Work detail

Stock Market Trend Prediction Using Genetic Algorithms

Author: Mgr. Jaromír Malenko
Year: 2008 - winter
Leaders:
Consultants:
Work type: Finance, Financial Markets and Banking
Masters
Language: English
Pages: 55
Awards and prizes:
Link:
Abstract: In this thesis the genetic algorithm is considered as a method of computer learning and it is applied to stock market returns predicting. In the theoretical part we define the prediction task and summarize the results on stock market predictability. We present the statistical methods and measures to evaluate the accuracy of predictor and related computer learning approaches. We introduce the genetic algorithm as a method able to evolve an arbitrary algorithm. In the implementation part a tree representation of any algorithm is employed, the related genetic operators and user application are implemented. In the application part we first evolve predictor as an arbitrary program –
the black-box predictor. The result does not differ gnificantly from liner model, but the economic measures improved. Next, we optimize the parameters of MACD and RSI technical indicators. Optimized indicators may generate profit on the contrary to unoptimized indicators. We find that genetic algorithm is a general technique that is not straightforward to be applied for prediction, but it is efficient in optimizing parameters of a specific model.

Partners

Deloitte

Sponsors

CRIF
McKinsey
Patria Finance