOUTH CAROLINA": "SC", "INDIANA": "IN", "NEVADA": "NV",
                     "LOUISIANA": "LA",
                     "NORTHERN MARIANA ISLANDS": "MP", "NEBRASKA": "NE", "ARIZONA": "AZ", "WISCONSIN": "WI",
                     "NORTH DAKOTA": "ND",
                     "Armed Forces Europe": "AE", "PENNSYLVANIA": "PA", "OKLAHOMA": "OK", "KENTUCKY": "KY",
                     "RHODE ISLAND": "RI",
                     "DISTRICT OF COLUMBIA": "DC", "ARKANSAS": "AR", "MISSOURI": "MO", "TEXAS": "TX", "MAINE": "ME"}
    states = state_to_code.keys()
    print fuzz.ratio('Python Package', 'PythonPackage')
    print process.extract('Mississippi', states)
    print process.extract('Mississipi', states, limit=1)
    print process.extractOne('Mississipi', states)
    data.apply(find_state_code, axis=1)

    print 'Before Correct State:\n', data['state']
    data['state'] = data.apply(correct_state, axis=1)
    print 'After Correct State:\n', data['state']
    data.insert(5, 'State Code', np.nan)
    data['State Code'] = data.apply(fill_state_code, axis=1)
    print data

    # group by
    print '==============group by================'
    print data.groupby('State Code')
    print 'All Columns:\n'
    print data.groupby('State Code').sum()
    print 'Short Columns:\n'
    print data[['State Code', 'Jan', 'Feb', 'Mar', 'total']].groupby('State Code').sum()

    # 写入文件
    data.to_excel('sales_result.xls', sheet_name='Sheet1', index=False)

主成分分析PCA代码实现：

import pandas as pd
import numpy as np
from sklearn.decomposition import PCA
from sklearn.linear_model import LogisticRegressionCV
from sklearn import metrics
from sklearn.model_selection import train_test_split
import matplotlib as mpl
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import PolynomialFeatures


def extend(a, b):
    return 1.05*a-0.05*b, 1.05*b-0.05*a


if __name__ == '__main__':
    pd.set_option('display.width', 200)
    data = pd.read_csv('iris.data', header=None)
    columns = ['sepal_length', 'sepal_width', 'petal_length', 'petal_width', 'type']
    data.rename(columns=dict(zip(np.arange(5), columns)), inplace=True)
    data['type'] = pd.Categorical(data['type']).codes
    print data.head(5)
    x = data.loc[:, columns[:-1]]
    y = data['type']

    pca = PCA(n_components=2, whiten=True, random_state=0)
    x = pca.fit_transform(x)
    print '各方向方差：', pca.explained_variance_
    print '方差所占比例：', pca.explained_variance_ratio_
    print x[:5]
    cm_light = mpl.colors.ListedColormap(['#77E0A0', '#FF8080', '#A0A0FF'])
    cm_dark = mpl.colors.ListedColormap(['g', 'r', 'b'])
    mpl.rcParams['font.sans-serif'] = u'SimHei'
    mpl.rcParams['axes.unicode_minus'] = False
    plt.figure(facecolor='w')
    plt.scatter(x[:, 0], x[:, 1], s=30, c=y, marker='o', cmap=cm_dark)
    plt.grid(b=True, ls=':')
    plt.xlabel(u'组份1', fontsize=14)
    plt.ylabel(u'组份2', fontsize=14)
    plt.title(u'鸢尾花数据PCA降维', fontsize=18)
    # plt.savefig('1.png')
    plt.show()

    x, x_test, y, y_test = train_test_split(x, y, train_size=0.7)
    model = Pipeline([
        ('poly', PolynomialFeatures(degree=2, include_bias=True)),
        ('lr', LogisticRegressionCV(Cs=np.logspace(-3, 4, 8), cv=5, fit_intercept=False))
    ])
    model.fit(x, y)
    print '最优参数：', model.get_params('lr')['lr'].C_
    y_hat = model.predict(x)
    print '训练集精确度：', metrics.accuracy_score(y, y_hat)
    y_test_hat = model.predict(x_test)
    print '测试集精确度：', metrics.accuracy_score(y_test, y_test_hat)

    N, M = 500, 500     # 横纵各采样多少个值
    x1_min, x1_max = extend(x[:, 0].min(), x[:, 0].max())   # 第0列的范围
    x2_min, x2_max = extend(x[:, 1].min(), x[:, 1].max())   # 第1列的范围
    t1 = np.linspace(x1_min, x1_max, N)
    t2 = np.linspace(x2_min, x2_max, M)
    x1, x2 = np.meshgrid(t1, t2)                    # 生成网格采样点
    x_show = np.stack((x1.flat, x2.flat), axis=1)   # 测试点
    y_hat = model.predict(x_show)  # 预测值
    y_hat = y_hat.reshape(x1.shape)  # 使之与输入的形状相同
    plt.figure(facecolor='w')
    plt.pcolormesh(x1, x2, y_hat, cmap=cm_light)  # 预测值的显示
    plt.scatter(x[:, 0], x[:, 1], s=30, c=y, edgecolors='k', cmap=cm_dark)  # 样本的显示
    plt.xlabel(u'组份1', fontsize=14)
    plt.ylabel(u'组份2', fontsize=14)
    plt.xlim(x1_min, x1_max)
    plt.ylim(x2_min, x2_max)
    plt.grid(b=True, ls=':')
    patchs = [mpatches.Patch(color='#77E0A0', label='Iris-setosa'),
              mpatches.Patch(color='#FF8080', label='Iris-versicolor'),
              mpatches.Patch(color='#A0A0FF', label='Iris-virginica')]
    plt.legend(handles=patchs, fancybox=True, framealpha=0.8, loc='lower right')
    plt.title(u'鸢尾花Logistic回归分类效果', fontsize=17)
    # plt.savefig('2.png')
    plt.show()