SVM
所属分类:数值算法/人工智能
开发工具:Python
文件大小:464KB
下载次数:17
上传日期:2018-11-09 14:20:33
上 传 者:
你好20182019
说明: 回归预测 股票预测 用到支持向量机的相关预测项目
(Regression forecast stock forecasting related forecasting project using support vector machine)
文件列表:
GYN.json (4932, 2017-01-21)
IASP520_ARMA_V1.01.py (4017, 2017-01-21)
IASP520_SVM.py (2381, 2017-01-21)
IASP520_SVM_V2.py (2970, 2017-01-21)
IASP520_SVM_V2_With_Polarity_Json.py (3775, 2017-01-21)
NQN.json (1706, 2017-01-21)
pic (0, 2017-01-21)
pic\ARIMA_EX.png (186338, 2017-01-21)
pic\ARIMA_V2.0.png (154056, 2017-01-21)
pic\SVM_V2.0.png (135022, 2017-01-21)
StockPrediction
=========
Stock data come from Yahoo_finance by Python.
News data come from tm.plugin by R.
ARMIA
===
Step
---
1.Use Daubechies 4 wavelet to transform the Stock Data which comes from Yahoo_finance.
2.Difference the time series make it stationary.
3.Create ACF & Pacf pictures to find out p & q which is the parameter in ARIMA.
4.Predict the stationary time series by ARIMA(p,q). Because this ARIMA package can't do difference bigger than 2, thus I don't use ARIMA(p,d,q).
5.Revert difference which we do in step 2.
| ARIMA |
|:----------------------------------:|
| ![Conclusion](pic/ARIMA_EX.png) |
SVM
===
Not good enough. I try to transform Price to the relation, like the relation between Open & Close attributes or today & yesterday.
Stock can't be Predicted only based on history stock data, so we pull in new data. It's still not good but much better than before.
| SVM |
|:----------------------------------:|
| ![Conclusion](pic/SVM_V2.0.png) |
近期下载者:
相关文件:
收藏者: