# Python-Semester-New

**Repository Path**: skyonedot/Python-Semester-New

## Basic Information

- **Project Name**: Python-Semester-New
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2020-07-11
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# 基于增量学习的变权重组合基分类器处理二分类问题

## 建模

### 基分类器

---

#### Logistic



#### Linear SVM



#### BernoulliNB



#### Perceptron



#### PassiveAggressiveClassifier



---

### 计算公式

### Calculate_Weight 

###### Parameter:  

![](https://gitee.com/song_tianyi/picture-host/raw/master/ndata//20200709142103.png)

###### Initialize:

![](https://gitee.com/song_tianyi/picture-host/raw/master/ndata//20200709142137.png)

###### Update rule

![](https://gitee.com/song_tianyi/picture-host/raw/master/ndata//20200709142158.png)

---

## 数据

### 方圆数据集

来源戳这里:point_right: [来源](http://sofasofa.io/competitions/6/data.zip)

展示如下

![](https://gitee.com/song_tianyi/picture-host/raw/master/ndata//20200709142332.png)

---

### 手写数字数据集

来源为Python内置数据集

展示如下

![](https://gitee.com/song_tianyi/picture-host/raw/master/ndata//20200709142426.png)

---

### Fashion-Mnist数据集

来源戳这里:point_right: [数据](https://storage.googleapis.com/kaggle-data-sets/2243%2F9243%2Fcompressed%2Ffashion-mnist_train.csv.zip?GoogleAccessId=gcp-kaggle-com@kaggle-161607.iam.gserviceaccount.com&Expires=1594535247&Signature=CBRCM%2B4JW6kJhk9x5FxZUIfmfCjb72B6RBx3%2FYNvnOE0QNgTkAMiSAoGca9%2Bn1YCPHJzpUHzlHCpKL0he3XF4%2FBaPxPVirwtFbtBS12rEwm5gAF0xsc2GOlb3%2F9qms1%2BB6eL0VTfdUJiRgY3IYL%2BDe4qv1O14mKj%2BtMnzt9l%2FxKZbspSKJaduqzDeK5p%2B7YQPPTtTk28z4ALXrR7wpCW6NBX%2BGUrwSf%2FgVKMtIVNKJEkvXoDoFG0vRfpi5tHvrAq2TpnR2qVRMjNdqiLr70v1zjdkIxEl7AEZsBokg3co1jT3JCYpMJ2DaWnh7nZLy4NCnBciBa5W95FfQG8q1Oi9Q%3D%3D)

数据展示如下

![](https://gitee.com/song_tianyi/picture-host/raw/master/ndata//20200709142501.png)

---

## 实际运行

:watermelon:   用到了三个数据集,其中手写数字和Fashion-Mnist数据集虽然是多分类数据集,但是可以单独吧ylabel为不同的		两类提取出来,因此也可以做很多二分类;

:strawberry:  Fashion-Mnist跑了两个二分类,手写数字数据集跑了一个;

:apple:  yita值得大小,对结果的影响颇大;

:bear:  综合四个model来看,组合分类器要略微优秀于基分类器.