Research On Quantification Strategies Using Filtering Techniques And Machine Learning

Quantitative investment is a new type of investment,although the existence of a short time,but quantitative investment with its strict discipline and efficient execution efficiency is favored and pursued by the majority of investors.Under the framework of quantitative investment,this paper makes an in-depth study of quantitative timing and quantitative stock selection strategies.Different from the general strategies constructed by financial,statistical and nonlinear theories,this paper adopts communication filtering and Hilbert transform to design the timing strategy on the one hand,and machine learning to quantify the classic investment experience on the other hand,so as to make it a repeatable rule.Finally,the quantitative timing and quantitative stock selection strategies designed in the broad platform using the real historical data of A shares to test and verify.The study of strategy in this paper mainly includes two aspects:(1)A stock price sequence can be considered to be composed of three parts:long-term trend item,short-term fluctuation item and noise item.In the research of timing strategy,starting from the moving average index in trend following technology,it proves that the application of filter to stock market is reasonable.Then,EMA index which is more consistent with the realistic significance is adopted to construct a low-pass filter to filter the noise items,and then the trend items are removed by difference.Finally Hilbert transform is carried out on the remaining fluctuation part and the in-phase orthogonal space is constructed to construct trading signals using periodic fluctuations.In the end we test the validity of the timing strategy through Ping An Bank,China Merchants Bank and ZTE.(2)Gather Bidding(before opening)is the first round of the day long short fight,is the first time to buy and sell stocks in each trading day.Some experienced investors can make a profit by watching the rally auction to predict the trend of stocks.In order to make these experiences can form the quantitative rule,this article a strategy to standardization of information call auction,using PCA technology will be the original data from 29 d drop to 8 d,multiple linear regression,fitting the same day opening price,using the value investment idea,by the difference between the predicted and actual values,looking for undervalued stocks,the warehouse every ten days.Finally,the A-share market on December 31,2012 and June 28,2019 was selected for verification:the strategy had A considerable return before 2018,but from February to May 2018,the return declined due to the impact of market and regulatory policies.