| 摘要: | 在鄭皓元(2018)的研究中，建立了一個能自動化分析與評估單一交易策略的平台。該平台具有能針對單一策略與多商品進行自動化運算，且運算過程中能透過多電腦式分散運算來減少運算時間，並根據視覺化圖表呈現分析結果等特性。
但該平台對於多策略多商品的運算分析，並沒有正式比較的模型與框架，且對於濾網組合在多策略中的表現也未提及。還有在運算過程中產生的中繼績效檔案沒有統一儲存與管理的結構化設計，在擴增為多策略分析的視覺化分析圖表的方式上也較不易等缺陷。
因此本研究將自動化交易平台移至容易新增的互動式平台(jupyter)來進行策略的運算與分析圖表的呈現，如此便可利用物件化的方式實現多分析圖表的擴充，透過這些圖表，能輔助呈現單一策略或者多策略的分析結果，以利後續比較，也能從中得出濾網組合的效益。最後，透過模組將分散至各電腦的績效檔案進行集中與再分配，來實現績效檔案統一管理的目的。
研究中在原有的單一交易策略(布林通道)後，再新增2支交易策略(均線策略&凱特通道)，並使用1690檔商品當作資料集，來驗證平台能否進行多商品多策略的運算分析。在驗證的同時，我們也能從中比較出交易策略的優劣，以及找出哪些濾網組合可以適用於多個交易策略，也能從平台產出的資料中找出能夠穩定獲利的商品有哪些。
本研究發現凱特通道的交易策略對於多商品具有「高獲利、高風險」的特性，為三支交易策略中表現最優異的策略。而在濾網組合的方面，「交易量小於1千萬不買、成交量超過前五日平均量的2~15倍不買、股價變動率小於0.5%~1%不買」的組合，是所有濾網組合中表現最為優異的。本研究也找出了284檔商品具有良好的獲利性，能在這三支交易策略上穩定獲利。
;In the study of Hao-Yuan Zheng (2018), a platform was established to automate the analysis and evaluation of a single trading strategy. The platform has the ability to automate operations for a single strategy and multi-portfolio, and the computational process can reduce computation time through multi-computer decentralized operations and present analysis results based on visual charts.
However, the platform does not have a formal comparison model and framework for the computational analysis of multi-strategy and multi-portfolio, and the performance of the filter combination in multi-strategy is not mentioned. There is also a structured design in which the relay performance files generated during the calculation process are not uniformly stored and managed, and it is not easy to be attenuated in the way of amplifying the visual analysis charts for multi-strategy analysis.
Therefore, this study moves the automated trading platform to an easy-to-add interactive platform (jupyter) for strategy calculation and analysis of the chart, so that the object analysis can be used to expand the multi-analysis chart. Auxiliary presentation of single strategy or multi-strategy analysis results, in order to facilitate subsequent comparisons, can also derive the benefits of the filter combination. Finally, through the module, the performance files distributed to each computer are centralized and redistributed to achieve the purpose of unified management of performance files.
In the study, after the original single trading strategy (Bolling Band), two additional trading strategies (Moving Average strategy & Keltner Channel) were added, and 1690 items of stocks were used as data sets to verify whether the platform can carry out the operational analysis of multi-portfolio and multi-strategy. At the same time of verification, we can also compare the advantages and disadvantages of the trading strategy, and find out which filter combinations can be applied to multiple trading strategies, and also find out which portfolio can be stably profited from the data produced by the platform. .
This study found that Keltner Channel′s trading strategy has the characteristics of “high profitability and high risk” for multi-portfolio, and is the best performing strategy among the three trading strategies. In the aspect of the filter combination, the combination of "the transaction volume is less than 10 million, the transaction volume is more than 2 to 15 times the average of the previous five days, and the stock price change rate is less than 0.5% to 1%." is the best performance in the combination. The study also found that 284 portfolio have good profitability and can be stable in these three trading strategies. |
