HDU Course: AI Intro - Assignment

• 训练数据集：[Breast Cancer Wisconsin (Diagnostic) Data Set] (https://archive.ics.uci.edu/ml/datasets/Breast+Cancer+Wisconsin+%28Diagnostic%29)

Data Analysis

• RangeIndex: 569 entries, 0 to 568

• Data Columns: total 32 columns

• Data Types: float64(30), int64(1), object(1)

• Memory Usage: 142.3+ KB

• Attribute Info：

  1. 身份证号码
  2. 诊断（M =恶性，B =良性）
  3. 3-32为每个细胞核计算十个实值特征：

     • a）半径（从中心到周边点的距离的平均值）

     • b）纹理（灰度值的标准偏差）

     • c）周界

     • d）区域

     • e）光滑度（半径长度的局部变化）

     • f）紧凑性（周长^ 2 /面积 - 1.0）

     • g）凹度（轮廓凹部的严重程度）

     • h）凹点（轮廓的凹入部分的数量）

     • i）对称

     • j）分形维数（“海岸线近似” - 1）

     • 对每个数据分别求平均值，标准误差，“最差”或最大，产生30个特征。

     • 所有功能值都用四位有效数字重新编码。

Data Visualization

• 相似相关性热力图

• 加入分类的相关性热力图

  • New Correlation Features:
    • 32 to 16 dimension
    • concavity_mean, concave points_mean, concave points_worst
    • compactness_mean, compactness_worst, concavity_worst
    • perimeter_mean, radius_mean, area_mean, perimeter_worst, radius_worst, area_worst
    • area_se, radius_se, perimeter_se
    • texture_mean, texture_worst
    • texture_se, smoothness_se, symmetry_se
    • fractal_dimesion_worst
    • smoothness_mean
    • smoothness_worst
    • symmetry_mean
    • symmetry_worst
    • concave points_se
    • compactness_se
    • concavity_se
    • fractal_dimession_mean
    • fractal_dimession_se

Classification

Diagnosis	Amount
B	268
M	158

32 features

• Logistic Regression

  • Original

  Item	Precision	Recall	f1-Score	Support
  B	0.94624	0.98876	0.96703	89
  M	0.98000	0.90741	0.94231	54
  Avg / Total	0.95899	0.95804	0.95770	143

  • Depth Optimized

  Accuracy is maximum at depth 100 and accuracy 0.979

  Item	Precision	Recall	f1-Score	Support
  B	0.96739	1.00000	0.98343	89
  M	1.00000	0.94444	0.97143	54
  Avg / Total	0.97971	0.97902	0.97890	143

• kNN

Accuracy is max in Sample 10 and accuracy 0.9301

Item	Precision	Recall	f1-Score	Support
B	0.92473	0.96629	0.94505	89
M	0.94000	0.87037	0.90385	54
Avg / Total	0.93050	0.93007	0.92949	143

• Gaussian Naive Bayes

Item	Precision	Recall	f1-Score	Support
B	0.92632	0.98876	0.95652	89
M	0.97917	0.87037	0.92157	54
Avg / Total	0.94627	0.94406	0.94332	143

16 features

• Drop Attributes:

  • concave points_mean
  • concave points_worst
  • compactness_worst
  • concavity_worst
  • radius_mean
  • area_mean
  • perimeter_worst
  • radius_worst
  • area_worst
  • radius_se
  • perimeter_se
  • texture_worst
  • smoothness_se
  • symmetry_se

• Logistic Regression

Accuracy is maximum at depth 1000 and accuracy 0.986

Item	Precision	Recall	f1-Score	Support
B	0.97802	1.00000	0.98889	89
M	1.00000	0.96296	0.98113	54
Avg / Total	0.98632	0.98601	0.98596	143

• kNN

Accuracy is max in Sample 5 and accuracy 0.9161

Item	Precision	Recall	f1-Score	Support
B	0.91398	0.95506	0.93407	89
M	0.92000	0.85185	0.88462	54
Avg / Total	0.91625	0.91608	0.91539	143

• Random Forest

Item	Precision	Recall	f1-Score	Support
B	0.94624	0.98876	0.96703	89
M	0.98000	0.90741	0.94231	54
Avg / Total	0.95899	0.95804	0.95770	143

Dimension Reduction* (Unfinished)

• PCA
  • Image: K-means clustering on the digits dataset (PCA-reduced data)