def summary_release_2_docx(title, img_meta_dic_list, stg_run_id=None, enable_clean_cache=True):
"""
生成 预测成功率趋势报告
:param title:
:param img_meta_dic_list:
:param stg_run_id:
:param enable_clean_cache:
:return:
"""
logger.debug('生成报告开始')
# 生成 docx 文件
document = docx.Document()
# 设置默认字体
document.styles['Normal'].font.name = '微软雅黑'
document.styles['Normal']._element.rPr.rFonts.set(docx.oxml.ns.qn('w:eastAsia'), '微软雅黑')
# 创建自定义段落样式(第一个参数为样式名, 第二个参数为样式类型, 1为段落样式, 2为字符样式, 3为表格样式)
UserStyle1 = document.styles.add_style('UserStyle1', 1)
# 设置字体尺寸
UserStyle1.font.size = docx.shared.Pt(40)
# 设置字体颜色
UserStyle1.font.color.rgb = docx.shared.RGBColor(0xff, 0xde, 0x00)
# 居中文本
UserStyle1.paragraph_format.alignment = docx.enum.text.WD_ALIGN_PARAGRAPH.CENTER
# 设置中文字体
UserStyle1.font.name = '微软雅黑'
UserStyle1._element.rPr.rFonts.set(docx.oxml.ns.qn('w:eastAsia'), '微软雅黑')
# 文件内容
document.add_heading(title, 0).alignment = docx.enum.text.WD_ALIGN_PARAGRAPH.CENTER
document.add_paragraph('')
document.add_paragraph('')
heading_size = 1
for num, info_dic in enumerate(img_meta_dic_list, start=1):
trade_date_last_train = info_dic['trade_date_last_train']
trade_date_end = info_dic['trade_date_end']
document.add_heading(
f"{num}、{date_2_str(trade_date_last_train)} - {date_2_str(trade_date_end)}", heading_size)
split_point_list = info_dic['split_point_list']
if split_point_list is None:
p = document.add_paragraph(f"{num}.1） 日期区间段1个:\n")
p.add_run(f'\t1) {date_2_str(trade_date_last_train)} ~ {date_2_str(trade_date_end)}\n')
else:
p = document.add_paragraph(f"{num}.1） 日期区间段{len(split_point_list) - 1}个:\n")
for num2, (point1, point2) in enumerate(
iter_2_range(split_point_list, has_left_outer=False, has_right_outer=False), start=1):
p.add_run(f'\t{num2}) {date_2_str(point1)} ~ {date_2_str(point2)}\n')
document.add_paragraph(f"{num}.2） 模型路径：\n\t{info_dic['module_file_path']}")
document.add_paragraph(f"{num}.3） 取样状态(random_state)：\n\t{info_dic['predict_test_random_state']}")
document.add_paragraph(f"{num}.4） 展示数据长度：\n\t{info_dic['in_range_count']}")
document.add_paragraph(f"{num}.5） 预测准确率趋势图：")
document.add_picture(info_dic['img_file_path'])
file_name = f"{title}.docx"
file_path = os.path.join(get_report_folder_path(stg_run_id), file_name)
document.save(file_path)
if enable_clean_cache:
clean_cache()
logger.debug('生成报告结束。%s', file_path)
return file_path
def summary_md_2_docx(
md_df: pd.DataFrame,
percentiles=[0.2, 0.33, 0.5, 0.66, 0.8],
risk_free=0.03,
close_key=None,
enable_show_plot=True,
enable_save_plot=False,
name=None,
func_kwargs_dic={},
enable_clean_cache=True,
):
"""
汇总展示数据分析结果，同时以 dict 形式返回各项指标分析结果
第一个返回值，df的各项分析结果
第二个返回值，各个列的各项分析结果
:param md_df:
:param percentiles:分为数信息
:param risk_free:无风险收益率
:param close_key:对哪些列的数据执行统计
:param enable_show_plot:显示plot
:param enable_save_plot:保存plot
:param name:
:param func_kwargs_dic:
:param enable_clean_cache:
:return:
"""
ret_dic, each_col_dic, file_path_dic = summary_md(
md_df,
percentiles=percentiles,
risk_free=risk_free,
close_key=close_key,
enable_show_plot=enable_show_plot,
enable_save_plot=enable_save_plot,
name=name,
func_kwargs_dic=func_kwargs_dic)
logger.debug('file_path_dic')
for num, (k, v) in enumerate(file_path_dic.items(), start=1):
if isinstance(v, dict):
for num2, (k2, v2) in enumerate(v.items(), start=1):
if isinstance(v2, dict):
for num3, (k3, v3) in enumerate(v2.items(), start=1):
logger.debug("%d.%d.%d) %s %s %s -> %s", num, num2,
num3, k, k2, k3, v3)
else:
logger.debug("%d.%d) %s %s -> %s", num, num2, k, k2, v2)
else:
logger.debug("%d) %s -> %s", num, k, v)
# 生成 docx 文档将所需变量
heading_title = f'数据分析报告 {date_2_str(min(md_df.index))} - {date_2_str(max(md_df.index))} ({md_df.shape[0]} days)'
# 生成 docx 文件
document = docx.Document()
# 设置默认字体
document.styles['Normal'].font.name = '微软雅黑'
document.styles['Normal']._element.rPr.rFonts.set(
docx.oxml.ns.qn('w:eastAsia'), '微软雅黑')
# 创建自定义段落样式(第一个参数为样式名, 第二个参数为样式类型, 1为段落样式, 2为字符样式, 3为表格样式)
UserStyle1 = document.styles.add_style('UserStyle1', 1)
# 设置字体尺寸
UserStyle1.font.size = docx.shared.Pt(40)
# 设置字体颜色
UserStyle1.font.color.rgb = docx.shared.RGBColor(0xff, 0xde, 0x00)
# 居中文本
UserStyle1.paragraph_format.alignment = WD_ALIGN_PARAGRAPH.CENTER
# 设置中文字体
UserStyle1.font.name = '微软雅黑'
UserStyle1._element.rPr.rFonts.set(docx.oxml.ns.qn('w:eastAsia'), '微软雅黑')
# 文件内容
document.add_heading(heading_title,
0).alignment = WD_ALIGN_PARAGRAPH.CENTER
document.add_paragraph('')
document.add_paragraph('')
heading_count = 1
if 'warning' in ret_dic:
document.add_heading(f'{heading_count}、警告信息（Warning）', 1)
warning_list = ret_dic['warning']
p = document.add_paragraph('')
for msg in warning_list:
r = p.add_run(msg)
# r.bold = True
r.font.color.rgb = RGBColor(0xaf, 0x26, 0x26)
p.add_run('\n')
heading_count += 1
document.add_page_break()
if 'rr' in file_path_dic:
document.add_heading(f'{heading_count}、行情曲线', 1)
# 增加图片（此处使用相对位置）
document.add_picture(
file_path_dic['rr']) # , width=docx.shared.Inches(1.25)
heading_count += 1
# 添加分页符
document.add_page_break()
if 'hist' in file_path_dic:
document.add_heading(f'{heading_count}、Histgram 分布图', 1)
document.add_picture(file_path_dic['hist'])
heading_count += 1
if 'rr_quantile' in ret_dic:
document.add_heading(f'{heading_count}、分位数信息（Quantile）', 1)
rr_quantile_df = ret_dic['rr_quantile']
format_by_col = {_: FORMAT_2_PERCENT for _ in rr_quantile_df.columns}
df_2_table(document,
rr_quantile_df,
format_by_col=format_by_col,
max_col_count=5)
heading_count += 1
document.add_page_break()
if 'hist_future_n_rr' in file_path_dic:
document.add_heading(f'{heading_count}、未来N日收益率最高最低值分布图', 1)
quantile_dic = ret_dic['hist_future_n_rr']['quantile_dic']
for num, ((n_day, col_name), file_path) in enumerate(
file_path_dic['hist_future_n_rr'].items(), start=1):
document.add_heading(
f'{heading_count}.{num}) 未来 {n_day} 日 {col_name} 收益率最高最低值分布图',
2)
document.add_picture(file_path)
document.add_heading(f'{heading_count}.{num}.1) 分位数信息', 3)
data_df = quantile_dic[(n_day, col_name)]
df_2_table(
document,
data_df,
format_by_index={_: FORMAT_2_PERCENT
for _ in data_df.index})
document.add_heading(f'{heading_count}.{num}.2) 三分类标签分布比例', 3)
for (min_pct, max_pct
), distribution_rate_df in ret_dic['label_distribution'][(
n_day, col_name)].items():
file_path = file_path_dic['label_distribution'][(
n_day, col_name)][(min_pct, max_pct)]
document.add_picture(file_path)
distribution_rate_df.rename(columns={
1: f'1 under {min_pct * 100:.2f}%',
2: f'2 over {max_pct * 100:.2f}%'
},
inplace=True)
df_2_table(document,
distribution_rate_df,
format_by_index={
_: FORMAT_2_PERCENT
for _ in distribution_rate_df.index
})
document.add_page_break()
heading_count += 1
if 'drawdown' in file_path_dic:
document.add_heading(f'{heading_count}、行情回撤曲线', 1)
document.add_picture(file_path_dic['drawdown'])
heading_count += 1
document.add_page_break()
if 'scatter_matrix' in file_path_dic:
document.add_heading(f'{heading_count}、散点图矩阵图（Scatter Matrix）', 1)
document.add_picture(file_path_dic['scatter_matrix'])
heading_count += 1
document.add_page_break()
if 'correlation' in file_path_dic:
document.add_heading(f'{heading_count}、相关性矩阵图（Correlation）', 1)
document.add_picture(file_path_dic['correlation'])
heading_count += 1
document.add_page_break()
if 'stats' in file_path_dic:
document.add_heading(f'{heading_count}、绩效统计数据（Porformance stat）', 1)
stats_df = ret_dic['stats'].stats
stats_df_2_docx_table(stats_df, document)
heading_count += 1
document.add_page_break()
# 逐列分析
col_name = close_key
# 文件内容
heading_title = f'{col_name} 数据分析结果'
document.add_heading(heading_title,
0).alignment = WD_ALIGN_PARAGRAPH.CENTER
document.add_paragraph('')
document.add_paragraph('')
heading_count = 1
if 'rr_quantile' in ret_dic:
document.add_heading(f'{heading_count}、分位数信息（Quantile）', 1)
data_df = each_col_dic[col_name]['rr_quantile']
format_by_col = {_: FORMAT_2_PERCENT for _ in data_df.columns}
df_2_table(document,
data_df,
format_by_col=format_by_col,
max_col_count=5)
heading_count += 1
if f'{col_name} hist' in file_path_dic:
document.add_heading(f'{heading_count}、Histgram 分布图', 1)
document.add_picture(file_path_dic[f'{col_name} hist'])
heading_count += 1
document.add_page_break()
# 保存文件
file_name = f"MD{' ' if name is None else ' ' + name} " \
f"{date_2_str(min(md_df.index))} - {date_2_str(max(md_df.index))} " \
f"({md_df.shape[0]} days) {datetime_2_str(datetime.datetime.now(), STR_FORMAT_DATETIME_4_FILE_NAME)}.docx"
file_path = os.path.join(get_report_folder_path(), file_name)
document.save(file_path)
if enable_clean_cache:
clean_cache()
return file_path
def summary_stg_2_docx(stg_run_id=None,
enable_save_plot=True,
enable_show_plot=False,
enable_clean_cache=True,
doc_file_path=None):
"""
生成策略分析报告
:param stg_run_id:
:param enable_save_plot:
:param enable_show_plot:
:param enable_clean_cache:
:param doc_file_path: 可以是目录 或 文件名路径
:return:
"""
info = get_stg_run_info(stg_run_id)
stg_run_id = info.stg_run_id
stg_name = info.stg_name
run_mode = RunMode(info.run_mode)
kwargs = {
"enable_show_plot": enable_show_plot,
"enable_save_plot": enable_save_plot,
"run_mode": run_mode,
"stg_run_id": stg_run_id
}
if run_mode == RunMode.Backtest_FixPercent:
sum_df, symbol_rr_dic = get_rr_with_md(stg_run_id, compound_rr=True)
else:
sum_df, symbol_rr_dic = get_rr_with_md(stg_run_id, compound_rr=False)
if sum_df is None or sum_df.shape[0] == 0:
logger.warning('stg_run_id=%d 没有获取到 sum_df', stg_run_id)
file_path = None
return file_path
ret_dic, each_col_dic, file_path_dic = summary_rr(sum_df, **kwargs)
_, file_path = show_trade(**kwargs)
file_path_dic['trade'] = file_path
df, file_path = show_cash_and_margin(**kwargs)
file_path_dic['cash_and_margin'] = file_path
# 生成 docx 文档将所需变量
heading_title = f'策略分析报告[{stg_run_id} {stg_name}] ' \
f'{date_2_str(min(sum_df.index))} - {date_2_str(max(sum_df.index))} ({sum_df.shape[0]} days)'
# 生成 docx 文件
document = docx.Document()
# 设置默认字体
document.styles['Normal'].font.name = '微软雅黑'
document.styles['Normal']._element.rPr.rFonts.set(
docx.oxml.ns.qn('w:eastAsia'), '微软雅黑')
# 创建自定义段落样式(第一个参数为样式名, 第二个参数为样式类型, 1为段落样式, 2为字符样式, 3为表格样式)
UserStyle1 = document.styles.add_style('UserStyle1', 1)
# 设置字体尺寸
UserStyle1.font.size = docx.shared.Pt(40)
# 设置字体颜色
UserStyle1.font.color.rgb = docx.shared.RGBColor(0xff, 0xde, 0x00)
# 居中文本
UserStyle1.paragraph_format.alignment = WD_ALIGN_PARAGRAPH.CENTER
# 设置中文字体
UserStyle1.font.name = '微软雅黑'
UserStyle1._element.rPr.rFonts.set(docx.oxml.ns.qn('w:eastAsia'), '微软雅黑')
# 文件内容
document.add_heading(heading_title,
0).alignment = WD_ALIGN_PARAGRAPH.CENTER
document.add_paragraph('')
document.add_paragraph('')
heading_count = 1
document.add_heading(f'{heading_count}、策略回测收益曲线', 1)
# 增加图片（此处使用相对位置）
document.add_picture(
file_path_dic['rr']) # , width=docx.shared.Inches(1.25)
heading_count += 1
# 添加分页符
document.add_page_break()
document.add_heading(f'{heading_count}、策略回撤曲线', 1)
document.add_picture(file_path_dic['drawdown'])
heading_count += 1
document.add_page_break()
if run_mode != RunMode.Backtest_FixPercent:
document.add_heading(f'{heading_count}、现金与仓位堆叠图', 1)
document.add_picture(file_path_dic['cash_and_margin'])
heading_count += 1
document.add_page_break()
document.add_heading(f'{heading_count}、散点图矩阵图（Scatter Matrix）', 1)
document.add_picture(file_path_dic['scatter_matrix'])
heading_count += 1
document.add_page_break()
document.add_heading(f'{heading_count}、相关性矩阵图（Correlation）', 1)
document.add_picture(file_path_dic['correlation'])
heading_count += 1
document.add_page_break()
document.add_heading(f'{heading_count}、绩效统计数据（Porformance stat）', 1)
stats_df = ret_dic['stats'].stats
stats_df_2_docx_table(stats_df, document)
heading_count += 1
document.add_page_break()
# 交易记录
document.add_heading(f'{heading_count}、买卖点记录', 1)
document.add_picture(file_path_dic['trade'])
heading_count += 1
document.add_page_break()
# 保存文件
if doc_file_path is not None:
if os.path.isdir(doc_file_path):
folder_path, file_name = doc_file_path, ''
else:
folder_path, file_name = os.path.split(doc_file_path)
else:
folder_path, file_name = get_report_folder_path(stg_run_id), ''
if folder_path != '' and not os.path.exists(folder_path):
os.makedirs(folder_path)
if file_name == '':
try:
calc_mode_str = CalcMode(
json.loads(
info.trade_agent_params_list)[0]['calc_mode']).name + " "
except:
calc_mode_str = " "
# run_mode_str = run_mode.name + " "
file_name = f"{stg_run_id} {stg_name} {calc_mode_str}" \
f"{date_2_str(min(sum_df.index))} - {date_2_str(max(sum_df.index))} ({sum_df.shape[0]} days) " \
f"{datetime_2_str(datetime.datetime.now(), STR_FORMAT_DATETIME_4_FILE_NAME)}.docx"
file_path = os.path.join(folder_path, file_name)
else:
file_path = doc_file_path
document.save(file_path)
if enable_clean_cache:
clean_cache()
return file_path