Traditional machine learning classification model predicts the rise and fall of stock prices under more than 40 indicators

Preface

The stock market is surging , Only a good forecast of the stock price can we better seize the profit opportunity . So what is the effect of the traditional machine learning classification model in this respect ？

This paper considers the moving average 、 Exponential moving average 、DMI、MACD、KDJ、RSI etc. More than 40 indicators Under the circumstances , Comparison of the Support vector machine 、 Decision tree 、 Random forests Three models predict the effect of stock price fluctuation

disclaimer

Nothing in this vision and analysis should be interpreted as investment advice , Past performance does not necessarily indicate future results .

---

We choose Maotai stock (600519) As an object of study , The data date range is 2020-01-01 To 2020-12-30. Uniform adoption of the last 55 Day data as test set , Last 64 The data of days ago is used as the training set

---

Support vector machine

Support vector machine （SVM） It is a kind of generalized linear classifier which classifies data according to supervised learning , The decision boundary is the maximum margin hyperplane for learning samples

In order to increase the nonlinearity of the model , use RBF Kernel function , The effect of the model is as follows
Compared with the previous article using only six indicators , from 42.47% The accuracy of is improved to 49.9%.

---

Decision tree

Decision tree (Decision Tree） On the basis of knowing the probability of occurrence of various situations , The probability that the expected value of NPV is greater than or equal to zero is calculated by constructing decision tree , Evaluate project risks , The decision analysis method to judge its feasibility , It is a graphic method of intuitively using probability analysis . Because this kind of decision branch is like a tree's branch , So it's called decision tree . In machine learning , Decision tree is a prediction model , It represents a mapping relationship between object attributes and object values

The depth of the defined number is 5, The effect of the model is as follows
Compared with the previous article, which only used six indicators , Overall, there is still some improvement
The accuracy on the training set ranges from 77.53% Greatly increased to 97.66%
The accuracy on the test set ranges from 56.25% Improved a little 56.36%
But the recall rate from 53.53% To improve the 56.35%

---

Random forests

Random forest is a classifier that contains multiple decision trees , And the output category is determined by the mode of the output category of the individual tree

Define the depth of the tree as 5, The number of random trees is 10, The effect of the model is as follows
Compared with the previous article, which only used six indicators , Overall, there is still some improvement
The accuracy on the training set ranges from 86.52% Greatly increased to 100%
The accuracy on the test set ranges from 57.81% Improved a little 58.18%
But the recall rate from 56.76% To improve the 58.40%

---

summary

After a simple test , The effect of random forest is relatively optimal

But there are still big problems in the prediction effect , The improvement effect is very small , Explain the need for more appropriate machine learning methods