Bluefissure / ECE568-hw3

Bayesian Curve Fitting Model

A basic Bayesian curve fitting model designed for StockProject.

Requirement

Python 3.6+ with

• codecs
• numpy

You should use pip install codecs and pip install numpy to install these dependencies.

Use

You can read the help information with command python main.py --help.

Examples are like:

• python .\src\main.py --file db/hist/ATVI.csv which
  • uses hist/ATVI.csv as data source
• python .\src\main.py --file db/hist/ATVI.csv -bf gaussian -d 128 which
  • uses hist/ATVI.csv as data source
  • has a 128-degree gaussian function as the basic function used for bayesian curve fitting
• python .\src\main.py --file db/hist/AAPL.csv -bf polynomial -d 64 which
  • uses hist/AAPL.csv as data source
  • has a 64-degree polynomial function as the basic function used for bayesian curve fitting

Performance

• Fitting is okay
• Predicting stock is dreaming
• Gaussian works better than others
• Easily get overfitting

The default command got an MAE(Mean absolute error) as 0.9272843719426256 and MRE(Mean relative error) as 0.00480401596775846.