ml3 1.2.7

TechYoung Machine Learning ToolKit

# ML3

Introduction

ML3是TechYoung课程辅助工具包.

## Distribution

Run the following commands to register, build and upload the package to PYPI.

python3 setup.py sdist upload

The home page on PYPI is: https://pypi.org/project/wcc/

Install

sudo pip3 install ml3

---

Usage

After installation, run the following command:

import ml3

Methods:

plot.histplot(data, column_name, **kwargs):

• *data:* dataframe

• *column_name:* column name of dataframe, 例如 “hr_mean”

• *kwargs:* “xmin”, “xmax”

plot.gmmplot(data, column_names, k_range, **kwargs):

• *data:* dataframe

• *column_names:* list of columns name of dataframe, 例如 [“hr_mean”, “hr_std”]

• *k_range:* the range of components (k), 例如 [2, 11] or (2, 11)

• *kwargs:* “xmin”, “xmax”, “ymax”, “bins”

plot.kmeansplot(data, column_names, k_range, **kwargs):

• *data:* dataframe

• *column_names:* list of columns name of dataframe, 例如 [“hr_mean”, “hr_std”]

• *k_range:* the range of clusters (k), 例如 [2, 11] or (2, 11)

• *kwargs:* “xmin”, “xmax” “ymax”, “bins”

plot.metricplot(n_clusters_range, scores, scores2=[], **kwargs):

• *n_clusters_range：* tuple or list of range，例如 (2, 10)

• *scores:* list of score

• *scores:* list of score2 (option)

• *kwargs:* “x_label”, “y_label”

plot.errorbarplot(data, x, y=[], y2=[], **kwargs):

• *data:* dataframe

• *x:* x-axis column name，例如 “ctime”

• *y:* y column name，例如 [“hr_mean”, “hr_std”]

• *y2:* y2 column name，例如 [“br_mean”, “br_std”] (option)

• *kwargs:* “X_LABEL”, “Y_LABEL”, “TITLE”, “LIMIT”

plot.pcaplot(data, column_names, **kwargs):

• *data:* dataframe

• *column_names:* list of columns name of dataframe, 例如 [“hr_mean”, “hr_std”]

• *kwargs:* “n_components”

plot.tsenplot(data, column_names, **kwargs):

• *data:* dataframe

• *column_names:* list of columns name of dataframe, 例如 [“hr_mean”, “hr_std”]

• *kwargs:* “n_components”

plot.kalmanplot(data, column_names, dim_x=2, dim_z=1, x=[], p=[], f=[], q=[], h=[], r=1, **kwargs):

• *data:* dataframe

• *column_names:* list of columns name of rawdata dataframe, 例如 [“hr”]

• *dim_x:* the size of the state vector，状态空间维度

  • 默认为2

• *dim_z:* the size of the measurement vector，观测矩阵维度

  • 默认为1

• *x:* filter state estimate，初始状态预测矩阵

  • 默认为[1, 0.1]，分别为心率和心率变化率

• *p:* covariance matrix，协方差矩阵

  • 默认为[[1, 0.1], [0.1, 1]]，心率变化率和人的心率是一定的关系，根据运动状态或者濒死会有很明显的差别，选择0.1代表有一定关系，但是不关系大

• *q:* process uncertainty/noise，噪声矩阵，此矩阵不能为0，否则数据会异常

  • 默认为[[0.0001, 0], [0, 0.0001]]，因为数据都是在cpu中进行，不会产生噪音

• *r:* measurement uncertainty/noise，测量误差

  • 默认为1，测量误差，医疗器械心率误差规定为+-1

• *h:* measurement function

  • Sometimes certain states are measured, when others are not. For example, the first, third and fifth states of a five-dimensional state vector are measurable, while second and fourth states are not measurable H = [[1, 0, 0, 0, 0], [0, 0, 1, 0, 0], [0, 0, 0, 0, 1]]

• *f:* state transistion matrix，状态转移矩阵

  • 默认为[[1, 0.5], [0, 1]]，此矩阵不能对称，否则会计算异常

• *kwargs:* None

seaborn.boxplot(x, y, **kwargs):

此函数需要ml4进行对原始数据进行窗口化分类

• *x:* the UNIX timestamp list from ml4

• *y:* the data list from ml4

• *kwargs:* “X_LABEL”, “Y_LABEL”, “TITLE”

seaborn.violinplot(x, y, **kwargs):

此函数需要ml4进行对原始数据进行窗口化分类

• *x:* the UNIX timestamp list from ml4

• *y:* the data list from ml4

• *kwargs:* “X_LABEL”, “Y_LABEL”, “TITLE”

Example

import ml3
import ml4
import pandas as pd

data = pd.read_csv("feature.csv")
# histogram
ml3.plot.histplot(data, "hr_mean")
# error bar
ml3.plot.errorbarplot(data, "ctime", ["hr_mean", "hr_std"], ["br_mean", "br_std"])
# single feature
ml3.plot.kmeansplot(data, "hr_mean", (2, 10))
ml3.plot.gmmplot(data, "hr_mean", (2, 10))
# multiple features
ml3.plot.gmmplot(data, ["hr_mean", "hr_std", "br_mean", "br_std", "mo_mean", "mo_std"], (2, 10))
ml3.plot.kmeansplot(data, ["hr_mean", "hr_std", "br_mean", "br_std", "mo_mean", "mo_std"], (2, 10))
# two scores metricplot
scores = [110704, 75304, 60731, 52297, 45675, 41231, 37744, 35247, 33263]
scores2 = [0.05, 0.09, 0.15, 0.2, 0.3, 0.5, 0.6, 0.9, 1]
ml3.plot.metricplot((2, 11), scores, scores2)
# boxplot and violoinplot
x, y = ml4.ml4.getWindowData(data, "ctime", "hr")
timeList = []
for i in x:
    tmp = datetime.fromtimestamp(i)
    timeList.append(tmp.strftime("%H:%M"))
ml3.seaborn.boxplot(timeList, y)
ml3.seaborn.violinplot(timeList, y)


data = pd.read_csv("rawdata.csv")
ml3.plot.kalmanplot(data, ["hr"])

data["log1p"] = np.log1p(data["br_std"])
ml3.plot.kmeansplot(data, ["log1p"], (2, 10), ymax=15, bins=0.01)

Note

版本里的1.2.4是旧的版本。1.2.5和以后的版本是用于函数计算的版本。 1.2.5以及以后版本将去掉wcc自动框架. 目录下的子目录：libwebp-0.4.1-linux-x86-64 需要从网上下载，然后把里面的bin下的gif2webp放到/usr/bin里。这样就可以在wcc里调用了.