Acting optimally in partially observable stochastic domains
AAAI'94 Proceedings of the twelfth national conference on Artificial intelligence (vol. 2)
Exact and approximate algorithms for partially observable markov decision processes
Exact and approximate algorithms for partially observable markov decision processes
Automated Planning: Theory & Practice
Automated Planning: Theory & Practice
Heuristic search value iteration for POMDPs
UAI '04 Proceedings of the 20th conference on Uncertainty in artificial intelligence
Reinforcement learning for optimized trade execution
ICML '06 Proceedings of the 23rd international conference on Machine learning
Technical analysis: the complete resource for financial market technicians
Technical analysis: the complete resource for financial market technicians
Point-based value iteration: an anytime algorithm for POMDPs
IJCAI'03 Proceedings of the 18th international joint conference on Artificial intelligence
Hi-index | 0.00 |
The stock market can be considered a nondeterministic and partially observable domain, because investors never know all information that affects prices and the result of an investment is always uncertain. Technical Analysis methods demand only data that are easily available, i.e. the series of prices and trade volumes, and are then very useful to predict current price trends. Analysts have however to deal with the fact that the indications of these methods are uncertain, having different interpretations. In this work, we assume the hypothesis that an investment context can be modeled as a Partially Observable Markov Decision Process (POMDP) and partial observations of price trends can be provided by Technical Analysis methods. A metamodel is proposed to specify POMDP problems, embedding formal interpretations of Technical Analysis methods. Planning algorithms can then try to create investment policies to maximize profits. Due to the complexity for solving POMDPs, algorithms that generate only approximate solutions have to be applied. Nevertheless, the results obtained by an implemented prototype are promising.