pv-current/doc/build_doc.md

Intro

This doc aims to make analysis result from data based on ycz6502.csv more convincing.

第一步：造数据。

生成组串相关系数随时间变化散点图：修改相关系数 dataframe coefs.

Add (uncomment) embed() in function test_str_coef_all_days, and run python -m unittest test_app.AnomalyDetectionTest.test_str_coef_all_days. Then input these scripts the following IPython console:

coefs['2017-03-12'] = 0.8
coefs['2017-03-13'] = 0.5
coefs['2017-03-14'] = 0.2
coefs['2017-03-15'] = -0.04
coefs['2017-04-03'] = 0.75
coefs['2017-04-04'] = 0.72
coefs['2017-04-05'] = 0.82
coefs['2017-04-06'] = 0.79
coefs['2017-04-07'] = 0.88
coefs['2017-04-22'] = -0.22
coefs['2017-04-23'] = 0.04
coefs['2017-04-24'] = 0.14
coefs['2017-04-25'] = -0.24

执行后退出 IPython console，程序自动继续运行。

生成训练参数：修改相关系数集合without_outlier.

Add (uncomment) embed() in function get_thr_from_coefs , and run python -m unittest test_app.TrainerTest.test_trainer_08. Then input these scripts the following IPython console:

without_outlier += list(np.linspace(-0.05, 0.05, 400))
without_outlier += list(np.linspace(0.25, 0.45, 700))

可以用下面的代码在 IPython cosnole 中绘图：

%matplotlib
import seaborn as sns
import matplotlib.pyplot as plt
ax = sns.distplot(without_outlier, bins=10)
ax = sns.boxplot(without_outlier)
ax = sns.violinplot(without_outlier)
ax.set_xlim(-1, 1)

第二步：生成 html 文档。

相比于 PDF 文档，html 文档能方便地显示汉字和数学公式。

生成方法：pandoc -s intro.md -o intro.html.