Research On The Intelligent Stock Selection Plan Of The Biomedical Sector Based On The Stacking Fusion Model

The biomedical industry is a national key strategic emerging industry,is a sunrise industry.In recent years,the annual compound growth rate of new drugs in the world is 8%,and the United States,Europe and other developed countries account for a large proportion of the number of drugs in research.Since our country’s biomedicine has been included in the key development plan of the 13 th Five-Year Plan,it has ushered in a relatively broad space for growth.Moreover,the growing frequency of investment and financing activities and the huge demand for innovative drugs from pharmaceutical companies will fuel the heat of M&A activity in this sector.The establishment of the mature US stock market,the rapid development of Hong Kong’s biotechnology board and China’s science and technology board all highlight the investment price of the biomedical industry value and investment opportunities.In recent years,with the gradual development of artificial intelligence,the way of integrating intelligent algorithm into stock selection strategy is increasingly accepted and used by investors because of its high stability,large capital capacity and considerable income.therefore,this paper,with the help of machine learning algorithm,selects the data of 356 enterprises in the global biomedical sector from2011 to 2018,and constructs a stock selection strategy based on the stacking framework of xgboost-rf-lightgbm fusion model.Firstly,this paper constructs the index database based on the characteristics of biomedical industry,uses the XGboost algorithm to eliminate the invalid index of the output result and the correlation thermal map of the feature importance,and makes the missing value of the data and so on.The experimental results show that the growth index and characteristic index of biomedical industry play an important role in the later strategy construction,which illustrates that the plate focuses on growth and R&D characteristics.Subsequently,the XGboost and random forest models are constructed in the first layer of Stacking.The LightGBM model is constructed in the second layer,and the three are combined to get the final stock selection strategy.The output index of this paper is to predict theprobability of rising,the standard of screening stocks is to predict the probability of the top 60 stocks to hold positions,a year to change positions.Finally,the paper evaluates and tests the stock selection strategy,and compares the model with several models.The results show that the average yield of stock selection strategy based on Stacking framework is 17.78%,8.99% higher than the average yield of industry benchmark,Sharpe’s ratio is 0.29,and the maximum retreat is 25.21%.On the whole,this strategy can obtain the excess return which far exceeds the industry benchmark,and its maximum retreat is relatively reasonable,so the model is effective.Therefore,compared with other models,the fusion model constructed in this paper has good classification ability,and its stability is higher,which can improve the accuracy and stability of stock selection,and can help investors to obtain stable excess income in this section.