[b]（4）支持向量机（SVM）大数据算法操作实践。[/b][br][br] ① 作业目的。[br][br] 旨在让学生了解硬间隔支持向量机、软间隔支持向量机及非线性支持向量机的算法含义及应用场景，了解四类不同核函数，即 Linear Kernel、Polynomial Kernel、Gaussian Kernel和 Sigmoid Kernel 对学习机（Learner）性状的影响，体会其中的异同点，从而加深对 Orange平台中各种支持向量机对分类功能的实现。[br][br] ② 作业准备。[br][br] [color=#0000ff][b][url=https://orangedatamining.com/download/]Orange3 软件下载[/url][icon]/images/ggb/toolbar/mode_zoomin.png[/icon][/b][/color]并安装。[br] [br] [color=#0000ff][b][url=https://pan.baidu.com/disk/main#/index?category=all&path=%2FSVM%2FSVM%E6%93%8D%E4%BD%9C%E5%AE%9E%E8%B7%B5]源数据[/url][icon]/images/ggb/toolbar/mode_zoomin.png[/icon][/b][/color]包含三个文件，adult-data.txt（训练集）、adult-test.txt（测试集）、adult-attribute.txt（数据来源及属性说明）。

[br]| This data was extracted from the census bureau database found at [br][br]| http://www.census.gov/ftp/pub/DES/www/welcome.html [br][br]| Split into train-test using MLC++ GenCVFiles （2/3，1/3 random）. [br][br]| 48842 instances，mix of continuous and discrete （train=32561，test=16281）[br][br]| 45222 if instances with unknown values are removed （train=30162，test=15060）[br][br]| Duplicate or conflicting instances : 6 [br][br]| Class probabilities for adult.all file [br][br]| Probability for the label '>50K' : 23.93% / 24.78% （without unknowns）[br][br]| Probability for the label '<=50K' : 76.07% / 75.22% （without unknowns）[br][br]| [br][br]| Extraction was done by Barry Becker from the 1994 Census database. A set of| reasonably clean [br][br]records was extracted using the following conditions: （（AAGE>16） && （AGI>100） && （AFNLWGT>1）[br][br]&& （HRSWK>0））. [br][br]|| Prediction task is to determine whether a person makes over 50K| a year. [br][br]| [br][br]| C4.5 : 84.46+-0.30 [br][br]| Naive-Bayes : 83.88+-0.30 [br][br]| NBTree : 85.90+-0.28 [br][br]| [br][br]|| Following algorithms were later run with the following error rates，all after removal of unknowns [br][br]and using the original train/test split. All these numbers are straight runs using MLC++ with default [br][br]values. [br][br]| [br][br]| Algorithm Error [br][br]| -- ---------------- ----- [br] [br]| 1 C4.5 15.54[br] [br][br]| 2 C4.5-auto 14.46 [br][br]| 3 C4.5 rules 14.94 [br][br]| 4 Voted ID3 （0.6） 15.64 [br][br]| 5 Voted ID3 （0.8） 16.47 [br][br]| 6 T2 16.84 [br][br]| 7 1R 19.54 [br][br]| 8 NBTree 14.10 [br][br]| 9 CN2 16.00 [br][br]| 10 HOODG 14.82 [br][br]| 11 FSS Naive Bayes 14.05 [br][br]| 12 IDTM （Decision table） 14.46 [br][br]| 13 Naive-Bayes 16.12 [br][br]| 14 Nearest-neighbor （1） 21.42 [br][br]| 15 Nearest-neighbor （3） 20.35 [br][br]| 16 OC1 15.04 [br][br]| Description of fnlwgt （final weight）:| The weights on the CPS files are controlled to independent estimates of the| civilian noninstitutional population of the US. These are prepared monthly| for us by Population Division here at the Census Bureau. We use 3 sets of| controls. [br][br]| These are: [br][br]| 1. A single cell estimate of the population 16+ for each state. [br][br]| 2. Controls for Hispanic Origin by age and sex. [br][br]| 3. Controls by Race，age and sex. [br][br]age: continuous. [br][br]workclass: Private，Self-emp-not-inc，Self-emp-inc，Federal-gov，Local-gov，State-gov，Without-pay，Never-worked. [br][br]fnlwgt: continuous. [br][br]education: Bachelors，Some-college，11th，HS-grad，Prof-school，Assoc-acdm，Assoc-voc，9th，7th-8th，12th，Masters，1st-4th，10th，Doctorate，5th-6th，Preschool. [br][br]education-num: continuous. [br][br]marital-status: Married-civ-spouse，Divorced，Never-married，Separated，Widowed，Married-spouse-absent，Married-AF-spouse. [br][br]occupation: Tech-support，Craft-repair，Other-service，Sales，Exec-managerial，Prof-specialty，Handlers-cleaners，Machine-op-inspct，Adm-clerical，Farming-fishing，Transport-moving，Priv-house-serv，Protective-serv，Armed-Forces. [br][br]relationship: Wife，Own-child，Husband，Not-in-family，Other-relative，Unmarried. [br][br]race: White，Asian-Pac-Islander，Amer-Indian-Eskimo，Other，Black.[br][br]sex: Female，Male. [br][br]capital-gain: continuous. [br][br]capital-loss: continuous. [br][br]hours-per-week: continuous. [br][br]native-country: United-States，Cambodia，England，Puerto-Rico，Canada，Germany，Outlying-US（Guam-USVI-etc），India，Japan，Greece，South，China，Cuba，Iran，Honduras，Philippines，Italy，Poland，Jamaica，Vietnam，Mexico，Portugal，Ireland，France，Dominican-Republic，Laos，Ecuador，Taiwan，Haiti，Columbia，Hungary，Guatemala，Nicaragua，Scotland，Thailand，Yugoslavia，El-Salvador，Trinadad&Tobago，Peru，Hong，Holand-Netherlands.

a. 数据源分析。[br][br] 首先分析数据源的数据出处，属于开源的机器学习数据库网站。[br][br] b. 数据属性及数据配置解析。[br][br] 数据配置：[br][br] Split into train-test using MLC++ GenCVFiles （2/3，1/3 random）. [br][br] 48842 instances，mix of continuous and discrete （train=32561，test=16281）[br][br] 45222 if instances with unknown values are removed （train=30162，test=15060）[br][br] Class probabilities for adult.all file [br][br] Probability for the label '>50K' ：23.93% / 24.78% （without unknowns）[br][br] Probability for the label '<=50K' ：76.07% / 75.22% （without unknowns）

[br]特征属性（共 15 个）：[br][br]age：continuous. [br][br]workclass：Private，Self-emp-not-inc，Self-emp-inc，Federal-gov，Local-gov，State-gov，[br]Without-pay，Never-worked. [br][br]fnlwgt：continuous. [br][br]education：Bachelors，Some-college，11th，HS-grad，Prof-school，Assoc-acdm，Assoc-voc，9th，7th-8th，12th，Masters，1st-4th，10th，Doctorate，5th-6th，Preschool. [br][br]education-num：continuous. [br][br]marital-status：Married-civ-spouse，Divorced，Never-married，Separated，Widowed，Married-spouse-absent，Married-AF-spouse. [br][br]occupation：Tech-support，Craft-repair，Other-service，Sales，Exec-managerial，Prof-specialty，Handlers-cleaners， Machine-op-inspct， Adm-clerical， Farming-fishing， Transport-moving，Priv-house-serv，Protective-serv，Armed-Forces. [br][br]relationship：Wife，Own-child，Husband，Not-in-family，Other-relative，Unmarried. [br][br]race：White，Asian-Pac-Islander，Amer-Indian-Eskimo，Other，Black. [br][br]sex：Female，Male. [br][br]capital-gain：continuous. [br][br]capital-loss：continuous. [br][br]hours-per-week：continuous. [br][br]native-country：United-States，Cambodia，England，Puerto-Rico，Canada，Germany，Outlying-US（Guam-USVI-etc），India，Japan，Greece，South，China，Cuba，Iran，Honduras，Philippines ， Italy ， Poland ， Jamaica ， Vietnam ， Mexico ， Portugal ， Ireland ， France ，Dominican-Republic，Laos，Ecuador，Taiwan，Haiti，Columbia，Hungary，Guatemala，Nicaragua，Scotland，Thailand，Yugoslavia，El-Salvador，Trinadad&Tobago，Peru，Hong，Holand-Netherlands.

③ 作业内容。[br][br] 作业包括三个部分：[br][br] ● 数据整理及转换；[br][br] ● Orange 平台上机操作；[br][br] ● 撰写分析报告。[br][br] a. 数据整理与转换。[br][br] 一般来讲，下载数据采用的是 txt 文件，而 txt 是一种纯文本文档，里面不会有任何字体[br]格式，直观性较差，同时也不便于 Orange 平台操作，因此需要进行转换并预处理。[br][br] ● 在 Excel 中打开 txt 文件。[br][br] 要求：创建训练集及测试集两个 Excel 数据集，文件名自定。[br][br] ● 预处理数据。[br][br] 要求：加标题表头，通过筛选批量删除含有“？”字符的记录。[br][br] b. Orange 平台上机操作。[br][br] 总要求是对四个核函数分别建立学习器，并比较各学习器的优劣。工作流完整，逻辑清[br]晰，产出合理。关键内容如下：[br][br] ▶ 设置四个核函数的学习器；[br][br] ▶ 训练集及测试集部署合理；[br][br] ▶ 调整惩罚项及参数设置，调优学习器；[br][br] ▶ 数据集在线端配属正确，不报错；[br][br] ▶ 调用可视化模块，对支持向量进行展示。[br][center][img]https://s21.ax1x.com/2025/02/20/pEQ3JfA.png[/img][/center][br][br] c. 撰写数据分析报告。[br] [br] 以上[color=#0000ff][b][url=https://pan.baidu.com/disk/main#/index?category=all&path=%2F%E6%95%B0%E6%8D%AE%E5%88%86%E6%9E%90%E6%8A%A5%E5%91%8A%E6%A8%A1%E6%9D%BF]资源下载[/url][icon]/images/ggb/toolbar/mode_zoomin.png[/icon][/b][/color]。[br]