Forecasting prices of financial assets using Novel SGWO-SVR Method (Sanitized Grey Wolf Optimizer):

*Copyright© Bhaskar Tripathi,2020

This work uses "Grey Wolf Optimizer" to forecast the day-ahead prices of a financial asset (like Ethereum cryptocurrency). The Predictor variable is "US dollar exchange price" of Ethereum's blockchain. The Hybrid algorithm demonstrates the modelling of a complex case study for a chaotic dataset exhibiting properties like high-dimensionality, multimodality, non-uniformity and non-linearity. Nature inspired models are much reliable than their Machine Learning and other optimization counterparts like Gradient based methods.

"The GWO algorithm mimics the leadership hierarchy and hunting mechanism of grey wolves in nature. Four types of grey wolves such as alpha, beta, delta, and omega are employed for simulating the leadership hierarchy. In addition, three main steps of hunting, searching for prey, encircling prey, and attacking prey, are implemented to perform optimization"[1]

• Sanitized GWO: My algorithm, named "Sanitized GWO" uses Chaos Theory to generate butterfly effect in GWO. This is done by introducing Time varying accelration coefficients for simulating the locations of Alpha, Beta and Delta wolves. This improves the balance between exploration and exploitation by emulating a naturally occuring reflex action process for the wolves. See this video for more details on Chaos theory: (https://www.youtube.com/watch?v=r_5shyQGIeA).

• Source: - https://www.sciencedirect.com/science/article/pii/S0965997813001853 [1]

Advanced econometric models like ARCH, GARH, ARIMA etc and deterministic models like Gradient Descent methods in Machine Learning, fail to address the aforementioned challanges associated with a noisy dataset. Therefore, using a stochastic model like GWO is much desired. A Support Vector Regressor(SVR) as an objective function with the intent of minimizing (or maximizing) its fitness value is used to optimize the Cost (C) and Gamma(γ) hyper-parameters of SVR. Once the SVR is trained by GWO, it can be used to forecast the prices of Ethereum.

The Hybrid Methodology is then compared with other State-of-the-art benchmark optimization methodologies like Grid Search, PSO, Ant Colony Method etc. The results indicate a superior performance of this model over all other comparable models on evaluation measures of RMSE, MSE, MAPE, R-Square.

[2]

• Usage: Simple replace your dataset, change the filename and treat the first column as "Predictor variable". rest of the columns will automatically be treated as independent variables.

• Credits:

• [1]: The project was inspired by Dr Seyedali Mirjalili, the original creator of GWO method and one of leaders in the area of metaheuristics. I convereted his Matlab code to Python and created a Sanitized Model that has been generalized on 36 benchmark functions and a chaotic Cryptocurrency dataset. I introduced Chaos for reflex positions of the wolves along with an objective function of SVR. The method is a Novel usage in the area of predicting financial securities. It is hoped to be of immense help to Intraday, interday and short term traders in the stock markets. Seyedali Mirjalili (2020). Grey Wolf Optimizer (GWO) (https://www.mathworks.com/matlabcentral/fileexchange/44974-grey-wolf-optimizer-gwo), MATLAB Central File Exchange. Retrieved May 28, 2020.

• [2]:. SVR Image taken for illustration from : https://jcheminf.biomedcentral.com/articles/10.1186/1758-2946-5-33

**If you like the project, Please leave a star to show your appreciation.