mrlukekr / automated-trader Goto Github PK

3rd Year University Undergraduate dissertation project, supervised by https://www.nottingham.ac.uk/computerscience/people/thomas.gaertner focusing on using Natural Language Processing, AI Portfolio Optimisation and Machine Learning to produce an Automated Trading Agent. Received 89% overall.

Home Page: https://www.researchgate.net/publication/328838338_Automated_Stock_Market_Trading_Using_Machine_Learning

Java 99.03% HTML 0.05% Shell 0.92%

finance fintech stock-market stock-trading stock-prediction stock-analysis automated-trading portfolio-optimization sentiment-analysis

automated-trader's People

Contributors

Stargazers

Watchers

Forkers

knut0815 sudarsan-sridharan laranea prajakta-collab msaberai nagendrasinh sonisatyam92 mobidian webclinic017

automated-trader's Issues

Data Leakage

Luke,

Your random forest is using a randomized train / test split. Your technical indicators embed prior day information into the calculations, which is leaking data into your randomized splits. Your reference [10] paper has this same issue (https://www.reddit.com/r/algotrading/comments/cv83yh/overfitting).

For time series analysis, you should not use randomized splits, as that is how not data would be received in a real environment.