Question

这是我第一次尝试使用DBSCAN对从网页中提取的文本内容块进行离散（数据点的绑定宽度）和连续特征（计算的CSS和数据点的路径）进行聚类。

我有7个样本（在第一个数据集中），因此当我将DBSCAN min_samples设置为1时，此输出是我期望的：

估计的簇数：7
估计的噪声点数：0
同质性：1.000
完整度：1.000

然后我尝试绘制群集以使其可视化。对于图，我使用了sklearn example，并根据我的数据进行了调整。但是，结果图看起来不太正确。

看起来大多数群集的y轴值都相同（-0.408）。我认为，这取决于在此步骤中使用StandardScaler（）：

feature_stack = np.hstack([continuous_features, discrete_features])
"""[[-1.31614507  0.          1.          1.          1.          1.          0.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          1.          0.          0.          0.          0.          1.          0.          0.          0.          0.          1.          1.          0.          1.          0.          1.          0.          0.          0.          1.          0.          0.          0.          0.          1.          0.          1.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          1.          0.          0.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          1.          0.          0.          0.          0.          1.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          0.          1.          0.        ]
    [-0.66130166  0.          1.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          1.          0.          1.          0.          1.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.        ]"""
features = feature_stack.astype(np.float32)
"""[[-1.3161451   0.          1.          1.          1.          1.          0.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          1.          0.          0.          0.          0.          1.          0.          0.          0.          0.          1.          1.          0.          1.          0.          1.          0.          0.          0.          1.          0.          0.          0.          0.          1.          0.          1.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          1.          0.          0.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          1.          0.          0.          0.          0.          1.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          0.          1.          0.        ]
    [-0.6613017   0.          1.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          1.          0.          1.          0.          1.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          0.          1.          0.          0.          0.          0.        ]"""

# CLUSTER DATA
scaled_data = StandardScaler().fit_transform(features)
# scaled_data
# [[-1.3161452  -0.4082483   0.40824828  2.4494898   2.4494898   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483   1.581139    1.581139  -0.4082483   1.1547004  -0.4082483   1.581139   -0.4082483  -0.4082483  -0.6324555   1.581139   -0.4082483  -0.4082483  -0.6324555  -0.4082483   2.4494898  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.6324556  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -1.1547006   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483   1.581139   -0.4082483 ]
#  [-0.66130173 -0.4082483   0.40824828 -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.6324556   1.581139  -0.4082483   1.1547004  -0.4082483   1.581139   -0.4082483  -0.4082483  -0.6324555   1.581139   -0.4082483  -0.4082483  -0.6324555  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.6324556  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -1.1547006  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483  -0.4082483   2.4494898  -0.4082483  -0.4082483  -0.6324556  -0.4082483 ]

我该怎么做才能改善我的模型？

这是我完整的代码（包括注释），以获取上面的情节：

# -*- coding: utf-8 -*-

# Main
import os
import simplejson as json
import random
import processors
import tokenizers
import analyzers
import clusterers
import numpy as np
from sklearn.feature_extraction import DictVectorizer
from sklearn import svm, preprocessing, cross_validation
from sklearn.metrics import precision_recall_curve, auc, classification_report, precision_recall_fscore_support
import collections

# Processor
from sklearn import preprocessing

# DBSCAN
from sklearn import cluster
from sklearn.preprocessing import StandardScaler
import numpy as np


class Processor(object):

    CONTINUOUS_FEATURES = {
    'width': lambda page, datapoint: float(datapoint['bound']['width']),}

    def __init__(self, data):

        self.data = data
        self.pages = []
        self.texts = []

        for page in self.data:
            for text in page['texts']:
                self.pages.append(page)
                self.texts.append(text)

    def extract(self):

        continuous_features = []
        discrete_features = []

        for page, text in zip(self.pages, self.texts):

            continuous_features.append([process(page, text) for key, process in self.CONTINUOUS_FEATURES.iteritems()])

            discrete_feature = dict(text['computed'].items())
            discrete_feature['path'] = ' > '.join(text['path'])
            discrete_features.append(discrete_feature)

        return continuous_features, discrete_features



def load_data(file):
    with open(file) as f:
        data = json.load(f)
    return data


def main():

    data = [{'body': {'scroll': {'top': 0, 'left': 0}, 'bound': {'width': 3983, 'top': 0, 'height': 1526, 'left': 0}}, 'texts': [{'computed': {'font-size': '15px', 'text-decoration-color': 'rgb(0, 0, 0)', 'color': 'rgb(0, 0, 0)', 'transform-origin': '15px 13px', 'margin-right': '10px', 'border-left-color': 'rgb(0, 0, 0)', 'background-repeat': 'no-repeat', 'caret-color': 'rgb(0, 0, 0)', 'border-top-color': 'rgb(0, 0, 0)', 'background-color': 'rgba(0, 0, 0, 0)', 'border-bottom-color': 'rgb(0, 0, 0)', 'outline-color': 'rgb(0, 0, 0)', 'border-right-color': 'rgb(0, 0, 0)', 'text-emphasis-color': 'rgb(0, 0, 0)', 'text-indent': '-9999px', 'unicode-bidi': 'normal', 'text-shadow': 'rgb(0, 0, 0) 0px 0px 0px', 'font-family': 'FuturaLight', 'background-image': 'url("file:///C:/Users/ronaldg/Documents/_Beauty/data/sites/adorebeauty/images/head/heart-icon.svg")', 'perspective-origin': '15px 13px', 'line-height': '20.25px', 'cursor': 'pointer', 'display': 'inline-block', 'column-rule-color': 'rgb(0, 0, 0)'}, 'text': ['Wishlist'], 'bound': {'width': 30, 'top': 30, 'height': 26, 'left': 2305.60009765625}, 'selector': [{'classes': ['wrapper'], 'id': '', 'name': 'div'}, {'classes': ['page'], 'id': '', 'name': 'div'}, {'classes': ['mage-header'], 'id': '', 'name': 'div'}, {'classes': [], 'id': 'header', 'name': 'header'}, {'classes': ['header-section'], 'id': '', 'name': 'div'}, {'classes': ['header-right-block'], 'id': '', 'name': 'div'}, {'classes': ['header-account'], 'id': 'header-account', 'name': 'div'}, {'classes': ['header-wishlist'], 'id': '', 'name': 'a'}], 'html': 'Wishlist', 'path': ['div', 'div', 'div', 'header', 'div', 'div', 'div', 'a'], 'element': {'classes': ['header-wishlist'], 'id': '', 'name': 'a'}}, {'computed': {'font-size': '15px', 'perspective-origin': '72.7px 15px', 'transform-origin': '72.7px 15px', 'display': 'inline-block', 'padding-top': '5px', 'font-family': 'FuturaLight', 'line-height': '20.25px', 'background-color': 'rgba(0, 0, 0, 0)'}, 'text': ['Sign in', ' | ', 'Register'], 'bound': {'width': 145.39999389648438, 'top': 25, 'height': 30, 'left': 2303.60009765625}, 'selector': [{'classes': ['wrapper'], 'id': '', 'name': 'div'}, {'classes': ['page'], 'id': '', 'name': 'div'}, {'classes': ['mage-header'], 'id': '', 'name': 'div'}, {'classes': [], 'id': 'header', 'name': 'header'}, {'classes': ['header-section'], 'id': '', 'name': 'div'}, {'classes': ['header-right-block'], 'id': '', 'name': 'div'}, {'classes': ['header-account'], 'id': 'header-account', 'name': 'div'}], 'html': '\n                <!-- -->\n            <a href="https://www.adorebeauty.com.au/wishlist/" rel="nofollow" class="header-wishlist" style="border: 1px solid red;">Wishlist</a><a href="https://www.adorebeauty.com.au/customer/account/login/" rel="nofollow" class="login">Sign in</a> | <a href="https://www.adorebeauty.com.au/customer/account/create/" rel="nofollow">Register</a>', 'path': ['div', 'div', 'div', 'header', 'div', 'div', 'div'], 'element': {'classes': ['header-account'], 'id': 'header-account', 'name': 'div'}}, {'computed': {'border-top-style': 'solid', 'font-size': '14px', 'text-decoration-color': 'rgb(255, 255, 255)', 'color': 'rgb(255, 255, 255)', 'letter-spacing': '1px', 'transform-origin': '95.0833px 22.5px', 'padding-bottom': '12px', 'padding-top': '12px', 'border-top-width': '1px', 'border-left-color': 'rgba(0, 0, 0, 0)', 'border-right-style': 'solid', 'padding-right': '18px', 'border-left-style': 'solid', 'caret-color': 'rgb(255, 255, 255)', 'border-top-color': 'rgba(0, 0, 0, 0)', 'background-color': 'rgba(0, 0, 0, 0)', 'border-bottom-color': 'rgb(255, 255, 255)', 'outline-color': 'rgb(255, 255, 255)', 'border-right-color': 'rgba(0, 0, 0, 0)', 'text-emphasis-color': 'rgb(255, 255, 255)', 'unicode-bidi': 'normal', 'text-shadow': 'rgb(255, 255, 255) 0px 0px 0px', 'list-style-type': 'none', 'font-family': 'FuturaLight', 'text-align': 'left', 'perspective-origin': '95.0833px 22.5px', 'cursor': 'pointer', 'border-right-width': '1px', 'column-rule-color': 'rgb(255, 255, 255)', 'text-transform': 'uppercase', 'line-height': '20px', 'border-left-width': '1px', 'padding-left': '18px'}, 'text': ['Shop By Category'], 'bound': {'width': 190.1666717529297, 'top': 80, 'height': 45, 'left': 1499}, 'selector': [{'classes': ['wrapper'], 'id': '', 'name': 'div'}, {'classes': ['page'], 'id': '', 'name': 'div'}, {'classes': ['nav-head'], 'id': '', 'name': 'nav'}, {'classes': ['top-nav'], 'id': 'top-nav', 'name': 'ul'}, {'classes': ['cat-item', 'top'], 'id': '', 'name': 'li'}, {'classes': [], 'id': '', 'name': 'a'}], 'html': 'Shop By Category', 'path': ['div', 'div', 'nav', 'ul', 'li', 'a'], 'element': {'classes': [], 'id': '', 'name': 'a'}}, {'computed': {'font-size': '16px', 'text-decoration-color': 'rgb(20, 179, 88)', 'color': 'rgb(20, 179, 88)', 'transform-origin': '270px 25.5333px', 'padding-bottom': '10px', 'padding-top': '10px', 'border-left-color': 'rgb(20, 179, 88)', 'margin-bottom': '28px', 'padding-right': '10px', 'caret-color': 'rgb(20, 179, 88)', 'border-top-color': 'rgb(20, 179, 88)', 'background-color': 'rgb(234, 248, 248)', 'border-bottom-color': 'rgb(20, 179, 88)', 'outline-color': 'rgb(20, 179, 88)', 'border-right-color': 'rgb(20, 179, 88)', 'text-emphasis-color': 'rgb(20, 179, 88)', 'text-shadow': 'rgb(20, 179, 88) 0px 0px 0px', 'perspective-origin': '270px 25.5333px', 'margin-top': '22px', 'line-height': '21.6px', 'column-rule-color': 'rgb(20, 179, 88)', 'padding-left': '10px'}, 'text': [u'\u2714\ufe0e ', 'In Stock.', '\n                                                                                    We ship today if you order before ', '3 am'], 'bound': {'width': 540, 'top': 479.9666748046875, 'height': 51.05000305175781, 'left': 1921.5}, 'selector': [{'classes': ['wrapper'], 'id': '', 'name': 'div'}, {'classes': ['page'], 'id': '', 'name': 'div'}, {'classes': ['col1-layout', 'main'], 'id': '', 'name': 'div'}, {'classes': ['col-main'], 'id': '', 'name': 'div'}, {'classes': [], 'id': '', 'name': 'div'}, {'classes': ['product-view'], 'id': '', 'name': 'div'}, {'classes': [], 'id': 'product_addtocart_form', 'name': 'form'}, {'classes': ['product-essential'], 'id': '', 'name': 'div'}, {'classes': ['product-shop'], 'id': 'product-shop', 'name': 'div'}, {'classes': ['add-to-box'], 'id': '', 'name': 'div'}, {'classes': ['is-before', 'new-in-stock'], 'id': '', 'name': 'div'}], 'html': u'\n                                        <span><span class="tick">\u2714\ufe0e </span>In Stock.</span>\n                                                                                    We ship today if you order before <span class="time" data-time="1539262800000">3 am</span>                                                                            ', 'path': ['div', 'div', 'div', 'div', 'div', 'div', 'form', 'div', 'div', 'div', 'div'], 'element': {'classes': ['is-before', 'new-in-stock'], 'id': '', 'name': 'div'}}, {'computed': {'float': 'left', 'transform-origin': '135px 18.5833px', 'perspective-origin': '135px 18.5833px', 'background-color': 'rgba(0, 0, 0, 0)', 'text-align': 'left'}, 'text': ['Qty'], 'bound': {'width': 270, 'top': 561.0166625976562, 'height': 37.15000915527344, 'left': 1921.5}, 'selector': [{'classes': ['wrapper'], 'id': '', 'name': 'div'}, {'classes': ['page'], 'id': '', 'name': 'div'}, {'classes': ['col1-layout', 'main'], 'id': '', 'name': 'div'}, {'classes': ['col-main'], 'id': '', 'name': 'div'}, {'classes': [], 'id': '', 'name': 'div'}, {'classes': ['product-view'], 'id': '', 'name': 'div'}, {'classes': [], 'id': 'product_addtocart_form', 'name': 'form'}, {'classes': ['product-essential'], 'id': '', 'name': 'div'}, {'classes': ['product-shop'], 'id': 'product-shop', 'name': 'div'}, {'classes': ['add-to-box'], 'id': '', 'name': 'div'}, {'classes': ['add-to-cart'], 'id': '', 'name': 'div'}], 'html': '\n\t<label for="qty">Qty</label>\n\t<select name="qty" id="qty" class="hasCustomSelect" style="-webkit-appearance: menulist-button; width: 60px; position: absolute; opacity: 0; height: 36px; font-size: 11px; left: 0px;">\n\t\t<option value="1" selected="">1</option>\n        \t\t<option value="2">2</option>\n        \t\t<option value="3">3</option>\n        \t\t<option value="4">4</option>\n        \t\t<option value="5">5</option>\n        \t\t<option value="6">6</option>\n        \t\t<option value="7">7</option>\n        \t\t<option value="8">8</option>\n        \t\t<option value="9">9</option>\n        \t\t<option value="10">10</option>\n        \t</select><span class="customSelect" style="display: inline-block;"><span class="customSelectInner" style="width: 49px; display: inline-block;">1</span></span>\n\t\t<button type="button" title="Add to Bag" class="button btn-cart"><span><span>Add to Bag</span></span></button>\n\t\t', 'path': ['div', 'div', 'div', 'div', 'div', 'div', 'form', 'div', 'div', 'div', 'div'], 'element': {'classes': ['add-to-cart'], 'id': '', 'name': 'div'}}, {'computed': {'text-decoration-color': 'rgb(102, 102, 102)', 'outline-color': 'rgb(102, 102, 102)', 'border-left-color': 'rgb(102, 102, 102)', 'perspective-origin': '250px 35px', 'color': 'rgb(102, 102, 102)', 'border-right-color': 'rgb(102, 102, 102)', 'text-emphasis-color': 'rgb(102, 102, 102)', 'transform-origin': '250px 35px', 'text-shadow': 'rgb(102, 102, 102) 0px 0px 0px', 'background-color': 'rgba(0, 0, 0, 0)', 'caret-color': 'rgb(102, 102, 102)', 'border-top-color': 'rgb(102, 102, 102)', 'border-bottom-color': 'rgb(102, 102, 102)', 'line-height': '14px', 'column-rule-color': 'rgb(102, 102, 102)', 'text-align': 'left'}, 'text': [u"Skin is visibly restored by morning, as added\xa0Lavender Essential Oil works to soothe inflamed skin and promote an even skin tone,\xa0 Evening Primrose Oil helps to repair skin and Squalane replenishes skin's\xa0moisture barrier, leaving skin feeling soft, supple and moisturised.\xa0This restoring facial serum improves firmness and elasticity while encouraging a radiant, youthful complexion.\xa0"], 'bound': {'width': 500, 'top': 734.1666870117188, 'height': 70, 'left': 1937.5}, 'selector': [{'classes': ['wrapper'], 'id': '', 'name': 'div'}, {'classes': ['page'], 'id': '', 'name': 'div'}, {'classes': ['col1-layout', 'main'], 'id': '', 'name': 'div'}, {'classes': ['col-main'], 'id': '', 'name': 'div'}, {'classes': [], 'id': '', 'name': 'div'}, {'classes': ['product-view'], 'id': '', 'name': 'div'}, {'classes': [], 'id': 'product_addtocart_form', 'name': 'form'}, {'classes': ['product-collateral'], 'id': '', 'name': 'div'}, {'classes': ['collateral-tabs', 'tab-list'], 'id': 'collateral-tabs', 'name': 'dl'}, {'classes': ['tab-container'], 'id': '', 'name': 'dd'}, {'classes': ['jspScrollable', 'tab-content'], 'id': '', 'name': 'div'}, {'classes': ['jspContainer'], 'id': '', 'name': 'div'}, {'classes': ['jspPane'], 'id': '', 'name': 'div'}, {'classes': ['jspContainer'], 'id': '', 'name': 'div'}, {'classes': ['jspPane'], 'id': '', 'name': 'div'}, {'classes': [], 'id': '', 'name': 'p'}], 'html': "Skin is visibly restored by morning, as added&nbsp;Lavender Essential Oil works to soothe inflamed skin and promote an even skin tone,&nbsp; Evening Primrose Oil helps to repair skin and Squalane replenishes skin's&nbsp;moisture barrier, leaving skin feeling soft, supple and moisturised.&nbsp;This restoring facial serum improves firmness and elasticity while encouraging a radiant, youthful complexion.&nbsp;<br><br>", 'path': ['div', 'div', 'div', 'div', 'div', 'div', 'form', 'div', 'dl', 'dd', 'div', 'div', 'div', 'div', 'div', 'p'], 'element': {'classes': [], 'id': '', 'name': 'p'}}, {'computed': {'text-decoration-color': 'rgb(153, 153, 153)', 'outline-color': 'rgb(153, 153, 153)', 'line-height': '14px', 'vertical-align': 'top', 'perspective-origin': '79px 7px', 'color': 'rgb(153, 153, 153)', 'border-right-color': 'rgb(153, 153, 153)', 'text-emphasis-color': 'rgb(153, 153, 153)', 'transform-origin': '79px 7px', 'text-shadow': 'rgb(153, 153, 153) 0px 0px 0px', 'background-color': 'rgba(0, 0, 0, 0)', 'border-left-color': 'rgb(153, 153, 153)', 'caret-color': 'rgb(153, 153, 153)', 'list-style-type': 'none', 'border-bottom-color': 'rgb(153, 153, 153)', 'border-top-color': 'rgb(153, 153, 153)', 'column-rule-color': 'rgb(153, 153, 153)', 'text-align': 'left'}, 'text': ['Free over $50'], 'bound': {'width': 158, 'top': 1910.75, 'height': 14, 'left': 1995.5}, 'selector': [{'classes': ['wrapper'], 'id': '', 'name': 'div'}, {'classes': ['page'], 'id': '', 'name': 'div'}, {'classes': ['footer-container'], 'id': '', 'name': 'div'}, {'classes': ['footer'], 'id': '', 'name': 'div'}, {'classes': ['footer-links-icons'], 'id': '', 'name': 'div'}, {'classes': ['footer-links'], 'id': '', 'name': 'div'}, {'classes': [], 'id': '', 'name': 'ul'}, {'classes': [], 'id': '', 'name': 'li'}], 'html': 'Free over $50', 'path': ['div', 'div', 'div', 'div', 'div', 'div', 'ul', 'li'], 'element': {'classes': [], 'id': '', 'name': 'li'}}]}]

    # PROCESS DATA
    processor = Processor(data)
    raw_continuous_features, raw_discrete_features = processor.extract()

    # ENCODE
    continuous_features = np.array(raw_continuous_features)
    scaled_continuous_features = preprocessing.scale(continuous_features)

    DV = DictVectorizer()

    discrete_features = DV.fit_transform(raw_discrete_features).toarray()

    features = np.hstack([continuous_features, discrete_features]).astype(np.float32)

    # CLUSTER DATA
    data = StandardScaler().fit_transform(features)
    db = cluster.DBSCAN(eps=0.5, min_samples=1).fit(data)


    ############################### DBSCAN PLOT DEMO/EXAMPLE ###############################

    from sklearn import metrics

    core_samples_mask = np.zeros_like(db.labels_, dtype=bool)
    core_samples_mask[db.core_sample_indices_] = True
    labels = db.labels_

    # Number of clusters in labels, ignoring noise if present.
    n_clusters_ = len(set(labels)) - (1 if -1 in labels else 0)
    n_noise_ = list(labels).count(-1)

    import matplotlib.pyplot as plt

    # Black removed and is used for noise instead.
    unique_labels = set(labels)
    colors = [plt.cm.Spectral(each)
              for each in np.linspace(0, 1, len(unique_labels))]
    for k, col in zip(unique_labels, colors):
        if k == -1:
            # Black used for noise.
            col = [0, 0, 0, 1]

        class_member_mask = (labels == k)

        xy = data[class_member_mask & core_samples_mask]
        plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(col),
                 markeredgecolor='k', markersize=14)

        xy = data[class_member_mask & ~core_samples_mask]
        plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(col),
                 markeredgecolor='k', markersize=6)

    plt.title('Estimated number of clusters: %d' % n_clusters_)
    plt.show()


if __name__ == '__main__':
    main()

感谢所有帮助/提示/指针。

Answer 1

您实际上并没有集群。拥有与数据点一样多的簇，您只拥有原始数据...对于只有7个样本的数据，DBSCAN并没有多大意义-那里没有“密集”的物体。

但是您的实际问题是关于标准缩放器。

如果将类别属性编码为0或1个二进制变量，然后应用标准缩放器，则0将成为某个负值，而1将是一个正值（通常是不同的）。

在您的情况下，只有一个点具有该特定值。

这说明了为什么整个一键编码和标准缩放方法实际上是一个非常糟糕的技巧。在DBSCAN上使用分类数据的正确方法是A）定义在此数据上定义的距离-无需将数据转换为矢量-或B）定义适当的邻居谓词，如通用DBSCAN后续论文中所述。额外的控制权。

同时使用DBSCAN对离散和连续特征进行聚类-确认设置正确吗？

1 个答案: