基于不同的技术指标进行量化分析，选择买入卖出点实现交易策略，实现稳健收益。
use time-train to train to select selected_code, and other period with the same time-length as each train-window to valid. eg:
train(2014-2017,2015-2018,2016-2019) - valid(2017-2020)
train(2015-2018,2016-2019,2017-2020) - valid(2018-2021)
train(2016-2019,2017-2020,2018-2021) - valid(2019-2022)
train(2017-2020,2018-2021,2019-2022) - valid(2020-2023)
1. Training pool: Initial-Codes
Each code should be trained epoch = [end-(start+batch+batch_length)+1] times.
2. Validation pool: After-Each-Training-Codes
During each Train-Valid process,
if (SUM(Train profit > 0)) is true, then the code was selected into Validation pool.
3. Winner pool: Winner-Codes
if (All(Train-Valid profit > 0)) is true, then the code was selected into Winner pool.